Re: [Wien] Questions about Na2IrO3 (PBE+SOC+U calculation with magnetic order)

[Hung Yu Yang]

Thanks a lot for the replies from both of you! And in this case the orbital
moment is indeed not negligible when I checked :ORB (~0.4).

Yours sincerely,

Hung-Yu

On Mon, Aug 7, 2017 at 4:27 AM, pieper  wrote:

> Of course! The local moments in an AF insulator can be anything. I just
> saw that there is only one type of magnetic atom in the unit cell (Ir) and
> did not think properly. Sorry for that one.
>
> ---
> Dr. Martin Pieper
> Karl-Franzens University
> Institute of Physics
> Universitätsplatz 5
> A-8010 Graz
> Austria
> Tel.: +43-(0)316-380-8564
>
>
> Am 07.08.2017 09:18, schrieb Peter Blaha:
>
>> The Ir moments are read from case.scf file, but why should it be near
>>> an integer, and what is the unit of the reported moments in case.scf
>>> file? If the unit is bohr magneton, then for a spin 1/2, should it be
>>> ~1.7?
>>>
>>
>> For a FERROmagnetic insulator, the TOTAL spin moment/cell must be
>> integer. The moments of individual atoms can have any value.
>>
>> For an ANTIferromagnet, the total spin moment/cell must be zero, and
>> of course the atomic moments can again have any value.
>>
>> What you get by :MMT or MMIxxx are spin moments only.
>>
>> The moments are given in Bohr magnetons (mu_B), but are the spin
>> moments, not the effective moments. Thus if you have one unpaired
>> electron the spin moment is one.
>>
>> PS: For your compound, the orbital moments are probably as important
>> as the spin moments and you must add them to get the total moment.
>>
>>
>>
>>
>>> The gap can be roughly seen along the k-path I plotted and is checked by
>>> DOS plot, which shows it's an insulator.
>>>
>>> Yours sincerely,
>>>
>>> Hung-Yu
>>>
>>> On Sun, Aug 6, 2017 at 11:03 AM, pieper >> > wrote:
>>>
>>> A few thoughts on this:
>>>
>>> ad 1)
>>> Presumably Ir is the magnetic ion, so what you describe seems to be
>>> ok with me. I did not check your .struct file, however.
>>>
>>> ad 2)
>>> See chapters 4.5.6 and 7.4 of the UG (Wien2k 17.1) on ther
>>> orb-program, and the references therein. Personally I found the .pdf
>>> file of his talk on LDA+U very usefull that E. Assmann posted on the
>>> Wien2k-site.
>>>
>>> Note the recommendation in the UG for the SIC-mode of orb for of
>>> strongly correlated systems: set J=0 and use only U_eff=U-J. The
>>> value of U_eff is something you will have to decide, perhaps based
>>> on the approach described by Madsen and Novak cited in the UG. You
>>> also might want to take a look at eece as an alternative (UG chapter
>>> 4.5.7)
>>>
>>> >From many comments here in the mailing list and from the UG (again
>>> e.g. chapter 5.5.7, 4.5.8) I take it that PBE is what you should do
>>> if you want to calculate spacial charge and spin distributions, but
>>> to calculate gaps you may have to switch to numerically much more
>>> costly hybrid methods.
>>>
>>> You probably can (mis)use U_eff as a free parameter to adjust the
>>> gap in your PBE calculation to your favorite value. However, the
>>> physical meaning of the value would be dubious (imho), and there is
>>> no guarantee that the Ir-moments simultaneously come near your
>>> favorit 'theoretical moments' (whatever the actual value and origin
>>> of those is).
>>>
>>> ad 3) Don't bother with the starting values of local moments for
>>> atomic configurations, and for an antiferromagnet the interstitial
>>> moment obviously should stay close to zero.
>>>
>>> But the moments you give in your table are very far from integer.
>>> Are this Ir-spin moments from  case.scf or did you add orbital
>>> moemnts calculated by lapwdm? The spin moments should be somewhere
>>> near integer for an insulator.
>>>
>>> So, how did you determine the gap in the table? Did you plot a DOS?
>>> Is this really an insulator, or are there in fact bands crossing
>>> E_F? You might severly misjudge the (direct?) gap depending on where
>>> in k-space it is and the points in your k-list.
>>>
>>> Good luck
>>>
>>> ---
>>> Dr. Martin Pieper
>>> Karl-Franzens University
>>> Institute of Physics
>>> Universitätsplatz 5
>>> A-8010 Graz
>>> Austria
>>> Tel.: +43-(0)316-380-8564 
>>>
>>>
>>>
>>> Am 04.08.2017 19:22, schrieb Hung Yu Yang:
>>>
>>> Dear WIEN2k developers and users,
>>>
>>> I am trying to do a calculation on Na2IrO3, which has a band gap
>>> ~340meV and a zigzag antiferromagnetic order in its ground
>>> state, and
>>> I have some questions as follow:
>>>
>>> 1. To assign the zigzag antiferromagnetic order, what I did is
>>> that I
>>> first made a cif file that has two inequivalent Ir atoms, and let
>>> WIEN2k decide the symmetry for me. I adopted the generated
>>> structural
>>> file (attached in this mail), checked the cif file in some
>>>   

Re: [Wien] Questions about Na2IrO3 (PBE+SOC+U calculation with magnetic order)

[pieper]

Of course! The local moments in an AF insulator can be anything. I just 
saw that there is only one type of magnetic atom in the unit cell (Ir) 
and did not think properly. Sorry for that one.


---
Dr. Martin Pieper
Karl-Franzens University
Institute of Physics
Universitätsplatz 5
A-8010 Graz
Austria
Tel.: +43-(0)316-380-8564


Am 07.08.2017 09:18, schrieb Peter Blaha:

The Ir moments are read from case.scf file, but why should it be near
an integer, and what is the unit of the reported moments in case.scf
file? If the unit is bohr magneton, then for a spin 1/2, should it be 
~1.7?


For a FERROmagnetic insulator, the TOTAL spin moment/cell must be
integer. The moments of individual atoms can have any value.

For an ANTIferromagnet, the total spin moment/cell must be zero, and
of course the atomic moments can again have any value.

What you get by :MMT or MMIxxx are spin moments only.

The moments are given in Bohr magnetons (mu_B), but are the spin
moments, not the effective moments. Thus if you have one unpaired
electron the spin moment is one.

PS: For your compound, the orbital moments are probably as important
as the spin moments and you must add them to get the total moment.





The gap can be roughly seen along the k-path I plotted and is checked 
by

DOS plot, which shows it's an insulator.

Yours sincerely,

Hung-Yu

On Sun, Aug 6, 2017 at 11:03 AM, pieper mailto:pie...@ifp.tuwien.ac.at>> wrote:

A few thoughts on this:

ad 1)
Presumably Ir is the magnetic ion, so what you describe seems to 
be

ok with me. I did not check your .struct file, however.

ad 2)
See chapters 4.5.6 and 7.4 of the UG (Wien2k 17.1) on ther
orb-program, and the references therein. Personally I found the 
.pdf
file of his talk on LDA+U very usefull that E. Assmann posted on 
the

Wien2k-site.

Note the recommendation in the UG for the SIC-mode of orb for of
strongly correlated systems: set J=0 and use only U_eff=U-J. The
value of U_eff is something you will have to decide, perhaps based
on the approach described by Madsen and Novak cited in the UG. You
also might want to take a look at eece as an alternative (UG 
chapter

4.5.7)

>From many comments here in the mailing list and from the UG 
(again
e.g. chapter 5.5.7, 4.5.8) I take it that PBE is what you should 
do
if you want to calculate spacial charge and spin distributions, 
but

to calculate gaps you may have to switch to numerically much more
costly hybrid methods.

You probably can (mis)use U_eff as a free parameter to adjust the
gap in your PBE calculation to your favorite value. However, the
physical meaning of the value would be dubious (imho), and there 
is

no guarantee that the Ir-moments simultaneously come near your
favorit 'theoretical moments' (whatever the actual value and 
origin

of those is).

ad 3) Don't bother with the starting values of local moments for
atomic configurations, and for an antiferromagnet the interstitial
moment obviously should stay close to zero.

But the moments you give in your table are very far from integer.
Are this Ir-spin moments from  case.scf or did you add orbital
moemnts calculated by lapwdm? The spin moments should be somewhere
near integer for an insulator.

So, how did you determine the gap in the table? Did you plot a 
DOS?

Is this really an insulator, or are there in fact bands crossing
E_F? You might severly misjudge the (direct?) gap depending on 
where

in k-space it is and the points in your k-list.

Good luck

---
Dr. Martin Pieper
Karl-Franzens University
Institute of Physics
Universitätsplatz 5
A-8010 Graz
Austria
Tel.: +43-(0)316-380-8564 



Am 04.08.2017 19:22, schrieb Hung Yu Yang:

Dear WIEN2k developers and users,

I am trying to do a calculation on Na2IrO3, which has a band 
gap

~340meV and a zigzag antiferromagnetic order in its ground
state, and
I have some questions as follow:

1. To assign the zigzag antiferromagnetic order, what I did is
that I
first made a cif file that has two inequivalent Ir atoms, and 
let

WIEN2k decide the symmetry for me. I adopted the generated
structural
file (attached in this mail), checked the cif file in some
visulization software and made sure they were structurally
equivalent
except that there were 2 inequivalent Ir atoms instead of 1. 
Is this
the proper way to generate structural files for magnetically 
ordered

systems?

2. After generating the desired structural file, I put up on 
Ir1

atom
and down on the other (Ir2), used several different 
combinations

of U
and J, and I got the following results:

U(eV) J(eV) Ueff=U-J (eV)  Moment(Ir1, 
Ir2)

  Gap(meV)
2.1

Re: [Wien] Questions about Na2IrO3 (PBE+SOC+U calculation with magnetic order)

[Peter Blaha]

The Ir moments are read from case.scf file, but why should it be near
an integer, and what is the unit of the reported moments in case.scf
file? If the unit is bohr magneton, then for a spin 1/2, should it be ~1.7?


For a FERROmagnetic insulator, the TOTAL spin moment/cell must be 
integer. The moments of individual atoms can have any value.


For an ANTIferromagnet, the total spin moment/cell must be zero, and of 
course the atomic moments can again have any value.


What you get by :MMT or MMIxxx are spin moments only.

The moments are given in Bohr magnetons (mu_B), but are the spin 
moments, not the effective moments. Thus if you have one unpaired 
electron the spin moment is one.


PS: For your compound, the orbital moments are probably as important as 
the spin moments and you must add them to get the total moment.






The gap can be roughly seen along the k-path I plotted and is checked by
DOS plot, which shows it's an insulator.

Yours sincerely,

Hung-Yu

On Sun, Aug 6, 2017 at 11:03 AM, pieper mailto:pie...@ifp.tuwien.ac.at>> wrote:

A few thoughts on this:

ad 1)
Presumably Ir is the magnetic ion, so what you describe seems to be
ok with me. I did not check your .struct file, however.

ad 2)
See chapters 4.5.6 and 7.4 of the UG (Wien2k 17.1) on ther
orb-program, and the references therein. Personally I found the .pdf
file of his talk on LDA+U very usefull that E. Assmann posted on the
Wien2k-site.

Note the recommendation in the UG for the SIC-mode of orb for of
strongly correlated systems: set J=0 and use only U_eff=U-J. The
value of U_eff is something you will have to decide, perhaps based
on the approach described by Madsen and Novak cited in the UG. You
also might want to take a look at eece as an alternative (UG chapter
4.5.7)

>From many comments here in the mailing list and from the UG (again
e.g. chapter 5.5.7, 4.5.8) I take it that PBE is what you should do
if you want to calculate spacial charge and spin distributions, but
to calculate gaps you may have to switch to numerically much more
costly hybrid methods.

You probably can (mis)use U_eff as a free parameter to adjust the
gap in your PBE calculation to your favorite value. However, the
physical meaning of the value would be dubious (imho), and there is
no guarantee that the Ir-moments simultaneously come near your
favorit 'theoretical moments' (whatever the actual value and origin
of those is).

ad 3) Don't bother with the starting values of local moments for
atomic configurations, and for an antiferromagnet the interstitial
moment obviously should stay close to zero.

But the moments you give in your table are very far from integer.
Are this Ir-spin moments from  case.scf or did you add orbital
moemnts calculated by lapwdm? The spin moments should be somewhere
near integer for an insulator.

So, how did you determine the gap in the table? Did you plot a DOS?
Is this really an insulator, or are there in fact bands crossing
E_F? You might severly misjudge the (direct?) gap depending on where
in k-space it is and the points in your k-list.

Good luck

---
Dr. Martin Pieper
Karl-Franzens University
Institute of Physics
Universitätsplatz 5
A-8010 Graz
Austria
Tel.: +43-(0)316-380-8564 



Am 04.08.2017 19:22, schrieb Hung Yu Yang:

Dear WIEN2k developers and users,

I am trying to do a calculation on Na2IrO3, which has a band gap
~340meV and a zigzag antiferromagnetic order in its ground
state, and
I have some questions as follow:

1. To assign the zigzag antiferromagnetic order, what I did is
that I
first made a cif file that has two inequivalent Ir atoms, and let
WIEN2k decide the symmetry for me. I adopted the generated
structural
file (attached in this mail), checked the cif file in some
visulization software and made sure they were structurally
equivalent
except that there were 2 inequivalent Ir atoms instead of 1. Is this
the proper way to generate structural files for magnetically ordered
systems?

2. After generating the desired structural file, I put up on Ir1
atom
and down on the other (Ir2), used several different combinations
of U
and J, and I got the following results:

U(eV) J(eV) Ueff=U-J (eV)  Moment(Ir1, Ir2)
  Gap(meV)
2.1 0.6 1.5
0.22532,-0.22439~700meV
2.4 02.4
0.30105,-0.30109~750meV
30.6 2.4
0.23225,-0.23235~900meV
31.5 1.5
0.17203,-0.17210~900meV

First, in this test, it seems that the gap is affected by U
on

Re: [Wien] Questions about Na2IrO3 (PBE+SOC+U calculation with magnetic order)

[Hung Yu Yang]

Hi,

For your doubts with regard to (3), my responses are as follow:

The Ir moments are read from case.scf file, but why should it be near
an integer, and what is the unit of the reported moments in case.scf file?
If the unit is bohr magneton, then for a spin 1/2, should it be ~1.7?

The gap can be roughly seen along the k-path I plotted and is checked by
DOS plot, which shows it's an insulator.

Yours sincerely,

Hung-Yu

On Sun, Aug 6, 2017 at 11:03 AM, pieper  wrote:

> A few thoughts on this:
>
> ad 1)
> Presumably Ir is the magnetic ion, so what you describe seems to be ok
> with me. I did not check your .struct file, however.
>
> ad 2)
> See chapters 4.5.6 and 7.4 of the UG (Wien2k 17.1) on ther orb-program,
> and the references therein. Personally I found the .pdf file of his talk on
> LDA+U very usefull that E. Assmann posted on the Wien2k-site.
>
> Note the recommendation in the UG for the SIC-mode of orb for of strongly
> correlated systems: set J=0 and use only U_eff=U-J. The value of U_eff is
> something you will have to decide, perhaps based on the approach described
> by Madsen and Novak cited in the UG. You also might want to take a look at
> eece as an alternative (UG chapter 4.5.7)
>
> From many comments here in the mailing list and from the UG (again e.g.
> chapter 5.5.7, 4.5.8) I take it that PBE is what you should do if you want
> to calculate spacial charge and spin distributions, but to calculate gaps
> you may have to switch to numerically much more costly hybrid methods.
>
> You probably can (mis)use U_eff as a free parameter to adjust the gap in
> your PBE calculation to your favorite value. However, the physical meaning
> of the value would be dubious (imho), and there is no guarantee that the
> Ir-moments simultaneously come near your favorit 'theoretical moments'
> (whatever the actual value and origin of those is).
>
> ad 3) Don't bother with the starting values of local moments for atomic
> configurations, and for an antiferromagnet the interstitial moment
> obviously should stay close to zero.
>
> But the moments you give in your table are very far from integer. Are this
> Ir-spin moments from  case.scf or did you add orbital moemnts calculated by
> lapwdm? The spin moments should be somewhere near integer for an insulator.
>
> So, how did you determine the gap in the table? Did you plot a DOS? Is
> this really an insulator, or are there in fact bands crossing E_F? You
> might severly misjudge the (direct?) gap depending on where in k-space it
> is and the points in your k-list.
>
> Good luck
>
> ---
> Dr. Martin Pieper
> Karl-Franzens University
> Institute of Physics
> Universitätsplatz 5
> A-8010 Graz
> Austria
> Tel.: +43-(0)316-380-8564
>
>
>
> Am 04.08.2017 19:22, schrieb Hung Yu Yang:
>
>> Dear WIEN2k developers and users,
>>
>> I am trying to do a calculation on Na2IrO3, which has a band gap
>> ~340meV and a zigzag antiferromagnetic order in its ground state, and
>> I have some questions as follow:
>>
>> 1. To assign the zigzag antiferromagnetic order, what I did is that I
>> first made a cif file that has two inequivalent Ir atoms, and let
>> WIEN2k decide the symmetry for me. I adopted the generated structural
>> file (attached in this mail), checked the cif file in some
>> visulization software and made sure they were structurally equivalent
>> except that there were 2 inequivalent Ir atoms instead of 1. Is this
>> the proper way to generate structural files for magnetically ordered
>> systems?
>>
>> 2. After generating the desired structural file, I put up on Ir1 atom
>> and down on the other (Ir2), used several different combinations of U
>> and J, and I got the following results:
>>
>> U(eV) J(eV) Ueff=U-J (eV)  Moment(Ir1, Ir2)
>>   Gap(meV)
>> 2.1 0.6 1.5
>> 0.22532,-0.22439~700meV
>> 2.4 02.4
>> 0.30105,-0.30109~750meV
>> 30.6 2.4
>> 0.23225,-0.23235~900meV
>> 31.5 1.5
>> 0.17203,-0.17210~900meV
>>
>> First, in this test, it seems that the gap is affected by U only, not
>> Ueff=U-J. Second, the moments of Ir seemed to be closely related to J.
>> In this situation, what is the proper way to assign U and J? I am not
>> sure how much it means if I just try to tune U and J until they match
>> the experimental gap and/or theoretical moments.
>>
>> 3. I also have a question about the unit of moment given here; I
>> understand that the moments depend on the RMT sizes and I assumed the
>> unit is in bohr magneton, but the first value shown for Ir atom is
>> 0.76697 and I am not sure in what unit is this value and how it
>> assigns the initial moment for a certain element (say, Ir in my case).
>> Also, the interstitial moments are nearly 0 among all the cycles.
>>
>> For this calculation, kpoints=450, RKmax=8 and I had to use the TEMP
>> scheme for it to converge. I appreciate any reply from you.
>>
>> Yours s

Re: [Wien] Questions about Na2IrO3 (PBE+SOC+U calculation with magnetic order)

[pieper]

A few thoughts on this:

ad 1)
Presumably Ir is the magnetic ion, so what you describe seems to be ok 
with me. I did not check your .struct file, however.


ad 2)
See chapters 4.5.6 and 7.4 of the UG (Wien2k 17.1) on ther orb-program, 
and the references therein. Personally I found the .pdf file of his talk 
on LDA+U very usefull that E. Assmann posted on the Wien2k-site.


Note the recommendation in the UG for the SIC-mode of orb for of 
strongly correlated systems: set J=0 and use only U_eff=U-J. The value 
of U_eff is something you will have to decide, perhaps based on the 
approach described by Madsen and Novak cited in the UG. You also might 
want to take a look at eece as an alternative (UG chapter 4.5.7)


From many comments here in the mailing list and from the UG (again e.g. 
chapter 5.5.7, 4.5.8) I take it that PBE is what you should do if you 
want to calculate spacial charge and spin distributions, but to 
calculate gaps you may have to switch to numerically much more costly 
hybrid methods.


You probably can (mis)use U_eff as a free parameter to adjust the gap in 
your PBE calculation to your favorite value. However, the physical 
meaning of the value would be dubious (imho), and there is no guarantee 
that the Ir-moments simultaneously come near your favorit 'theoretical 
moments' (whatever the actual value and origin of those is).


ad 3) Don't bother with the starting values of local moments for atomic 
configurations, and for an antiferromagnet the interstitial moment 
obviously should stay close to zero.


But the moments you give in your table are very far from integer. Are 
this Ir-spin moments from  case.scf or did you add orbital moemnts 
calculated by lapwdm? The spin moments should be somewhere near integer 
for an insulator.


So, how did you determine the gap in the table? Did you plot a DOS? Is 
this really an insulator, or are there in fact bands crossing E_F? You 
might severly misjudge the (direct?) gap depending on where in k-space 
it is and the points in your k-list.


Good luck

---
Dr. Martin Pieper
Karl-Franzens University
Institute of Physics
Universitätsplatz 5
A-8010 Graz
Austria
Tel.: +43-(0)316-380-8564


Am 04.08.2017 19:22, schrieb Hung Yu Yang:

Dear WIEN2k developers and users,

I am trying to do a calculation on Na2IrO3, which has a band gap
~340meV and a zigzag antiferromagnetic order in its ground state, and
I have some questions as follow:

1. To assign the zigzag antiferromagnetic order, what I did is that I
first made a cif file that has two inequivalent Ir atoms, and let
WIEN2k decide the symmetry for me. I adopted the generated structural
file (attached in this mail), checked the cif file in some
visulization software and made sure they were structurally equivalent
except that there were 2 inequivalent Ir atoms instead of 1. Is this
the proper way to generate structural files for magnetically ordered
systems?

2. After generating the desired structural file, I put up on Ir1 atom
and down on the other (Ir2), used several different combinations of U
and J, and I got the following results:

U(eV) J(eV) Ueff=U-J (eV)  Moment(Ir1, Ir2)
  Gap(meV)
2.1 0.6 1.5
0.22532,-0.22439~700meV
2.4 02.4
0.30105,-0.30109~750meV
30.6 2.4
0.23225,-0.23235~900meV
31.5 1.5
0.17203,-0.17210~900meV

First, in this test, it seems that the gap is affected by U only, not
Ueff=U-J. Second, the moments of Ir seemed to be closely related to J.
In this situation, what is the proper way to assign U and J? I am not
sure how much it means if I just try to tune U and J until they match
the experimental gap and/or theoretical moments.

3. I also have a question about the unit of moment given here; I
understand that the moments depend on the RMT sizes and I assumed the
unit is in bohr magneton, but the first value shown for Ir atom is
0.76697 and I am not sure in what unit is this value and how it
assigns the initial moment for a certain element (say, Ir in my case).
Also, the interstitial moments are nearly 0 among all the cycles.

For this calculation, kpoints=450, RKmax=8 and I had to use the TEMP
scheme for it to converge. I appreciate any reply from you.

Yours sincerely,

Hung-Yu
___
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html

___
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:  
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html