QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: postnikov on September 13, 2011, 08:39

Title: The error about repeating previous PRL result.
Post by: postnikov on September 13, 2011, 08:39
The error about repeating previous PRL result.

These days, I repeat the calculation results of PRL 97,09201(2006).
My obtained magnetic moment per unit cell in the one-dimensional vanadium-benzene wire system is
about 0.55ub,however, the above PRL paper show the magnetic moment should be 1 ub unit cell.

Is it my own input file error,or the atk code itself?

Could someone give me some advice?

The input file is attached!
Title: Re: The error about repeating previous PRL result.
Post by: zh on September 13, 2011, 16:06
1). You had better use a same exchange correlation functional with the one used in literature.
2). In the setup of initial spins of atoms, only the vanadium atom is required.
Title: Re: The error about repeating previous PRL result.
Post by: postnikov on September 14, 2011, 10:45
1). You had better use a same exchange correlation functional with the one used in literature.
2). In the setup of initial spins of atoms, only the vanadium atom is required.

I have change the XC function into GGA, and the inital spin is set only for V, the

result is very similar to the previous calculation, ie, M is also 0.6 ub

 ATK developers could like to check this case for me?
Title: Re: The error about repeating previous PRL result.
Post by: zh on September 14, 2011, 12:45
I also checked it using the  Atomistix ToolKit 11.2.2. The total magnetic moment is about 0.6 uB. The density of states gives a metallic ferromagnetism, rather than the half-metallic behavior. The pseudopotential of vanadium provided in the the code may be problematic, and it needs to be improved.
Title: Re: The error about repeating previous PRL result.
Post by: Anders Blom on September 14, 2011, 13:47
@zh: Do not use ATK 11.2.2 for anything (or 11.2.1 or 11.2.0 for that matter). It contains bugs and has been replaced by 11.2.3.

@postnikov: The article reference in your first post is wrong, so I cannot check the original results.
Title: Re: The error about repeating previous PRL result.
Post by: zh on September 14, 2011, 13:49
The reference:
Volodymyr V. Maslyuk et al, Organometallic Benzene-Vanadium Wire: A One-Dimensional Half-Metallic Ferromagnet, Phys. Rev. Lett. 97, 097201 (2006)
Title: Re: The error about repeating previous PRL result.
Post by: Anders Blom on September 14, 2011, 13:54
Thanks, I also just found it via Google :) I will see if I can check the results. The basis set of V look ok at a first glance, but that's no guarantee.
Title: Re: The error about repeating previous PRL result.
Post by: Anders Blom on September 14, 2011, 16:00
I can confirm ATK predicts a metallic ferromagnetic state. The DOS resembles that in the article (you need 200 kz-points or so) but the majority states are slightly shifted so there is finite DOS at the Fermi level.

This might be a basis set effect, the article uses a rather different basis set. Maybe with LDA+U things would change a bit in ATK? They try this in the article without seeing a major effect, but perhaps ATK needs it to shift the bands slightly? You can try this!
Title: Re: The error about repeating previous PRL result.
Post by: postnikov on September 14, 2011, 16:26
I can confirm ATK predicts a metallic ferromagnetic state. The DOS resembles that in the article (you need 200 kz-points or so) but the majority states are slightly shifted so there is finite DOS at the Fermi level.

This might be a basis set effect, the article uses a rather different basis set. Maybe with LDA+U things would change a bit in ATK? They try this in the article without seeing a major effect, but perhaps ATK needs it to shift the bands slightly? You can try this!


How about the  magnetic moment?
Can you show your input file ?
Title: Re: The error about repeating previous PRL result.
Post by: Anders Blom on September 14, 2011, 16:34
My Mulliken populations show Q(up)-Q(down) about 0.7, I'm just trying to wrap my head around if there is a factor of 2 for the Bohr magneton, as that would put it around 1.4 as in the article...?

Input attached; I also tried the hexagonal configuration they use in the article, but it makes no difference.
Title: Re: The error about repeating previous PRL result.
Post by: postnikov on September 14, 2011, 17:37
My Mulliken populations show Q(up)-Q(down) about 0.7, I'm just trying to wrap my head around if there is a factor of 2 for the Bohr magneton, as that would put it around 1.4 as in the article...?

Input attached; I also tried the hexagonal configuration they use in the article, but it makes no difference.


This is my siesta calculation result.

Q(up)-Q(down) =0.9993

In my SIESTA calculation, V 3p is considered as the semicore state in the used PP and basis.

How about the ATK code?
Title: Re: The error about repeating previous PRL result.
Post by: Anders Blom on September 14, 2011, 17:47
In the pp we have in ATK right now, 3p is pure core. You can try the SIESTA pp in ATK, as long as it's a UPF file with the right format and no off-diagonal projectors. You do however have to make a customized basis set then to include 3p. Not too hard, the pp is the tricky bit.

My DOS is similar, just shifted a bit. Is your majority DOS @ Ef zero or finite?
Title: Re: The error about repeating previous PRL result.
Post by: postnikov on September 14, 2011, 17:54
In the pp we have in ATK right now, 3p is pure core. You can try the SIESTA pp in ATK, as long as it's a UPF file with the right format and no off-diagonal projectors. You do however have to make a customized basis set then to include 3p. Not too hard, the pp is the tricky bit.

My DOS is similar, just shifted a bit. Is your majority DOS @ Ef zero or finite?


The majority DOS @ Ef is zero, as shown in the above fig.

Could you provide the modified V pp and basis including 3p for me and other ATK users.

As you know, the Up-Down is about 0.6 or 0.7, it is very different from the result published in Phys. Rev. Lett. and the SIESTA code.

So if the modified V pp and basis are provided, the other people can also adopt them in their research work about V.
Title: Re: The error about repeating previous PRL result.
Post by: Anders Blom on September 14, 2011, 18:02
Ok (was hard to tell if it was just small or zero). But the band edge is very close to Ef, closer than in the PRL I think, so I guess the V core/basis set matters a lot.
Title: Re: The error about repeating previous PRL result.
Post by: postnikov on September 15, 2011, 19:46
Ok (was hard to tell if it was just small or zero). But the band edge is very close to Ef, closer than in the PRL I think, so I guess the V core/basis set matters a lot.


How about the PP and basis?
Title: Re: The error about repeating previous PRL result.
Post by: kstokbro on September 16, 2011, 23:57
We are working on trying to find a pseudopotential and basis for you, but it might be a little delayed since we are also busy finalizing the release.
Title: Re: The error about repeating previous PRL result.
Post by: postnikov on September 18, 2011, 03:37
We are working on trying to find a pseudopotential and basis for you, but it might be a little delayed since we are also busy finalizing the release.



No problem!

Thank all of you for your work!
Title: Re: The error about repeating previous PRL result.
Post by: kstokbro on September 21, 2011, 23:14
please find attached a gga semicore pseudopotentials for Vanadium
it is from M. Krack, see Theor. Chem. Acc. 114, 145 (2005) and converted to the upf format. There is also an example of how to setup the basis set.

The inclusion of semicore  states gives a large difference for the gga calculations for the present system. As you can see in the attached
the momen on V is 1.25, and the total moment of the molecule is  0.99.

We plan to include all the pseudo potentials from the above paper in the final release of ATK11.8, i.e. an additional set of  semi-core and only valence pseudo potentials for LDA and GGA.
Title: Re: The error about repeating previous PRL result.
Post by: postnikov on September 22, 2011, 04:01
please find attached a gga semicore pseudopotentials for Vanadium
it is from M. Krack, see Theor. Chem. Acc. 114, 145 (2005) and converted to the upf format. There is also an example of how to setup the basis set.

The inclusion of semicore  states gives a large difference for the gga calculations for the present system. As you can see in the attached
the momen on V is 1.25, and the total moment of the molecule is  0.99.

We plan to include all the pseudo potentials from the above paper in the final release of ATK11.8, i.e. an additional set of  semi-core and only valence pseudo potentials for LDA and GGA.

Thank all of you for your hard work!
Title: Re: The error about repeating previous PRL result.
Post by: postnikov on September 22, 2011, 04:21
please find attached a gga semicore pseudopotentials for Vanadium
it is from M. Krack, see Theor. Chem. Acc. 114, 145 (2005) and converted to the upf format. There is also an example of how to setup the basis set.

The inclusion of semicore  states gives a large difference for the gga calculations for the present system. As you can see in the attached
the momen on V is 1.25, and the total moment of the molecule is  0.99.

We plan to include all the pseudo potentials from the above paper in the final release of ATK11.8, i.e. an additional set of  semi-core and only valence pseudo potentials for LDA and GGA.

kstokbro, thank you for your work very much.
However, in our group the used atk version is 2008.10, so could you provide me the corresponding basis for the atk 2008.10?
Thank again!
Title: Re: The error about repeating previous PRL result.
Post by: kstokbro on September 22, 2011, 10:08
ATK2008.10 is no longer supported, and I strongly recommend upgrading to the latest version.
The latest version is better in all respects, accuracy, convergence, ease of user, number of features.

You can get a free trial version of the latest version and test it for 30 days and upgrade by paying a maintainance fee.

I see that I attached wrong input and output files, please find the correct below.
Title: Re: The error about repeating previous PRL result.
Post by: Anders Blom on November 18, 2011, 10:45
I have turned this discussion into a mini-tutorial, showing how to reproduce the results from the article with ATK 11.8: http://quantumwise.com/publications/tutorials/mini-tutorials/141.