[CP2K-user] DFT+U with Ti and O

[kazimieras tamoliunas]

Hello,
 i'm calculating Na(+1)Ti2(+4)(PO4)3(-3) using DFT+U method and the system 
is quite big ( 108 atoms, calculating 2x2x1, so 432 atoms in total) and i'm 
trying to fit the U parameter. Now i'm using U for both Ti and O (obviously 
Ti d orbitals). However, what's less obvious is what kind of orbitals 
should i apply the U correction to, regarding oxygen? Whenever i apply L=0 
or L=1 ( p sounds reasonable to me) it prints out a following message:

*** 04:19:11 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!

Now when i set L=2 for O, the warning disappears and i get a great 
convergence. 
That i do like, however, i cannot understand how this method works. I 
suppose i have oxygen mostly -2, therefore it should have it's 2p fully 
filled. From my point of view, if i want to localise the electron i should 
think about 3s or 2p, but certainly not 3d orbital. Can somebody help me 
with that?
Appreciate any help, comments or possible literature from the ones familiar 
with the problem!

PS. i know that these oxidation states are not correct and you cannot 
precisely have them in a periodic system. However, i suppose, that DFT+U 
method is used for certain electron localisation, therefore, i feel that 
they provide some ground to rely upon. Correct me if i'm mistaken!!

Regards,
Kazimieras
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190617/7748eab0/attachment.htm>