My Community

Hi,

One vasp user (see Feb 01 2005 in bug report forum) has already reported this.

For non-magnetic bulk calculation with LDA+U, I used

LDAUTYPE =4

(ISPIN =1 default) and a particular value of U & J (from the online manual double counting correction for LDA+U containts J so J needs to be given even for non-magnetic runs, although not conceivable why for non-magnetic system one needs J).

But the run stops before the first electronic SCF starts.

In the Feb 01 2005 reply to the vasp user, admin suggested to use ISPIN =2 with LDAUTYPE =4.

But is n't ISPIN=2, LDAUTYPE =4 means one is doing SPIN-POLARIZED calculation for non-magnetic system?

Also the vasp user (Feb 01, 2005 query) said ISPIN=1 and LDAUTYPE= 2 worked for his non-magnetic calculation. If
Dudarev's apparoach (LDAUTYPE =2 ) is for spin-polarized system, then how ISPIN=1, LDAUTYPE=2 worked for him/her?

Sahu

Dear Sahu,

in general for the two DFT+U approaches implemented in VASP you have to
set ISPIN=2. The Lichtenstein technique as well as the Dudarev approach are
formulated such that the electron spin, i.e. the z-component, is explicitely included. Hence, you have to perform spin-polarized calculations and all DFT+U calculations with ISPIN=1 are useless. When you describe a system with DFT+U you introduce by definition magnetic moments and impose a particular ordering of the moments although a system might be non-magnetic. For example, most of the physical and structural properties of delta-Pu are well described by DFT+U. However, delta-Pu is non-magnetic
and DFT+U imposes magnetism artificially.
The only difference between the DFT+U methods defined by LDAUTYPE=1 and LDAUTYPE=2 is which double counting correction is used. Thats all.

René Windiks