ABINIT Discussion Forums

Hello,

I would like to know if the implementation of the linear response method to calculate U has been tested in comparison with other codes. This method is also implemented in Quantum Espresso (QE).
According to the abinit tutorial, U for bcc Fe is ca. 7.9eV whereas I get 2.2eV from QE. Even if XC potentials and different potentials are different, the difference in U seems to be too large.
Can I rely on the values provided by abinit? What could be the reason for this huge discrepancy?

Thank in advance,

Ider

Hi

I tested it on actinide dioxides and it worked rather well up to Am. There was a 0.3-0.5 eV difference compared to the experimental based value.

To my knowledge the implementation is not fully friendly user yet. You have to do some maths to interpret the results yielded by abinit. For instance for bcc Fe you should not use directly the value given, which is too large.

It is still in development.

Boris wrote:
You have to do some maths to interpret the results yielded by abinit. For instance for bcc Fe you should not use directly the value given, which is too large.

.

Hi, Thank for your reply!

What do you mean with "do some maths to interpret the results"? The tutorial does not mention at all that the value for bcc Fe is somehow too high. I am not sure if the developer are aware of that.

I don't know much, just that the developers are aware of that and they're working on making the implementation more userfriendly.

Dear all,
according to my experience that part of the code is still to be considered not stable: I got segmentation fault errors trying to determine U (and J) for FeO, Fe2O3 and LiFePO4.
Moreover I'm not sure that you can compare data obtained using Abinit and QE because the two implementatons could not be fully equivalent.
Mauro.

Thank you for yours answers!