The ABINIT Users Mailing List ( CLOSED )

[matthieu verstraete <matthieu.jean.verstraete@gmail.com>]

On Thu, Feb 19, 2009 at 3:37 AM, <6671011@163.com> wrote:

Dear Matthieu,

While calculating the band structure (BS) of a 1D semiconductor (Eg ~0.4 eV), I
found the following strange behaviors:

In 1D things are much simpler, and the BS and GS kpoint grids are basically identical (the BS is simply denser, it doesn't need to follow high symmetry lines etc...). There isn't really a need to use the GS Ef in this case, you can just count the bands and see where the valence band edge is. Why are you worrying about it?
 

(1) In Abinit 5.6.4 (tsmear=0)

               plot of GS-Eigenvalues --->  semiconductor,
               plot of BS-Eigenvalues --->  metal (obviously wrong).

you mean it is a metal according to the position of the GS Ef? Normally the BS run should not output a Fermi energy (maybe 0 or something). If it does it will be calculated based on the BS kpoints only.

What is probably happening is the following: your GS kpoint grid is sparse, and does not contain a kpoint which is at the band edge of the semiconductor. With occopt 0, abinit outputs the maximum of the valence band, not the Fermi energy (this is written in the output), and as a result this value is lower than the true valence band edge, which you then get correctly out of the BS run. If you know, and can check with the BS, that the system is not metallic, there is no reason to use the Fermi energy at all (as in CASTEP). You could also use the middle of the gap, calculated from the GS, which should be precise enough and inside the real gap as well.

 
M.

To solve the problem, I have to use much denser kpoints in the GS.


(2) In CATSEP (included in the GUI---Materials Studio (MS)), I used the exactly
same kpoints and similar other parameters, but got:

               plot of GS-Eigenvalues --->  semiconductor,
               plot of BS-Eigenvalues --->  semiconductor.

------------------

The above difference confused me a lot. Today, after your reply, I did some
further tests and got the answer. Actually, CATEP give almost the same original
results as Abinit, but the MS does some extra process and causes the above
difference. The only file used by MS in Band Structures Analysis is *.bands,
which contains the following information:

               A.  Number of electrons in a cell
               B.  Fermi energy of GS
               C.  Eigenvalues

The trick used by MS is: If A and C deduce a semiconductor, then ignore B and
get a new Fermi energy according to A and C; Or else, use B and C.

-----------------


Sincerely,
Gunagfu Luo


----------------------------------------------------------------------------------------------------

CASTEP should be calculating the EF in the same way. What do you mean that it
is not good enough? The non-SCF GS is a good way of getting a more precise EF,
and I would also decrease tsmear (also make sure that this is identical btw
abinit and CASTEP - it is the most influential factor).

Matthieu


----------------------------------------------------------------------------------------------------
On Sat, Feb 14, 2009 at 6:46 PM, <6671011@163.com> wrote:

Dear Matthieu,

Thanks for explanation.

The problem I met is that the EF in the SCF-GS is not accurate enough for the
BS. So, according to your reply, the only reasonable solution is improving the
SCF-GS calculation?  Or maybe I could add a non-SCF GS calculation between the
SCF-GS and the BS calculation?

p.s. I am dealing with 1D system. And there is no such problem when I work with
CASTEP at almost the same accuracy level.


Sincerely,
Guangfu Luo


----------------------------------------------------------------------------------------------------


The Fermi energy should be taken from the full GS calculation, not the BS,
whose k-points are not weighted and distributed evenly throughout the Brillouin
Zone. Also, there is no need to calculate the Fermi energy as the density is
already fixed. This is why EF is not output for the non scf case.

Matthieu


----------------------------------------------------------------------------------------------------

On Sat, Feb 14, 2009 at 1:31 AM, <6671011@163.com> wrote:

Dear Friends,

 I hope to obtain the Fermi Energy in the Band Structure calculation.
However, Abinit does not export it in the default situation, though it does
that in the GS calculation.

 Is it possible to get it easily by setting some parameter? (I can get the
Fermi Energy according to its definition in the Band Structure, but probably
Abinit has this function already.)


Thanks a lot.

Sincerely,
Guangfu Luo

-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dr. Matthieu Verstraete

European Theoretical Spectroscopy Facility (ETSF) 
Dpto. Fisica de Materiales, 
U. del Pais Vasco,            
Centro Joxe Mari Korta, Av. de Tolosa, 72,   Phone: +34-943018393
E-20018 Donostia-San Sebastian, Spain        Fax  : +34-943018390

Mail : matthieu.jean.verstraete@gmail.com
http://www-users.york.ac.uk/~mjv500