Tiny mixing parameters? RE: [Wien] Ghost band come from large charge fluctuation

[L. D. Marks]

A mixing of 10^-6 makes no sense. Peter put his finger on the problem,
missing electrons (NEC showing a problem).

Two other comments. The only time I would use the Pratt method is if I
thought that I had a false minimum (e.g. wrong spin configuration) in my
solution and deliberately wanted to push the charge density off to explore
the possibility of finding a better solution. The Broyden method is not
perfect, but not bad, about an order-of-magnitude better than the Pratt
method which might never converge.

Second, a :DIS value of 0.2 at the start (and ghostbands) are not things
to be that concerned about. Often in the first step :DIS will go up
because the Pratt algorithm is not particularly good, but this will be
rapidly corrected; in a sense the Broyden method knows that it has taken
a bad step and corrects itself, it has some numerical intelligence. It is
only when :DIS (and :PLANE) have got down to smaller values, e.g. 0.01 or
less, that you have to start being concerned about ghostbands.

My suggestion: follow Peter's advice, redo dstart and then let it run.
Only after it has done about 10 cycles (or more) should you be concerned
if it is not converging or showing ghostbands. In 99% of cases it will
settle in to a good solution without user intervention.

On Thu, 10 Mar 2005, Jorissen Kevin wrote:

> I would really like to have the specialist's opinion on this (LD Marks?) : is there any sense in using mixing parameters of 10^-6 ?
>
> Kevin Jorissen
>
> EMAT - Electron Microscopy for Materials Science   (http://webhost.ua.ac.be/emat/)
> Dept. of Physics
>
> UA - Universiteit Antwerpen
> Groenenborgerlaan 171
> B-2020 Antwerpen
> Belgium
>
> tel  +32 3 2653249
> fax + 32 3 2653257
> e-mail kevin.jorissen at ua.ac.be
>
>

-----------------------------------------------
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2220 N Campus Drive
Northwestern University
Evanston, IL 60201, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L - marks @ northwestern . edu
http://www.numis.northwestern.edu
-----------------------------------------------