[QE-users] QE-7.1 and DFT+U

Wed Jun 29 19:22:43 CEST 2022

Dear Iurii,

Thank you so much for your detailed explanation.

Now I feel good with the new input syntax, but I think I have to explore 
more abilities.

By the way, I checked if the results of the reference:

https://arxiv.org/abs/2108.12758

is reproducible using QE-7.1, and found out that it is "YES" !

So, I am mostly convinced that in the new release the "hard-coded" has been 
promoted to the "easy-coded"

without any significant changes in the code.

Thank you so much for your patience.

Best Wishes,

Mahmoud

From: Iurii TIMROV via users <users at lists.quantum-espresso.org>
To: Mahmoud Payami Shabestari <mpayami at aeoi.org.ir>, "Quantum ESPRESSO users 
Forum" < @lists.quantum-espresso.org>
Date: Wed, 29 Jun 2022 16:29:58 +0000
Subject: Re: [QE-users] QE-7.1 and DFT+U

Dear Mahmoud,

Please have a look at the new documentation in QE/Doc/Hubbard_input.pdf.

In the new input you need to specify to which manifold you are applying the 
Hubbard U correction. For Uranium it could be 5f or 6d or both. So if you 
want to apply U to 6d states then the input syntax is:

HUBBARD {atomic}
U U1-6d 4.0
U U2-6d 4.0

If instead you want to apply U to 5f states then:

HUBBARD {atomic}
U U1-5f 4.0
U U2-5f 4.0

If you want to apply U to both then

HUBBARD {atomic}
U U1-5f  4.0

U U1-6d 4.0
U U2-5f  4.0

U U2-6d 4.0

Of course the value of U should be different for 5f and 6d manifolds.

>From our experience, "ortho-atomic" Hubbard projectors are more accurate 
than "atomic". Note that when one changes the projector type also the value 
of U changes.

The value of U can be computed from first-principles using the HP code: 
https://www.sciencedirect.com/science/article/pii/S0010465522001746

However, currently the HP code cannot be used to compute U when there are 
two (or more) Hubbard channels per atomic type.

Please let us know if there are still any issues with the new Hubbard input 
syntax (in this case provide the full input and output files).

Greetings,

Iurii

--
Dr. Iurii TIMROV
Senior Research Scientist
Theory and Simulation of Materials (THEOS)
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
From: Mahmoud Payami Shabestari <mpayami at aeoi.org.ir>
Sent: Wednesday, June 29, 2022 4:34:44 PM
To: Iurii TIMROV; Quantum ESPRESSO users Forum
Subject: Re: [QE-users] QE-7.1 and DFT+U

Dear Iurii,

Thank you so much for your comments.
In trying to run my case using QE-7.1,
I get the error message:

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     Error in routine card_hubbard (1):
     Hubbard atom name missing or wrong or too long
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

While I have prepared the input similar to that of example:

HUBBARD {atomic}
U U1 4.0
U U2 4.0

where U1 and U2 are two kinds of U atoms in 1k-AFM.

Should I add something more in the input?

Thank you in advance.

Best regards,
Mahmoud

From: Iurii TIMROV via users <users at lists.quantum-espresso.org>
To: Mahmoud Payami Shabestari <mpayami at aeoi.org.ir>, "Quantum ESPRESSO users 
Forum" <users at lists.quantum-espresso.org>
Date: Wed, 29 Jun 2022 14:01:39 +0000
Subject: Re: [QE-users] QE-7.1 and DFT+U

Dear Mahmoud,

> https://arxiv.org/abs/2108.12758

Thank you for this reference! I am going to read it. 

> You argue that since O atom is not magnetic, the starting_magnetization 
for that atom is taken to be zero.

I was not precise enough. I meant for FeO, while in general this might not 
be the case (I don't have much experience with this though).

In the older versions of QE the starting_magnetization for O was 0.0. And 
the calculations were converging to the solution where the magnetic moments 
on O are zero. That's why I removed starting_magnetization for O in the new 
example which means that the default value of 0.0 is used for O. I tried to 
specify the finite value of starting_magnetization for O in FeO but still 
the calculation convergences to the solution where the magnetic moments on O 
are zero.

> By "hard-coded" I do not understand what you mean.

I mean that it is implemented in the routine PW/src/tabd.f90 in the older 
version of QE and that it was not possible to change these parameters from 
the input file. In contrast, in QE7.1 the starting atomic occupations of 
Hubbard atoms can be controlled from the input (but by default they are read 
from the pseudopotentials).

Greetings,
Iurii

--
Dr. Iurii TIMROV
Senior Research Scientist
Theory and Simulation of Materials (THEOS)
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
From: Mahmoud Payami Shabestari <mpayami at aeoi.org.ir>
Sent: Wednesday, June 29, 2022 3:38:54 PM
To: Iurii TIMROV; Quantum ESPRESSO users Forum
Subject: Re: [QE-users] QE-7.1 and DFT+U

Dear Iurii,

You argue that since O atom is not magnetic, the starting_magnetization for 
that atom is taken to be zero.
I would like to attract your attention to the work:

https://arxiv.org/abs/2108.12758

in which it is shown that the GS of UO2 is obtained for non-zero 
starting_magnetization of O-atom.
I have to check the results from QE-7.1 of my inputs (of course in new 
syntax) of the above work to see if I get the same results. Otherwise, the 
occupations were wrongly dependent on the starting_magnetization of O-atoms, 
and now it is bug-fixed. By "hard-coded" I do not understand what you mean.
Thank you again.

Best regards,
Mahmoud

From: Iurii TIMROV via users <users at lists.quantum-espresso.org>
To: Mahmoud Payami Shabestari <mpayami at aeoi.org.ir>, "Quantum ESPRESSO users 
Forum" <users at lists.quantum-espresso.org>
Date: Wed, 29 Jun 2022 11:44:03 +0000
Subject: Re: [QE-users] QE-7.1 and DFT+U

Dear Mahmoud,

> Explicitly, the starting_magnetization for O is removed in the new one. 
Does it mean the default value is still zero or the code searches for a 
better starting magnetization?

O atoms are nonmagnetic so that's why starting_magnetization for O was 
removed. 

> But, comparing the results from qe-7.1 and qe-6.7, one notices significant 
changes in the input and outputs of feo_user_ns files.

FeO is metallic in DFT while it is insulting in nature. The difference 
between the new and old behaviour of QE is the starting occupations of the 
Fe-3d manifold. In older versions of QE (before 7.1) the starting 
occupations are hard-coded to be 6 (see PW/src/tabd.f90) while in QE7.1 the 
starting occupations are read from the pseudopotential (in this particular 
example it is 7). It turns out that in this example different starting 
occupations for Fe-3d states lead to the convergence to different metallic 
ground states of FeO, but in any case both are wrong ground states (one 
needs +U and to use starting_ns_eignevalue to open a gap and thus obtain the 
correct ground state). But if you want to restore the old behaviour of QE in 
the new version (i.e. 7.1) you need to specify in the input the new keyword 
Hubbard_occ(1,1) = 6.0 and Hubbard_occ(2,1) = 6.0 (because you have Fe1 and 
Fe2) - in this case the code reads the starting occupations from the input 
and not from the pseudopotential.  

Greetings,
Iurii

--
Dr. Iurii TIMROV
Senior Research Scientist
Theory and Simulation of Materials (THEOS)
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 Mahmoud 
Payami Shabestari via users <users at lists.quantum-espresso.org>
Sent: Wednesday, June 29, 2022 1:16:53 PM
To: users at lists.quantum-espresso.org
Subject: [QE-users] QE-7.1 and DFT+U

Dear QE developers,

Hi.
In the new release QE-7.1, I investigated through the example08 of PW, the 
example of FeO.
In the release-notes, Incompatible changes, line 3, it is mentioned that 
"input syntax" is changed.
But, comparing the results from qe-7.1 and qe-6.7, one notices significant 
changes in the input and outputs of feo_user_ns files.
Explicitly, the starting_magnetization for O is removed in the new one. Does 
it mean the default value is still zero or the code searches for a better 
starting magnetization?
Any comments is highly appreciated.

With best regards,
Mahmoud Payami
NSTRI, AEOI, Tehran, Iran

Email: mpayami at aeoi.org.ir
Phone: +98 (0)21 82066504

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20220629/849a1e9e/attachment.html>