[Wien] Magnetic material relaxation, RMT, MAE calculations

[Jihoon Park]

Dear Ciao,


Thank you for your reply. Your advice was very helpful for me to understand.
I have used 10,000 k-points for MnAl calculation. 372,248 is 37 times larger
than the k-points that I used.

And I did not mean the experimental RMT. I meant changing RMT to fit an
experimental K value.

So, is it a good number of k-points (~300,000) for 2 to 4 atoms phase to
calculated K?
And should I do the relaxation for magnetic materials? Somebody say that we
do not need it for magnetic materials; simply take an experimental lattice
constant.


All my best,
Jihoon Park


To find the magntic anisotropy is a "brute force" job (see chapter 4.6.4 of
the textbook of J K?bler, Theory of itinerant electron magnetism),
that means you need really many k-points
(The question is what is "many" ? As answer from experience I would say: If
you think you have enough then just take a lot more !
There is an example in K?blers book where Halilov used 372 248 k-points in
the full Brillouinn zone.)
So how many k-points did you use ?

There is another thing one should keep in mind, some experimental  values
may be influenced by the shape anisotropy
of the sample that was used (unfortunately in many experimental work not
enough information is given).
MnAl has a rich binary phase diagramm, are you shure that the experimental
values are for a single phase sample ?

What do you mean with experimental RMT ?
There is some older LEED theory where the muffin tin radii were calculated
from the energy dependence of
low energy electron diffraction pattern. I guess that these have nothing to
do with the RMT you set in Wien2k.


Ciao
Gerhard


Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz

Von: wien-bounces at
zeus.theochem.tuwien.ac.at<http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien>[wien-bounces
at 
zeus.theochem.tuwien.ac.at<http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien>]"
im Auftrag von "Jihoon Park [jpark61 at
crimson.ua.edu<http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien>
]
Gesendet: Samstag, 4. Juni 2011 04:37
Bis: wien at 
zeus.theochem.tuwien.ac.at<http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien>
Betreff: [Wien] Magnetic material relaxation, RMT, MAE calculations

Dear users,


I am really wondering if we have to "relax" structure everytime when
calculating magnetic materials.
When relaxed, the magnetic properties, especially K value, become too far
from experimental data; K value calculated from not relaxed structure is
also far from the experimental one.

I tried force theorem as well, but that is still far from the experimental
data too.

Everytime my calculated K value is reversed.
For example, energies for <001> and <100> are compared and used to calculate
K.
But I do not understand why the energy for <100> is lower than that of
<001>.
Experimentally, <001> is stable for MnAl.


And another qeustion is about RMT.
Is it okay to change RMT to fit experimental data?
I tried several times to fit experimental one, but still question myself if
that is okay.
-- next part --
An HTML attachment was scrubbed...
URL: 
<http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20110604/81ae36cc/attachment.htm>

[Wien] Magnetic material relaxation, RMT, MAE calculations

[Gerhard Fecher]

To find the magntic anisotropy is a "brute force" job (see chapter 4.6.4 of the 
textbook of J K?bler, Theory of itinerant electron magnetism),
that means you need really many k-points
(The question is what is "many" ? As answer from experience I would say: If you 
think you have enough then just take a lot more ! 
There is an example in K?blers book where Halilov used 372 248 k-points in the 
full Brillouinn zone.)
So how many k-points did you use ?

There is another thing one should keep in mind, some experimental  values may 
be influenced by the shape anisotropy
of the sample that was used (unfortunately in many experimental work not enough 
information is given).
MnAl has a rich binary phase diagramm, are you shure that the experimental 
values are for a single phase sample ?

What do you mean with experimental RMT ?
There is some older LEED theory where the muffin tin radii were calculated from 
the energy dependence of
low energy electron diffraction pattern. I guess that these have nothing to do 
with the RMT you set in Wien2k.


Ciao
Gerhard


Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz

Von: wien-bounces at zeus.theochem.tuwien.ac.at [wien-bounces at 
zeus.theochem.tuwien.ac.at]" im Auftrag von "Jihoon Park [jpark61 at 
crimson.ua.edu]
Gesendet: Samstag, 4. Juni 2011 04:37
Bis: wien at zeus.theochem.tuwien.ac.at
Betreff: [Wien] Magnetic material relaxation, RMT, MAE calculations

Dear users,


I am really wondering if we have to "relax" structure everytime when 
calculating magnetic materials.
When relaxed, the magnetic properties, especially K value, become too far from 
experimental data; K value calculated from not relaxed structure is also far 
from the experimental one.

I tried force theorem as well, but that is still far from the experimental data 
too.

Everytime my calculated K value is reversed.
For example, energies for <001> and <100> are compared and used to calculate K.
But I do not understand why the energy for <100> is lower than that of <001>.
Experimentally, <001> is stable for MnAl.


And another qeustion is about RMT.
Is it okay to change RMT to fit experimental data?
I tried several times to fit experimental one, but still question myself if 
that is okay.

[Wien] Magnetic material relaxation, RMT, MAE calculations

[Stefaan Cottenier]

> I am really wondering if we have to "relax" structure everytime when
> calculating magnetic materials.
> When relaxed, the magnetic properties, especially K value, become too
> far from experimental data; K value calculated from not relaxed
> structure is also far from the experimental one.

Hard to give any specific comment. This could have several explanations:

1) either your calculation procedure is not accurate enough (k-points? 
very important for MAE)

2) or there is physics going on that is not described in your model

3) or DFT at the LDA/GGA level is not capable of describing this effect 
for this material.

> And another qeustion is about RMT.
> Is it okay to change RMT to fit experimental data?
> I tried several times to fit experimental one, but still question myself
> if that is okay.

See the FAQ (http://www.wien2k.at/reg_user/faq/rmt.html). And even 
safer: accept the values suggested by setrmt_lapw (take the minimal 
values for the relaxed and unrelaxed cases, then you'r sure you can 
cover 'any' situation).

Stefaan

[Wien] Magnetic properties or 5d metals

[Pablo de la Mora]

El 03/06/11 08:06, Peter Blaha escribi?:
> If it gives a small U for a 5d system, it gives the expected result.
>
> The method is reliable if   a)  most of the considered electrons (5d) 
> are localized
> within the atomic sphere and b) is only reliable when "LDA+U is 
> appicable, that means
> if "atomic physics" dominates and states are localized similar as in 
> an atom.
>
> Spin-orbit ??

Thank you, the Madsen and Novak method gives for Ir a value of Uh=0.7, 
but people use Uh=2.

For a metallic system the method seems to require a larger Uh, as you 
mention, the states are not localized.
Yes, SO is includded.

> Am 02.06.2011 14:13, schrieb Pablo de la Mora:
>> Dear WIEN users,
>> Compounds with 5d metals have magnetic properties, but due to the 
>> extended nature of the 5d orbitals the calculations give non-magnetic 
>> results, includding the Hubbard U (Uh).
>> I have calculated the Uh using the method proposed by G. Madsen and 
>> P. Novak:
>> Notes about constraint LDA calculations to determine U 
>> 
>> which gives me a small Uh.
>> To stabilize the magnetic state I need to use an unphysically large 
>> Uh. Any suggestions?
>>
>> Is this method suggested by Madsen and Novak accurate?
>> I have used in other compounds and it seems to give reasonable 
>> results, but these notes are not very clear (I could expand some 
>> issues in the notes to make them clearer).
>>
>>
>>
>> ___
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>

[Wien] Where is Orbital moment?

[Stefaan Cottenier]

> I have calculated LS coupling with dm, but have not found where
> calculated orbital moments are.

With values '1 3' in the last line of case.indmc, you find the orbital 
moments of the orbitals that are specified in case.indmc in case.scfdmup.

Stefaan

[Wien] lapw2

[Stefaan Cottenier]

>>lapw2 -up (17:59:01) Segmentation fault
> 0.084u 0.004s 0:00.08 100.0%  0+0k 0+0io 0pf+0w

Search the mailing list archive for 'segmentation fault' : this has been 
discussed many times before.

Stefaan

[Wien] Volume optimization and Struct File

[Stefaan Cottenier]

> I am working with the TiC example in the user guide. I run the StructGen
> command, the SCF calculations and then do the volume optimization. I
> notice that the case.struct file changes after volume optimization,
> however the lattice parameter in case.struct file doesn't match up with
> the optimized lattice parameter that is given by the .outputeos file.
> If we want to perform further calculations using a volume optimized
> structure, are we supposed to use the updated case.struct file that the
> program creates, or do we need to manually edit the case.struct file to
> provide the optimized lattice parameter given in the .outputeos file?

Lattice parameters in case.struct are changed one after the other, to 
assume the values you have specified during 'x optimize'. If you want to 
do a final calculation for the optimized case, you have indeed to put 
the values from case.outputeos into case.struct.

Stefaan