DFT+U method LOWDIN Warning: Forces are not yet fully implemented

[Vyacheslav Bryantsev]

Dear CP2K Developers,

I am writing to check the meaning of the following warning:

*** WARNING in dft_plus_u.F:785 :: Forces are not yet fully implemented ***
 *** for DFT+U method LOWDIN    

The MD simulation is running with the warning message above.

How are the forces  estimated at each MD step? Can I trust these calculations?

I note that using DFT+U method MULLIKEN generates another warning:

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

Therefore I switched from MULLIKEN to LOWDIN method for DFT+U calculations that are stated in the manual to be more robust.

Thank you,

Slava Bryantsev

[Krack Matthias (PSI)]

Hi Slava

 

the forces for DFT+U with Lowdin are wrong. An MD run should not continue with that choice. I will fix that. On the hand, an MD using DFT+U with Mulliken should work fine in spite of the warning about unphysical charges.

 

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.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/f157f7fd-af6f-4b57-8e40-d11afe7aa802%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Vyacheslav Bryantsev]

Hi Matthias,

Thank you very much for your notes.

Slava

On Saturday, May 18, 2019 at 1:01:46 PM UTC-4, Matthias Krack wrote:

Hi Slava

 

the forces for DFT+U with Lowdin are wrong. An MD run should not continue with that choice. I will fix that. On the hand, an MD using DFT+U with Mulliken should work fine in spite of the warning about unphysical charges.

 

Matthias

 

Von: cp...@googlegroups.com <cp...@googlegroups.com> Im Auftrag von Vyacheslav Bryantsev
Gesendet: Freitag, 17. Mai 2019 20:13
An: cp2k <cp...@googlegroups.com>
Betreff: [CP2K:11752] DFT+U method LOWDIN Warning: Forces are not yet fully implemented

 

Dear CP2K Developers,

 

I am writing to check the meaning of the following warning:

 

*** WARNING in dft_plus_u.F:785 :: Forces are not yet fully implemented ***
 *** for DFT+U method LOWDIN    

 

The MD simulation is running with the warning message above.

How are the forces  estimated at each MD step? Can I trust these calculations?

 

I note that using DFT+U method MULLIKEN generates another warning:

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

 

Therefore I switched from MULLIKEN to LOWDIN method for DFT+U calculations that are stated in the manual to be more robust.

 

Thank you,

Slava Bryantsev

--
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 cp...@googlegroups.com.