Calculation of U (DFT+U)

Message

aarondesk · #1 Post by **aarondesk** » Tue Jul 18, 2006 8:12 pm

I'm looking at a paper on calculating the U term in the DFT+U method.

PRB 71, 035105 (2005).

The idea is to perturb the occupation of the electrons at a particular atom by applying a potential to the l orbitals of interest on that site. For example, getting the U term for the d electrons would mathematically be given as the sum of these derivatives:

U = d(alpha)/dn(SC) - d(alpha)/dn,

where alpha is the applied potential on the d orbitals and n is the density of the d orbitals. The first derivative is from a self-consistent calculation, while the second isn't.

Does anyone have experience with this method and VASP? How is a potential applied at a particular atomic site and orbital? What is the best way to get the orbital specific density?

Thank you.

#2 Post by **admin** » Fri Jul 21, 2006 10:23 am

1) please set LDAUTYPE=3 in INCAR
2) the orbital-specific densities are given in the output if you set LORBIT (please have a look to the manual for the appropriate setting)

jcconesa · #3 Post by **jcconesa** » Sat Oct 07, 2006 4:14 pm

1) LDAUTYPE=3 is not documented in the VASP guide. What is its use?
2) Is there a way of introducing the alpha value to VASP in order to put this method in practice?
3) Such method is actually implemented already in Espresso. It would be a nice thing to have it in VASP...

askhetan · #4 Post by **askhetan** » Mon Oct 29, 2012 2:41 pm

I have been wondering over the exact question of the meaning of LDAUTYPE=3. Could you please tell know what it is ?

icall · #5 Post by **icall** » Tue Oct 30, 2012 12:26 am

I also want to know if LDAUTYPE = 3 can be used to calculate U.