[CP2K-user] [CP2K:14133] MIX_CL_TRUNC + ADMM in cp2k development branch

[Nicholas Winner]

Thanks for the reply. I do restart from GGA; I am also using a range 
separated functional as the hf_admm_utils.F module actually takes the 
coulomb and longrange scales and uses them to activate PBE_HOLE_TRUNC_LR 
and XWPBE sections. 

On the point of the short cutoff radius: there is actually a little 
uncertainty in the ideal cutoff radius. In one sense, we would like to have 
the truncated PBE0 converged w.r.t full 1/r, but there is some literature 
evidence that the PBE0 form should be treated as a short-range interaction, 
and the cutoff should be short. Since the short and long range terms have 
different ideal mixing fractions, if we extend PBE0 cutoff out too far 
(like 6-8Å), then while we are reproducing exact 1/r pretty well, we are 
applying short-range mixing fraction beyond where we should. 

Anyway, that's just an aside if you're interested. Thanks for the 
assistance.

-Nick

On Monday, November 2, 2020 at 2:29:40 AM UTC-8 Augustin Bussy wrote:

> Dear Nicholas,
>
> your XC input looks fine to me, but I have a few remarks. The cutoffs of 
> the truncated Coulomb potential and the long range erf functions are very 
> short. We typically use cutoffs that are around 6 Angstrom, which ensures 
> that results are close enough to a calculation with standard Coulomb 
> potential. Unless you are restarting your calculation from a well converged 
> GGA level calculation, I would refrain from using the SCREEN_ON_INITIAL_P 
> keyword, as the ATOMIC_GUESS starting point might not be good enough for 
> porper screening. Finally, if you are using a range spearated HF 
> potentials, you might want to combine it with a range separated DFT 
> exchange functional as well.
>
> I assume you set up the ADMM section properly and made a sensible choice 
> of primary and auxiliary basis sets. Without your full input file however, 
> this is impossible to say whether there is an issue there.
>
> Hope this helps. Best, 
> Augustin
>
> ----... at googlegroups.com wrote: -----
> To: "cp2k" <c... at googlegroups.com>
> From: "Nicholas Winner" 
> Sent by: c... at googlegroups.com
> Date: 10/28/2020 05:14PM
> Subject: [CP2K:14133] MIX_CL_TRUNC + ADMM in cp2k development branch
>
>
> Hello all,
>
> I need to make use of range separated hybrid functionals for periodic 
> systems, so I have pulled the latest development version of cp2k from 
> github which adds this support to admm module (in hfx_admm_utils.F).
>
> To the best of my knowledge, I have implemented it correctly, but the 
> difference in final energy between PBE0 and MIX_CL_TRUNC is ~7 Ha, which 
> seems rather large. When going from GGA -> PBE0 the energy difference <1 
> Ha. Obviously there should be some change when changing the interaction 
> potential, but this has me considered.
>
> *GGA: * ENERGY| Total FORCE_EVAL ( QS ) energy (a.u.):             
> -983.840366364871670 *PBE0* : ENERGY| Total FORCE_EVAL ( QS ) energy 
> (a.u.):             -983.680119471562080
> *MIX_CL_TRUNC*: ENERGY| Total FORCE_EVAL ( QS ) energy [a.u.]:    
> -974.290021294811368
>
> Here is the XC section I am using to activate MIX_CL_TRUNC with ADMM. Can 
> someone please sanity check and tell me if this is appropriate? 
>
>                 &XC
>
>                         &XC_FUNCTIONAL PBE
>
>                                 &PBE
>
>                                         PARAMETRIZATION ORIG
>
>                                         SCALE_C 1
>
>                                         SCALE_X 0.665 # 1 - alpha - beta
>
>                                 &END PBE
>
>                         &END XC_FUNCTIONAL
>
>                         &HF
>
>                                 FRACTION 1
>
>                                 &SCREENING
>
>                                         EPS_SCHWARZ 1e-07
>
>                                         EPS_SCHWARZ_FORCES 1e-06
>
>                                         SCREEN_ON_INITIAL_P True
>
>                                         SCREEN_P_FORCES True
>
>                                 &END SCREENING
>
>                                 &INTERACTION_POTENTIAL
>
>                                         POTENTIAL_TYPE MIX_CL_TRUNC
>
>                                         CUTOFF_RADIUS 1.8
>
>                                         T_C_G_DATA t_c_g.dat
>
>                                         SCALE_LONGRANGE 0.085 # alpha
>
>                                         SCALE_COULOMB 0.25 # beta
>
>                                         OMEGA 0.94486
>
>                                 &END INTERACTION_POTENTIAL
>
>                                 &LOAD_BALANCE
>
>                                         RANDOMIZE True
>
>                                 &END LOAD_BALANCE
>
>                                 &MEMORY
>
>                                         EPS_STORAGE_SCALING 0.1
>
>                                         MAX_MEMORY 2000
>
>                                 &END MEMORY
>
>                         &END HF
>
>                 &END XC
> Best,
> Nicholas Winner
>
> -- 
> 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.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/cp2k/06e0d703-46a0-4e05-873a-3852d52f8d66n%40googlegroups.com 
> <https://groups.google.com/d/msgid/cp2k/06e0d703-46a0-4e05-873a-3852d52f8d66n%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20201102/a8fa9d5b/attachment.htm>