Physics of LOCPOT

[protonic]

This seems like an embarrassingly basic question, but I just thought I have to ask this anyway to be 100% clear: What exactly does the LOCPOT file give?

More precisely, what is meant by "total local potential" in the LOCPOT section of the VASP manual? Is this simply the effective potential under Kohn-Sham DFT (without XC), i.e. Eq. (2.6) of Kohn and Sham, Phys. Rev. 140 (1965) A1133? If not, I would appreciate it if someone can kindly point out a paper or book section that explicitly explains the Physics behind and/or how the LOCPOT file values are calculated. Thanks.

[admin]

by default, the local potential as given in LOCPOT does not include the XC-contributions, because they are converge very slowly into the vacuum.
If you want to keep the XC-contributions as well, please comment the line
IF (IO%LVHAR) CALL SET_LEXCH(-1) (vasp.5.2)
or
INFO%LEXCHG=-1 (vasp.4.6)
before the
CALL POTLOK
line in main.F

[protonic]

Thanks admin for the prompt response. That part for vasp 5.2, to my knowledge, isn't in the latest VASP manual yet, so that will be useful.

Can I assume then that the answer to my previous question "Is the LOCPOT simply Eq. (2.6) of Kohn and Sham, Phys. Rev. 140 (1965) A1133?" is a resounding "Yes"? (Sorry -- I really just need an explicit yes/no response to this question to help me sleep tonight.)

[tlchan]

The local part of the self-consistent potential refers to the sum of the local part of the pseudopotentials, Hartree potential, and exchange-correlation potential. It is called local because the potentials act on the wave function by a simple multiplication. The non-local part of the pseudopotential is not printed out in LOCPOT. What the admin is telling you is that LOCPOT does not print out the exchange-correlation potential by default. Unless you modify the code and recompile as above, LOCPOT only contains the local electrostatic potential.

[juhL]

hi tIchan, nice post, however to make sure, i'm asking once more in detail:

based on the review from kresse DOI: 10.1103/PhysRevB.59.1758

(http://link.aps.org/doi/10.1103/PhysRevB.59.1758)

we have the one-electron potential v_{eff}=v_H + v_{XC} from Eq. (43)

here v_H is the coulomb like interaction of the valence electrons (n~+n^) aswell as the pseudo cores (n~_{Zc}) and the exchange correlation potential, which is not written (at least by default) into the LOCPOT file.

i am not sure, what else is written into the LOCPOT file, because the manual says on the one hand: everything but v_{XC}, but since v_{eff} - v_{XC} = v_H, but on the other hand it is possible to write v_H only, by setting VHAR = .TRUE.

when i'm comparing the normal LOCPOT file with the potential i get when setting VHAR=.TRUE. they are different, in addition in the VHAR case the potential looks reasonable, if one expects v_H to be the coulomb part of electron&electron AND electron&core interactions, as it is defined in this formalism (the AND because in other formalisms, v_H is only the coulomb part of the electron&electron interaction)

this leads to the question: what else is written into the LOCPOT file?

you say "the local part of the pseudopotentials"

with this, do you mean v^1_eff and v~^1_eff from Eq. (45) and (46), or more precise, just the v_H contributions of these potentials ?


(i'm sry to put this Eq. references, but it just seems not possible to show the formula in readable shape otherwise)

[juhL]

hi tIchan, nice post, however to make sure, i'm asking once more in detail:

based on the review from kresse DOI: 10.1103/PhysRevB.59.1758

(http://link.aps.org/doi/10.1103/PhysRevB.59.1758)

we have the one-electron potential v_{eff}=v_H + v_{XC} from Eq. (43)

here v_H is the coulomb like interaction of the valence electrons (n~+n^) aswell as the pseudo cores (n~_{Zc}) and the exchange correlation potential, which is not written (at least by default) into the LOCPOT file.

i am not sure, what else is written into the LOCPOT file, because the manual says on the one hand: everything but v_{XC}, but since v_{eff} - v_{XC} = v_H, but on the other hand it is possible to write v_H only, by setting VHAR = .TRUE.

when i'm comparing the normal LOCPOT file with the potential i get when setting VHAR=.TRUE. they are different, in addition in the VHAR case the potential looks reasonable, if one expects v_H to be the coulomb part of electron&electron AND electron&core interactions, as it is defined in this formalism (the AND because in other formalisms, v_H is only the coulomb part of the electron&electron interaction)

this leads to the question: what else is written into the LOCPOT file?

you say "the local part of the pseudopotentials"

with this, do you mean v^1_eff and v~^1_eff from Eq. (45) and (46), or more precise, just the v_H contributions of these potentials ?


(i'm sry to put this Eq. references, but it just seems not possible to show the formula in readable shape otherwise)

[tlchan]

The VHAR and LOCPOT is different because VHAR contains the Coulomb potential due to the electrons, but LOCPOT contains both the Coulomb potential from electrons and atomic ions. The contribution from the ions should come from the ionic pseudopotential. Only the local part of the ionic pseudopotential is written out because the non-local part is an operator.

I believe your confusion comes from reading the PRB that you cited. In the PRB paper, the Hartree potential comes from the total charge density, including both the electrons and atomic ions, which is different from the VHAR file.

[juhL]

hi,

first: thanks for your reply

second: from some test calculations i've done in the past, i tend to disagree with you:

i compiled vasp versions, where the potentials calculated in the pot.F routine, are set to 0 after calculation, before adding to vtot.

from this i can see:
if v_XC, v_H and v_ION is set to zero, i get a zero potential, so there is no additional part of the potential i'm over looking
for v_H only and for v_ION only, contributions are written into LOCPOT for both cases VHAR and VTOT = TRUE and the contributions are identical

last but not least, for v_XC only, a nonzero potential is written only for LVTOT = TRUE

finally i can add up the potentials for v_H only + v_ION only and get the exact potential as the potential of the original vasp for LVHAR = TRUE

the same works for adding the potentials of v_H only + v_ION only
+ v_XC only which gives the exact potential for LVTOT = TRUE of the original code


to sum up we have:

v_H + v_ION for LVHAR and v_H + v_ION + v_XC for LVTOT

although it is not written in the manual like this, i think my tests proof, that that's what written for the both cases (at least for vasp 5.2.11)
<span class='smallblacktext'>[ Edited Mon Jun 27 2011, 01:41AM ]</span>

[tlchan]

I didn't goto the code to verify the output. If you have tested it, I believe that you are right.

There should be no confusion about LVTOT. The tag is to write out the total local self-consistent potential. The issue is whether VHAR writes out the total electrostatic potential (which includes the electronic and ionic contribution), or just the electronic contribution of the electrostatic potential (which is the Hartree potential). It is quite misleading for the VHAR tag to write out not just the Hartree potential.

[juhL]

yep, i agree with you that it is indeed misleading :-)

however, thank you for your time and the communication.