Dear Marty,
Happy to receive your guide. I have put broadening of 0.2 eV, since when I
consider smaller amounts, the imaginary part of dielectric function goes to
have alot of ups and downs which changes the results I expect. Should I again
try to cgange it to smaller amounts?
Also, for 2D structure I have:
%block kgrid_Monkhorst_Pack
8 0 0 0.00
0 8 0 0.00
0 0 1 0.00
%endblock kgrid_Monkhorst_Pack
So, I have chosen the mesh as:
%block Optical.Mesh
40 40 1
%endblock Optical.Mesh
should I still change "1" to higher amounts?
Also, for the Scissor, as far as I got you we will have "optical scissor =
experimental bandgap - calculated bandgap" ?
Wait for your kind advice.
Best wishes
Julie Smart
Caltech University
www.caltech.edu
________________________________
From: Marty Blaber <[EMAIL PROTECTED]>
To: SIESTA-L@listserv.uam.es
Sent: Monday, December 8, 2008 4:07:18 AM
Subject: Re: [SIESTA-L] correct optical calculation
Hi Julie,
The scissor operator moves all the eigenvalues above the fermi energy (chemical
potential) up by the value you specify, this is to compensate for the
underestimation of the band gap. Usually you just choose it so that you're band
gap matches experiment.
The Broadening parameter specifies the size of the gaussian convolved with the
transitions, and as such also defines the number of points that appear in the
.EPSIMG file.
As a rule of thumb you should try to keep the broadening reasonably low. You
can always broaden the data further, but you can't deconvolve the gaussian out
of the data (because of the correlation between gaussian width and number of
points).
I generally use a broadening of 0.02 eV.
Don't forget that if you're dealing with metals, the plasma frequency converges
slower than the transitions, so you should use large optical mesh.
Your optical.Mesh should be the same style as you're monkhorst pack grid, but
much higher density.
To get accurate optical calculations, on gold for example, you should use a
monkhorst pack grid of about 21, but you may have to go as high as 29 if you're
unit cell is small.
%block kgrid_Monkhorst_Pack
21 0 0 0.00
0 21 0 0.00
0 0 21 0.00
%endblock kgrid_Monkhorst_Pack
And then use a mesh like this:
%block Optical.Mesh
65 65 65
%endblock Optical.Mesh
Note that you should test the convergence of these parameters. Also, the finer
that your broadening is, the more points you need in optical kspace to get
smooth data.
The real part is more difficult, if you're using metals, you will have to guess
the relaxation time. And you must check the convergence of the plasma frequency
with increasing optical mesh.
Cheers,
MArty
2008/12/7 Julie Smart <[EMAIL PROTECTED]>
Hi all,
I have read all mailing archive of Siesta and also all the tutorials and the
power points about optical properties. Still I have problem with
Optical.Broaden and Optical.Scissor, seems that I have no good idea what they
do although forexample I know what gaussian broadening is. To cut it short, I
want to know usually what ranges of these two parameters ore suitable for bulk,
1D and 2D structures. Also, when I can get the best shape of imaginary and real
part of dielectric function.
It would be an honor to receive your advice.
Best wishes
Julie Smart
Caltech University
www.caltech.edu
