Dear Andrei Postnikov,
thank you very much for your
answer (one more time!).
what I mean when I said that that the results did not change, was that we
used a denser k-mesh along gamma to X (600 points) and it was still very
difficult to decide how the bands were connected.
Thus, I guess that we need to do all the stuff, analyze the orbital
character of the bands (something difficult), denser k-mesh, check the DOS,
although as you said this won`t give 100 % guarantee.
In others PW codes it is not different because of the plane-waves. The
difference is that you can use symmetry, and that can help you.
Many thanks for you help,
pablo
----- Original Message -----
From: <apost...@uni-osnabrueck.de>
To: <SIESTA-L@listserv.uam.es>
Sent: Tuesday, April 14, 2009 7:11 PM
Subject: Re: [SIESTA-L] band mixing problem with siesta
Dear Pablo,
I see from your data that the crossing/no crossing happens
between the argument values 0.020781 and 0.24937:
.016625 -.069279
.020781 -.026979 ( up to here increasing )
.024937 -.040879 ( from here on decreasing )
.029093 -.077879
......................
.016625 .034121
.020781 .003579 ( up to here decreasing )
.024937 .014821 ( from here on increasing )
.029093 .056121
In my understanding, you'll see it with better "resolution"
if using a more dense k-mesh along the Gamma-X path.
In this sense, I don'understand how it can be that, as you write,
We have (...) increased the number of points used from gamma to X
and the results did not change.
What exactly did not change? You'll have different argument values now
and new function values corresponding to them!
On the contrary, it is clear that whatever you tried otherwise
We have increased the number of k points, mesh cut-off, orbital confining
cut-off,
won't have effect on how you see the bands "connected" or not
(unless these experiments distort the band structure completely).
In principle, there is no rocksolid way to see how the bands are
connected UNLESS you take a limit of dense enough k-mesh
(along the path you are interested in), in order to "resolve"
a crossing from an avoided crossing. You can analyze the orbital
character of bands before and after the suspected crossing
but even this won't give the 100% guarantee because the orbital
composition may change through the crossing.
In order to distinguish a zero-gap semiconductor, I'd suggest,
explore some vicinity of the suspected region...
In practical sense, in a calculation with probably any code,
the band energies come out numbered in the order of their
increasing energies, separately in each k-points.
So - sorry - I fail to see how it can be
different in a PW code you refer to. Could you please explain it better?
Best regards
Andrei Postnikov
Dear Siesta users,
We are having some problems with the band
structure. The face this problem when we
have metallic systems which have for example
2 bands which crosses the fermi level ( a
(n,n) SWCNT for example). When we use
gnubands to plot get the bands.dat, these
two bands are mixed. In the case of a (5,5)
SWCNT we get (from gamma to X)
.000000 -.241079
.004156 -.198779
.008312 -.155179
.012468 -.111979
.016625 -.069279
.020781 -.026979
.024937 -.040879
.029093 -.077879
.033249 -.114579
.037405 -.150979
.041561 -.186979
.045718 -.222679
.049874 -.258079
.054030 -.293179
.058186 -.327879
.062342 -.362379
.066498 -.396379
.070654 -.430179
.074811 -.463579
.078967 -.496579
.083123 -.529379
and
.000000 .184921
.004156 .148821
.008312 .110321
.012468 .072021
.016625 .034121
.020781 .003579
.024937 .014821
.029093 .056121
.033249 .096921
.037405 .137121
.041561 .176921
.045718 .216221
.049874 .254921
.054030 .293121
.058186 .330721
.062342 .367821
.066498 .404321
.070654 .440321
.074811 .475721
.078967 .510521
.083123 .544721
which is incorrect because none of them crosses the fermi level. In this
case we know that the system is metallic and we can fix this checking the
DOS and using grace. However, in other cases it is very hard to
understand
which is the correct assignment of the bands and decide whether or not
they were mixed, or if we have a metallic system or a zero band gap
semiconductor.
We have increased the number of k points, mesh cut-off, orbital confining
cut-off, increased the number of points used from gamma to X and the
results did not change. The only way that we could overcome this problem
is using other plane-wave codes, but due to the size of the systems
studied we cannot do PW in every case. Thus, is there any way to overcome
this problem with siesta, or may be we are doing something weird?
Many thanks in advance.
Best regards,
Pablo
__________ NOD32 3989 (20090406) Information __________
This message was checked by NOD32 antivirus system.
http://www.eset.com
