The ABINIT Users Mailing List ( CLOSED )

["Anglade Pierre-Matthieu" <anglade@gmail.com>]

Since you are dealing with metals at finite electronic temperature the
band occupancy is not an integer. That is you may have occupancies
like the following:

band 0 : 2 e-
band 1 : 2 e-
band 2 : 2 e-
band 3 : 2 e-
band 4 : 2 e-
band 5 : 2 e-
band 6 : 1.98 e-
band 7 : 0.015 e-
band 8 : 0.005 e-

That's the why the formula I've sent adds a few more bands than just
(nb val. e-)/2.
I suppose that your problem arises from the lake of empty bands.
Anyway you'll need to have at least one band with zero occupancy at
the end of your calculation if you aim at a converged result.

On 4/20/06, jmlv@imr.ac.cn <jmlv@imr.ac.cn> wrote:
> Dear Mr. Anglade Pierre-Matthieu and Nuno Galamba,
>
>    Firstly, thanks for immediate replies from you two guys.
>
>    As having been followed your suggestions, I checked out the 
> pseudopotentials for number of electrons. I found that, the number of 
> electrons in 22-Ti.GGA.fhi file is 4, while that for 28-Ni.GGA.fhi is 10. 
> Accordingly, nband should be (4+10)/2=7. I set nband to 7 and the run went 
> well for a while. But then there was a BUG message coming out at the end of 
> the log file, saying:
>
>   newocc : BUG -
>   In order to get the right number of electrons,
>   it seems that the Fermi energy must be outside the range  of 
> eigenenergies, plus 6 or 30 times the smearing, which is strange.
>   The calling routine gives nelect=  0.14000000D+02
>   The lowest bound is  -0.16987487D+01, with nelect=  0.00000000D+00
>   The highest bound is   0.12421480D+01, with nelect=  0.14000000D+02.
>   Action : contact ABINIT group.
>
>   Now I am confused again. I've sent a BUG report about this to Prof. 
> Xavier Gonze, hoping for a solution from him.
>
>   Also I hope you guys help me out. Thanks again for you time.
>
> Jianmin Lu
> imr, cas, China
>


--
Pierre-Matthieu Anglade