[QE-users] Problems with hp.x
Giuseppe Mattioli
giuseppe.mattioli at ism.cnr.it
Thu Jun 18 12:24:22 CEST 2020
Out of further curiosity, a few years ago there has been a fuss over
the "global U" approach of the pseudo-hybrid ACBN0 functional, which
had been tested on ZnO among other cases (Phys. Rev. X 2015, 5, 011006).
It seems that the ACBN0 implementation "didn't hit the market"... Is
it an active project?
Best
Giuseppe
Quoting Timrov Iurii <iurii.timrov at epfl.ch>:
>> ...because the 2 shell is fully occupied
>
>
> because the d shell is fully occupied.
>
>
> Sorry!
>
>
> Iurii
>
>
> ________________________________
> From: users <users-bounces at lists.quantum-espresso.org> on behalf of
> Timrov Iurii <iurii.timrov at epfl.ch>
> Sent: Thursday, June 18, 2020 10:08:19 AM
> To: Quantum ESPRESSO users Forum
> Subject: Re: [QE-users] Problems with hp.x
>
>
>> Since it's Cu(I) isn't
>> that a closed shell, a d10 system aswell? I was wondering why the
>> linear-response theory works in this case but not for Zn2+.
>
>
> Indeed, Cu2O has the same problem as ZnO, because the 2 shell is
> fully occupied. In our paper, for Cu2O we obtained U for Cu-3d
> states of 11.3 eV - this is from a one-shot calculation. This value
> of U is already large. If U is computed in a self-consistent way,
> then at 2nd, 3rd, and next iterations U for Cu-3d states will be
> larger and larger. This is because the d shell is full and it is
> pushed deeper and deeper in energy with increasing U, and the
> linear-response theory gives larger and larger values of U at the
> subsequent iterations. I realized quite late that Cu2O is not really
> a good "demonstrator" for our PRB 2018 paper - however, on the other
> hand, still it is OK just for the sake of benchmarking DFPT versus
> the finite-difference supercell approach of PRB 71, 035105 (2005).
>
>
> Concerning the question why in the one-shot calculation Cu-3d (in
> Cu2O) has U of ~11 eV and Zn-3d (in ZnO) has U of ~20 eV, is because
> Zn-3d states are deeper in energy than Cu-3d states (at the DFT
> level - which is typically the starting point for the HP calculation).
>
>
> HTH
>
>
> 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
> dv009200 at fh-muenster.de <dv009200 at fh-muenster.de>
> Sent: Wednesday, June 17, 2020 11:42:28 PM
> To: Quantum ESPRESSO users Forum
> Subject: Re: [QE-users] Problems with hp.x
>
> Thanks again thats for sure very helpful.
>
> Just out of curiosity I may have one more question: In the paper thats the
> basis of the hp.x code (I. Timrov, N. Marzari, M. Cococcioni, Phys. Rev. B
> 98, 085127, DOI:10.1103/PhysRevB.98.085127) the Hubbard U parameter of
> Cu2O is computed by the linear-response theory. Since it's Cu(I) isn't
> that a closed shell, a d10 system aswell? I was wondering why the
> linear-response theory works in this case but not for Zn2+.
>
> Best regards
>
> Dominik Voigt
>
> Dominik Voigt
> PhD Student University of Applied Sciences Münster
> Department of Physical Chemistry
>
>
>
>>
>> Dear Dominik
>>
>>> the method you suggested Giuseppe. I don't quite know whats the 'right'
>>> distance between the valence band and the d-band? Can I use experimental
>>> data from xps/ups spectra for that matter or do you have other
>>> suggestions?
>>
>> Yes, it is actually what I did with ZnO. As I remarked in my previous
>> message, it is going to work well if you apply the correction also to
>> the S 3p shell, in a DFT+U(Zn,S) fashion. In wurtzite ZnO the strong
>> curvature of the Zn 4s conduction band is such that the CB is
>> substantially free from delocalization error and it is a good absolute
>> reference. If you know the position of the Fermi level (which can
>> depend on intrinsic defects in real samples) with respect to the CB,
>> then you can reasonably estimate the position of the Zn 3d band.
>> I hope I've not added entropy...
>> Best
>> Giuseppe
>>
>> Quoting dv009200 at fh-muenster.de:
>>
>>> Thanks all for the helpful comments and interesting papers. I'm fine
>>> correcting the delocalization error of the Zn 3d states 'semiempirical'
>>> by
>>> the method you suggested Giuseppe. I don't quite know whats the 'right'
>>> distance between the valence band and the d-band? Can I use experimental
>>> data from xps/ups spectra for that matter or do you have other
>>> suggestions?
>>>
>>> Best regards
>>>
>>> Dominik Voigt
>>>
>>> Dominik Voigt
>>> PhD Student University of Applied Sciences Münster
>>> Department of Physical Chemistry
>>>
>>>
>>>
>>>> Dear All,
>>>>
>>>>
>>>> Giuseppe is right, this is a limitation of the current implementation
>>>> of
>>>> the linear-response theory to compute Hubbard parameters for
>>>> closed-shell
>>>> systems (like ZnO, ZnS, etc.). This "limitation" is known, and there is
>>>> some comment about this in q-e/HP/Doc/README. Please check this paper
>>>> for
>>>> more details:
>>>> <https://aip.scitation.org/doi/10.1063/1.4869718>
>>>>
>>>> https://aip.scitation.org/doi/10.1063/1.4869718
>>>>
>>>>
>>>> The developers of the HP code are planning to investigate this issue in
>>>> more detail, and hopefully we will come out with some solution in the
>>>> near
>>>> future.
>>>>
>>>>
>>>> Best regards,
>>>>
>>>> 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: Monday, June 15, 2020 10:55:59 PM
>>>> To: Quantum ESPRESSO users Forum
>>>> Subject: Re: [QE-users] Problems with hp.x
>>>>
>>>>
>>>> Dear Dominik
>>>> I suppose that the problem is not in hp.x, but in the application of
>>>> the linear-response method itself to Zn(2+). Zn(2+) is a d10
>>>> transition metal, with the 3d band fully occupied. In ZnO, e.g., the
>>>> Zn 3d band is quite narrow and placed below the O 2p valence band, and
>>>> I suppose that the same holds for ZnS, with the Zn 3d band pushing up
>>>> the S 3p band. When you apply the LR method to Zn, you compute
>>>> quantities such as d(alpha)/dn, where alpha is the (small)
>>>> perturbation and n is the occupation number of d orbitals on site I
>>>> (see International Journal of Quantum Chemistry 2014, 114, 14 for
>>>> details). If the shell is full, then you can perturb whatever you want
>>>> but you will never obtain more than the full occupation of the shell
>>>> that you already have in the unperturbed system. This is likely the
>>>> reason for the crazy values of LR U you obtain. If you want to correct
>>>> the strong delocalization error of the Zn 3d narrow band within the
>>>> DFT+U formalism, then you must use a "semiempirical" approach,
>>>> choosing, e.g., the U value that places the Zn 3d shell at the correct
>>>> distance from the valence band maximum. In this case, I would
>>>> recommend the use of a second +U correction on the S 3p shell, which
>>>> should ensure a good recovery of the ZnS band gap. I've satisfactorily
>>>> used this scheme in the case of ZnO in several publications, from
>>>> which you may want to take inspiration (Adv. Energy Mater. 2014, 4,
>>>> 1301694).
>>>> HTH
>>>> Giuseppe
>>>>
>>>> Quoting dv009200 at fh-muenster.de:
>>>>
>>>>> Hello everyone,
>>>>>
>>>>> I'm trying to calculate the hubbard u parameter for Zn in Zinc sulfide
>>>>> (sphalerite structure) with the help of the hp.x code. The
>>>>> calculations
>>>>> terminate normally without any errors. The problem is that I get
>>>>> (presumably) way too high values for U that also won't converge (if I
>>>>> take
>>>>> the value I got from a one-shot calculation and plug it in the SCF
>>>>> input
>>>>> and then redo the HP calculation).
>>>>>
>>>>> For example in the first step I calculate a U = 75.7035 in the second
>>>>> iteration I get U = 804.2405 and in the third U = 30999.2684.
>>>>>
>>>>> This seems unreasonable considering that the calculations for the
>>>>> provided
>>>>> examples in the 'HP' folder work fine and converge fast without such a
>>>>> massive change to a certain value for U using the above described
>>>>> scheme.
>>>>>
>>>>> Has someone an idea what is causing this trouble in my system? I
>>>>> already
>>>>> tried different PPs, functionals, U_projection_type, thresholds and k
>>>>> and
>>>>> q point grids all without success.
>>>>>
>>>>> Below is my input for the scf and hp calculation
>>>>>
>>>>> SCF-input:
>>>>> &control
>>>>> calculation='scf'
>>>>> restart_mode='from_scratch',
>>>>> pseudo_dir = '/home/dominik/codes/QE6.5/pseudo/'
>>>>> outdir='/home/dominik/codes/QE6.5/tempdir/'
>>>>> prefix='zns'
>>>>> /
>>>>> &SYSTEM
>>>>> ibrav = 2
>>>>> celldm(1)=10.291937439
>>>>> nat = 2
>>>>> ntyp = 2
>>>>> ecutwfc = 60.0
>>>>> ecutrho= 720.0
>>>>> lda_plus_u = .true.
>>>>> lda_plus_u_kind = 0
>>>>> U_projection_type = 'atomic'
>>>>> Hubbard_U(1) = 1d-8
>>>>> /
>>>>> &electrons
>>>>> mixing_beta=0.7
>>>>> conv_thr=1d-15
>>>>> /
>>>>> ATOMIC_SPECIES
>>>>> Zn 65.39 Zn.pbe-dn-rrkjus_psl.0.2.2.UPF
>>>>> S 32.07 S.pbe-n-rrkjus_psl.0.1.UPF
>>>>> ATOMIC_POSITIONS {alat}
>>>>> Zn 0.000000 0.000000 0.000000
>>>>> S 0.250000 0.250000 0.250000
>>>>> K_POINTS automatic
>>>>> 12 12 12 0 0 0
>>>>>
>>>>>
>>>>> HP-input:
>>>>> &inputhp
>>>>> prefix='zns'
>>>>> outdir='/home/dominik/codes/QE6.5/tempdir/'
>>>>> nq1 = 2
>>>>> nq2 = 2
>>>>> nq3 = 2
>>>>> conv_thr_chi = 1.0d-10
>>>>> iverbosity =2
>>>>> /
>>>>>
>>>>>
>>>>> Best regards
>>>>>
>>>>>
>>>>> Dominik Voigt
>>>>>
>>>>> Dominik Voigt
>>>>> PhD Student University of Applied Sciences Münster
>>>>> Department of Physical Chemistry
>>>>>
>>>>> _______________________________________________
>>>>> Quantum ESPRESSO is supported by MaX
>>>>> (www.max-centre.eu/quantum-espresso<http://www.max-centre.eu/quantum-espresso<http://www.max-centre.eu/quantum-espresso<http://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<http://www.max-centre.eu/quantum-espresso<http://www.max-centre.eu/quantum-espresso<http://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<http://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<http://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<http://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<http://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>
More information about the users
mailing list