[QE-users] epsilon.x and hybrids

Wed Apr 8 22:18:13 CEST 2020

Dear Iurii,
thank you very much for the explanation and for the references.

Best,
Michal

On Wed, 8 Apr 2020 at 21:09, Timrov Iurii <iurii.timrov at epfl.ch> wrote:

> Dear Michal,
>
>
> > BTW would you say that for calculation of the dielectric function of
> not-so-large (whatever this means...) periodic systems it is better to try
> GW with the GWW code instead?
>
>
> I am not familiar with GWW, but as far as I know it can be used only with
> a gamma point (k=0). I do not know what are other limitations, and whether
> you can apply it to compute optimal absorption spectra in solids. You may
> contact the developer Paolo Umari.
>
>
> > Is it why TDDFT with hybrids for periodic systems is missing in QE -
> because GW is better and tractable so why bother with TDDFT?
>
>
> It is a long story, but let me write briefly what I know (please correct
> me if I am wrong). I suggest to check the literature on this topic.
>
>
> GW+BSE is the standard approach to compute optical absorption spectra in
> periodic systems (semiconductors). This is so because the non-local
> interactions are present, and because the dielectric function (r-dependent
> and w-dependent) is present in the formalism which describes the
> dielectric properties of the periodic system.
>
>
> TDDFT in the adiabatic approximation with local/semilocal functionals
> (like LDA/GGA) is not accurate for a description of optical absorption
> spectra of periodic solids (please check the literature). However, TDDFT
> with non-local functionals is an interesting alternative to GW+BSE.
> TDDFT at PBE0 improves considerably w.r.t TDDFT at PBE, but still with just a
> constant (alpha=0.25) in the EXX part it is hard to do a descent job for
> solids (actually alpha is related to 1/epsilon : Phys. Rev. B 89, 195112
> (2014)). Much better is to use TDDFT in combination with rage-separated
> hybrid functionals, where 1/epsilon is not just a constant but a function
> 1/epsilon(r) (with models for epsilon(r)), which better represents the
> dielectric properties of the solid (Phys. Rev. B 93, 235106 (2016), check
> in particular the introduction in this reference about the history of
> various non-local XC kernels).
>
>
> Greetings,
>
> Iurii
>
>
> --
> Dr. Iurii Timrov
> Postdoctoral Researcher
> STI - IMX - THEOS and NCCR - MARVEL
> Swiss Federal Institute of Technology Lausanne (EPFL)
> CH-1015 Lausanne, Switzerland
> +41 21 69 34 881
> http://people.epfl.ch/265334
> ------------------------------
> *From:* users <users-bounces at lists.quantum-espresso.org> on behalf of
> Michal Krompiec <michal.krompiec at gmail.com>
> *Sent:* Wednesday, April 8, 2020 9:40:01 PM
>
> *To:* Quantum ESPRESSO users Forum
> *Subject:* Re: [QE-users] epsilon.x and hybrids
>
> BTW would you say that for calculation of the dielectric function of
> not-so-large (whatever this means...) periodic systems it is better to try
> GW with the GWW code instead? Is it why TDDFT with hybrids for periodic
> systems is missing in QE - because GW is better and tractable so why bother
> with TDDFT?
> Best,
> Michal
>
> On Wed, Apr 8, 2020 at 19:15 Timrov Iurii <iurii.timrov at epfl.ch> wrote:
>
>> Dear All,
>>
>>
>> > Do I remember correctly that epsilon.x also does not take into account
>> the nonlocal pseudopotential contribution?
>>
>>
>> Correct
>>
>>
>> > ...there used to be an option in the  turbo_davidson.x and
>> turbo_lanczos.x codes, namely no_hxc=.true.,  which permits an
>> independent-electron calculation.
>>
>>
>> Correct. By default, the matrix element of the dipole operator is
>> computed (in reciprocal space) via the matrix element of the commutator
>> that Paolo mentioned [H,x] (see Eq.(14) in S. Baroni and R. Resta, Phys.
>> Rev. B 33, 7017 (1986)): but in this case there is a kinetic term and the
>> part coming from the non-local part of a pseudopotential. In epsilon.x,
>> only the kinetic term is present, while in TDDFT codes both terms are
>> present (the second term was implement long ago in the Phonon code to
>> compute the dielectric tensor). But still, in the case of hybrid
>> functionals, in [H,x] there is another term which is missing - the
>> commutator with another non-local potential which is EXX. Stefano Baroni
>> and I developed a way how to compute this missing term several years ago: I
>> implemented it in QE and it worked well, but I never had time to release it
>> and publish some paper about it. But there is a workaround for finite
>> systems: the matrix element of the dipole operator can be computed in real
>> space based on the observation that the charge density of the finite system
>> decays fast outside the finite system, and hence the non-periodicity
>> problem of the dipole operator is no longer a problem. But this trick in
>> real space is not gonna work for periodic systems, because the charge
>> density if non-zero in the whole simulation cell. Thus, the only way to
>> overcome this problem is to implement the missing term for hybrids in [H,x].
>>
>>
>> Greetings,
>>
>> Iurii
>>
>>
>> --
>> Dr. Iurii Timrov
>> Postdoctoral Researcher
>> STI - IMX - THEOS and NCCR - MARVEL
>> Swiss Federal Institute of Technology Lausanne (EPFL)
>> CH-1015 Lausanne, Switzerland
>> +41 21 69 34 881
>> http://people.epfl.ch/265334
>> ------------------------------
>> *From:* users <users-bounces at lists.quantum-espresso.org> on behalf of
>> Giuseppe Mattioli <giuseppe.mattioli at ism.cnr.it>
>> *Sent:* Wednesday, April 8, 2020 7:42:35 PM
>> *To:* Quantum ESPRESSO users Forum
>> *Subject:* Re: [QE-users] epsilon.x and hybrids
>>
>>
>> Dear all
>> I don't want to raise the confusion level, so please correct me if I'm
>> wrong... If you want to calculate a heavily approximate absorption
>> spectrum of a (large and non-symmetrical) periodic system after a
>> ground state hybrid calculation, there used to be an option in the
>> turbo_davidson.x and turbo_lanczos.x codes, namely no_hxc=.true.,
>> which permits an independent-electron calculation. At least, hybrid
>> functionals and Gamma ground states should be properly treated by such
>> codes, resulting in an absorption spectrum compatible with those
>> obtained by using epsilon.x, which, AFAIK, calculates
>> <occupied|r|virtual> contributions to the absorption spectrum.
>> However, I don't know how much this kind of calculation is expensive
>> for large supercells. Of course if you are not interested in
>> absorption, then my suggestion is nonsense...
>> HTH
>> Giuseppe
>>
>> Quoting Lorenzo Paulatto <paulatz at gmail.com>:
>>
>> > Also, epsilon.x cannot use symmetry-reduced grids, which would be a
>> > huge wast of time with hybrids, but you can use open_grid.x after
>> > the pw.x calculation too obtain the full grid and work around this
>> > problem.
>> >
>> > cheers
>> >
>> > On 4/8/20 6:38 PM, Paolo Giannozzi wrote:
>> >> I think epsilon.x assumes that the dipole element of x can be
>> >> computed using [H,x]=p\hbar/m. The exchange potential is nonlocal,
>> >> so its commutator with x will yield an additional term that is not
>> >> accounted for. Not sure how important it is in practice. Do I
>> >> remember correctly that epsilon.x also does not take into account
>> >> the nonlocal pseudopotential contribution?
>> >>
>> >> Paolo
>> >>
>> >> On Wed, Apr 8, 2020 at 4:29 PM Manu Hegde <mhegde at sfu.ca
>> >> <mailto:mhegde at sfu.ca <mhegde at sfu.ca>>> wrote:
>> >>
>> >>    Hi Michal,
>> >>    Yes, it is possible.I did use both supercell and hybrid
>> >>    calculations. It did work.
>> >>    Manu
>> >>
>> >>    On Wed, Apr 8, 2020 at 10:08 AM Michal Krompiec
>> >>    <michal.krompiec at gmail.com <mailto:michal.krompiec at gmail.com
>> <michal.krompiec at gmail.com>>> wrote:
>> >>
>> >>        Hello,
>> >>        Is it possible to use epsilon.x on results of a calculation
>> with a
>> >>        hybrid functional (supercell, gamma point only)?
>> >>
>> >>        Thanks,
>> >>
>> >>        Michal Krompiec
>> >>        Merck KGaA
>> >>        _______________________________________________
>> >>        Quantum ESPRESSO is supported by MaX
>> >>        (www.max-centre.eu/quantum-espresso
>> >>        <http://www.max-centre.eu/quantum-espresso>)
>> >>        users mailing list users at lists.quantum-espresso.org
>> >>        <mailto:users at lists.quantum-espresso.org
>> <users at lists.quantum-espresso.org>>
>> >>        https://lists.quantum-espresso.org/mailman/listinfo/users
>> >>
>> >>    _______________________________________________
>> >>    Quantum ESPRESSO is supported by MaX
>> >>    (www.max-centre.eu/quantum-espresso
>> >>    <http://www.max-centre.eu/quantum-espresso>)
>> >>    users mailing list users at lists.quantum-espresso.org
>> >>    <mailto:users at lists.quantum-espresso.org
>> <users at lists.quantum-espresso.org>>
>> >>    https://lists.quantum-espresso.org/mailman/listinfo/users
>> >>
>> >>
>> >>
>> >> --
>> >> Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
>> >> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
>> <https://www.google.com/maps/search/Udine,+via+delle+Scienze+208,+33100+Udine,+Italy?entry=gmail&source=g>
>> >> Phone +39-0432-558216, fax +39-0432-558222
>> >>
>> >>
>> >> _______________________________________________
>> >> Quantum ESPRESSO is supported by MaX (
>> www.max-centre.eu/quantum-espresso)
>> >> users mailing list users at lists.quantum-espresso.org
>> >> https://lists.quantum-espresso.org/mailman/listinfo/users
>> >>
>> >
>> > --
>> > Lorenzo Paulatto - Paris
>> > _______________________________________________
>> > Quantum ESPRESSO is supported by MaX (
>> www.max-centre.eu/quantum-espresso)
>> > users mailing list users at lists.quantum-espresso.org
>> > https://lists.quantum-espresso.org/mailman/listinfo/users
>>
>>
>>
>> GIUSEPPE MATTIOLI
>> CNR - ISTITUTO DI STRUTTURA DELLA MATERIA
>> Via Salaria Km 29,300 - C.P. 10
>> I-00015 - Monterotondo Scalo (RM)
>> Mob (*preferred*) +39 373 7305625
>> Tel + 39 06 90672342 - Fax +39 06 90672316
>> E-mail: <giuseppe.mattioli at ism.cnr.it>
>>
>> _______________________________________________
>> Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso)
>> users mailing list users at lists.quantum-espresso.org
>> https://lists.quantum-espresso.org/mailman/listinfo/users
>> _______________________________________________
>> Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso)
>> users mailing list users at lists.quantum-espresso.org
>> https://lists.quantum-espresso.org/mailman/listinfo/users
>
> _______________________________________________
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso)
> users mailing list users at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20200408/b9f309df/attachment.html>