[CP2K-user] [CP2K:20572] Re: Au pseudopotential and basis set refinement for range separated functional
Vladislav Sláma
slamavl at gmail.com
Thu Aug 15 10:24:00 UTC 2024
Dear Frederick,
Thank you again for your very helpful insight. It starts to come together
now. I've checked the excited state properties for few smaller
bio-molecules with different approaches and also with different software
packages and I was getting reasonable results (with the PBE0
pseudopotential). However, I haven't done the same for the gold and I was
blindly assuming that the pseudopotential will be good as well. I'll repeat
the calculations with the new generated pseudopotential for the specific
range separated functional I use to check the effects.
Thank you also for the links concerning the basis set fitting. I'll take a
look at them.
Best wishes,
Vladislav
Dne čtvrtek 15. srpna 2024 v 11:31:14 UTC+2 uživatel Frederick Stein napsal:
> Dear Vladislav,
> If the basis set is large enough, it does not matter for what functional
> it has been optimized. This does not apply to pseudopotentials. Those
> optimized for PBE0 may work in case of standard elements (HCNO) but not
> necessarily for transition metals. In case of an optimization, they should
> be a reasonable first guess. Try to optimize them and compare the results.
> I have no experience in creating basis sets with the atom code. Please
> consult Jürg Hutter, he is the most experience here regarding basis set and
> pseudopotential optimizations (compare his extensive collection of
> self-optimized basis sets and pseudopotentials
> https://github.com/juerghutter/BASIS and
> https://github.com/juerghutter/GTH). I do have some experience with
> optimizing basis sets using a script provided by CP2K (
> https://github.com/cp2k/cp2k/tree/master/tools/scriptmini) to optimize
> correlation consistent basis sets.
> Best,
> Frederick
>
> Vladislav Sláma schrieb am Donnerstag, 15. August 2024 um 10:50:43 UTC+2:
>
>> Dear Frederick,
>> Thank you very much for your suggestions regarding the pseudopotential. Indeed,
>> I copied the XC section from my testing calculations with the
>> pseudopotentials and basis optimized for the PBE0 without any big changes.
>> I'll try to look more into ATOM%USE_GAUSS_HERMITE and ATOM%GRID_POINTS_GH
>> options to refine the pseudopotential even more. In general do you think
>> that the pseudopotentials and basis optimized for hybrid functionals (for
>> example PBE0) should be also good enough to be used with the range
>> separated hybrid functionals? I did some calculations of the excited states
>> of a small gold nanoparticle functionalized with one photoactive
>> biomolecule (with linker long enough tot to have charge transfer between
>> them) and I see quite large mixing between the gold and molecular
>> excitations. I wanted to be sure that the mixing is the real effect or just
>> an artifact originating from the improper pseudopotential and basis for the
>> range separated functional. This was my original motivation for the fitting.
>> I would like to ask you one more question related to the basis set
>> fitting. Is the correct procedure to fit the GAUSSIAN orbitals and then do
>> the contraction to DZVP basis with some other code, or is it possible to do
>> it directly also in CP2K within the ATOM code and obtain coefficients and
>> exponents of the contracted basis?
>>
>> Thank you again for your insight. Best wishes,
>> Vladislav
>>
>> Dne středa 14. srpna 2024 v 18:03:21 UTC+2 uživatel Frederick Stein
>> napsal:
>>
>>> Dear Vladislav,
>>> Generally, pseudopotentials optimized for hybrid functionals are a good
>>> first guess. I implemented pseudopotential optimization with the
>>> longrange-operator some years ago and optimized it for a few elements.
>>> IIRC, the relevant keywords are here:
>>> - ATOM%USE_GAUSS_HERMITE (better convergence) and ATOM%GRID_POINTS_GH to
>>> tune the accuracy of the longrange operator
>>> - ATOM%EXCHANGE_INTEGRALS requires the default NUMERIC (analytic
>>> integrals are not available for the longrange operator)
>>> - HF%SCREENING and HF%MEMORY are ignored (the sections were just copied
>>> from the original HF section)
>>> - check convergence of EPS_SCF and GRID_POINTS keywords and the ACCURACY
>>> keyword in the POWELL section
>>> - The assignment of orbitals to occupied/semicore/core must be correct
>>> Consult the regtests ATOM/regtest-pseudo/C-rsPBE-* for further reference
>>> (regtests are supposed to run fast).
>>> HTH,
>>> Frederick
>>>
>>> Vladislav Sláma schrieb am Mittwoch, 14. August 2024 um 14:37:11 UTC+2:
>>>
>>>> Hello,
>>>>
>>>> I would like to use wB97XD functional (or some other range-separated
>>>> one) with Au atoms for which I would like to fit pseudopotential and DZVP
>>>> basis set. With pseudopotential I tried to follow what I have found in the
>>>> test folder and started from the PBE pseudopotential parameters.
>>>>
>>>> 1) I would like to ask you if the input for the pseudopotential
>>>> parameters fitting with wB97XD is reasonable, because I’ve never done such
>>>> calculations?
>>>>
>>>> 2) What would be the correct strategy to get the DZVP basis set for the
>>>> Au atoms with fitted pseudopotential? The only calculation I managed to run
>>>> successfully (not sure if it is correct) was to fit the GAUSSIAN basis set.
>>>> However, for the CONTRACTED_GTO calculation always failed, sometimes
>>>> without specifying the error. Could you please suggest me how to do this
>>>> calculation? I’ve never needed to fit my own basis and pseudopotential so
>>>> any suggestion or advice is greatly appreciated.
>>>>
>>>> I’m including the input and output files in the attachment and below
>>>> also the input for the GAUSSIAN basis set generation.
>>>>
>>>> Best wishes and thank you for your help,
>>>>
>>>> Vladislav
>>>>
>>>>
>>>> Input I used for GAUSSIAN basis set fitting:
>>>>
>>>> &GLOBAL
>>>> PROGRAM_NAME ATOM
>>>> &END GLOBAL
>>>> &ATOM
>>>> ELEMENT Au
>>>> RUN_TYPE BASIS_OPTIMIZATION
>>>> ELECTRON_CONFIGURATION CORE 5d10 6s1
>>>> CORE [Xe] 4f14
>>>> MAX_ANGULAR_MOMENTUM 3
>>>> COULOMB_INTEGRALS ANALYTIC
>>>> EXCHANGE_INTEGRALS NUMERIC
>>>> &METHOD
>>>> METHOD_TYPE KOHN-SHAM
>>>> RELATIVISTIC DKH(3)
>>>> &XC
>>>> &XC_FUNCTIONAL
>>>> &HYB_GGA_XC_WB97X_D
>>>> &END HYB_GGA_XC_WB97X_D
>>>> &END XC_FUNCTIONAL
>>>> &HF
>>>> FRACTION 1.000
>>>> &SCREENING
>>>> EPS_SCHWARZ 1.0E-14
>>>> SCREEN_ON_INITIAL_P TRUE
>>>> &END
>>>> &INTERACTION_POTENTIAL
>>>> POTENTIAL_TYPE MIX_CL
>>>> SCALE_COULOMB 0.22036
>>>> SCALE_LONGRANGE 0.77964
>>>> OMEGA 0.20
>>>> &END
>>>> &MEMORY
>>>> MAX_MEMORY 4000
>>>> EPS_STORAGE_SCALING 0.1
>>>> &END
>>>> &END
>>>> &END XC
>>>> &END METHOD
>>>> &OPTIMIZATION
>>>> MAX_ITER 500
>>>> EPS_SCF 1.e-08
>>>> &END
>>>> &PP_BASIS
>>>> NUM_GTO 6 6 6
>>>> S_EXPONENTS 3.73260507 1.83419039 0.80906390 0.34515101
>>>> 0.13836655 0.04967010
>>>> P_EXPONENTS 3.73260507 1.83419039 0.80906390 0.34515101
>>>> 0.13836655 0.04967010
>>>> D_EXPONENTS 3.73260507 1.83419039 0.80906390 0.34515101
>>>> 0.13836655 0.04967010
>>>> EPS_EIGENVALUE 1.E-14
>>>> &END PP_BASIS
>>>> &POTENTIAL
>>>> PSEUDO_TYPE GTH
>>>> >H_POTENTIAL
>>>> 1 0 10 0
>>>> 0.59017106458211 1 11.68962795194189
>>>> 3
>>>> 0.52046766556862 2 2.20132630023183 -1.04609353504732
>>>> 2.87007982647554
>>>> 0.63876105996853 2 0.42941169563709 -0.86977143557004
>>>> 2.07606707053651
>>>> 0.44087154543382 2 -4.71769832312536 0.72776406084227
>>>> -1.72921082494821
>>>> &END
>>>> &END POTENTIAL
>>>> &POWELL
>>>> ACCURACY 1.e-8
>>>> STEP_SIZE 0.3
>>>> &END POWELL
>>>> &END ATOM
>>>>
>>>
--
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+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/07ee05ba-9e27-4460-a278-e9389c3347ecn%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20240815/5e1300cc/attachment.htm>
More information about the CP2K-user
mailing list