The ABINIT Users Mailing List ( CLOSED )

[Nicola Marzari <marzari@mit.edu>]

Anurag Chaudhry wrote:

> Is there a way to calculate/estimate the U parameter ?  


Dear Anurag,


I personally like the approach of Cococcioni and De Gironcoli
(PRB 2005) - it's elegant and insightful, based on linear-response
theory, and easy to implement.

Note that in that formulation U (U-J, actually) becomes a
linear-response property of your system. This means that it is
truly an ab-initio value, and not a fitting parameter. At the same
time, U becomes in principle dependent on the chemistry and structure
and electronic ground state you are considering (in practice, not
much variation, but it can become relevant as the bond between your
transition metal and its neighbors is broken).

Also, note that in my experience pseudopotentials generated with
different oxidation states (say, the neutral atom, vs the 2+ or
3+ positive cation) can lead to U's that are very different - even
if the physics/chemistry that they end up reproducing is very similar.

For this last reason alone, I have become hesitant in giving much
weight to the actual values of the U that have been published, although
of course they are often a good ballpark.


                                nicola


---------------------------------------------------------------------
Prof Nicola Marzari   Department of Materials Science and Engineering
13-5066   MIT   77 Massachusetts Avenue   Cambridge MA 02139-4307 USA
tel 617.4522758 fax 2586534 marzari@mit.edu http://quasiamore.mit.edu