DFT+U Calculation

[sudhir kumar Sahoo]

Dear cp2k users,

I am using DFT+U method for modelling CuInSe2 (U=5.0 eV for 3d electrons of Cu), particularly, using Mulliken method as implemented in cp2k program.
However, a warning is repeatedly written in the output file. 

 *** WARNING in dft_plus_u.F:1455 :: DFT+U energy contibution is negative ***
 *** possibly due to unphysical Mulliken charges. Check your input, if    ***
 *** this warning persists or try a different method!                     ***

After reading the cp2k google group discussion, I have tried to perform DFT+U calculations as explained below,
but each time I got the same warning in the output file.

1) Restarting the calculation U=5 eV, from an optimized wavefunction obtained from U=0 eV.
2) Point 1 + u_ramping option
3) Point 2 + &enforce_occupation

Can someone please help me in this.

Thanking You,
Sudhir

[Krack Matthias (PSI)]

Dear Sudhir

 

This warning pops up when the Mulliken population analysis (MPA) returns an orbital occupation larger than the maximum number of electrons allowed for the actual l value specified in the +U section, e.g. more than 10 electrons are assigned by the MPA to the Cu 3d (l=2) orbitals. This can be an indicator for a too large U value enforcing a too high on-site localisation. Please, note that the U(eff) values successfully employed by PW codes are not transferable to CP2K. The corresponding U values for CP2K are usually smaller, roughly by a factor of two, e.g. with respect to their impact on the band gap. You can specifically print the detailed orbital occupations of the +U atoms using this CP2K print key with “QS_SCF 1” in its &EACH section. The warning appears easily for systems like ZnO with a 3d10 configuration when the MPA assigns for instance 10.04 electrons to the Zn 3d orbitals of some Zn atoms. The warning can then be ignored possibly for such spurious violations of the maximum orbital occupation. It is up to the user eventually to decide, if the obtained results are still physically meaningful. There is always the option to employ another method when DFT+U troubles, e.g. DFT with hybrid functionals.

 

Best regards

 

Matthias

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

[Sudhir K. Sahoo]

Dear Matthias,

Many thanks for your reply and explanation. I have done one calculation with a lower value of U (2 eV), but got the same warning. 

Below is the input for DFT+U. 

       &DFT_PLUS_U  T
         L  2
         U_MINUS_J [eV] 2.0
         INIT_U_RAMPING_EACH_SCF
         U_RAMPING [eV] 0.10
         &ENFORCE_OCCUPATION
            EPS_SCF 1.0E-8
            MAX_SCF 10
            ORBITALS -2 -1 0 1 2
            SMEAR
         &END ENFORCE_OCCUPATION
       &END DFT_PLUS_U

I will give a try to print the detail orbital occupation.

Thanks,

Sudhir

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/t0tJh9KSCBs/unsubscribe.
To unsubscribe from this group and all its topics, send an email to cp2k+unsubscribe@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--

Sudhir Kumar Sahoo
C/O - Prof. Thomas D. Kühne
Post-doc Researcher

Department of Chemistry

University of Paderborn

[Krack Matthias (PSI)]

Dear Sudhir

 

You enforce a smearing of the projected d orbital occupations until an SCF convergence threshold of 1.0E-8 is reached (and for at least 10 SCF iteration steps). This will imply that the occupation constraint is never released during the SCF iteration loop unless your SCF convergence limit is even smaller than 1.0E-8 (well, you possibly never get there). That is also most likely the reason for the persistence of the warning. I suggest to comment/remove the EPS_SCF 1.0E-8 line. After 10 SCF iteration steps the SCF cycle will then evolve without occupation constraint. You may even set “&ENFORCE_OCCUPATION off” and drop the occupation constraint completely, since you do not want to enforce any specific d orbital occupation pattern obviously as you are smearing over all d orbitals anyway. You may also drop the U_RAMPING unless that is crucially needed to achieve convergence.

 

Best regards

 

Matthias

--

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/t0tJh9KSCBs/unsubscribe.

To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--

Sudhir Kumar Sahoo
C/O - Prof. Thomas D. Kühne
Post-doc Researcher

Department of Chemistry

University of Paderborn

 

--

You received this message because you are subscribed to the Google Groups "cp2k" group.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.

[Sudhir K. Sahoo]

Dear Matthias,

You are correct. The occupation of d-orbital is larger than 10 when large value of U is used,
and occupation for d-orbital is less than 10 only when value of U is very low (=1 eV).  

Do you think, this small value of U would be sufficient. 

Thanks,

Sudhir

--

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/t0tJh9KSCBs/unsubscribe.

To unsubscribe from this group and all its topics, send an email to cp2k+unsubscribe@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--

Sudhir Kumar Sahoo
C/O - Prof. Thomas D. Kühne
Post-doc Researcher

Department of Chemistry

University of Paderborn

 

--
You received this message because you are subscribed to the Google Groups "cp2k" group.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/t0tJh9KSCBs/unsubscribe.

To unsubscribe from this group and all its topics, send an email to cp2k+unsubscribe@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

[Krack Matthias (PSI)]

Dear Sudhir

 

I have no clue, since I don’t know your system. I suggest to calculate some structural or electronic properties of your system like lattice constants or the band gap for different U values and to compare these to experimental or theoretical literature data to get confidence concerning your model setup.

--

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/t0tJh9KSCBs/unsubscribe.

To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--

Sudhir Kumar Sahoo
C/O - Prof. Thomas D. Kühne
Post-doc Researcher

Department of Chemistry

University of Paderborn

 

--
You received this message because you are subscribed to the Google Groups "cp2k" group.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/t0tJh9KSCBs/unsubscribe.

To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.

To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

--

Sudhir Kumar Sahoo
C/O - Prof. Thomas D. Kühne
Post-doc Researcher

Department of Chemistry

University of Paderborn

 

--
You received this message because you are subscribed to the Google Groups "cp2k" group.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.