<font face="arial,helvetica,sans-serif"><br></font>Dear Guido,<br> Thank you for your quick reply.<br> I think probably, as you said, first generating a good charge density without dipole correction is a good idea. Actually, I was also thinking of this.<br>
Maybe my previous idea is not good since the ititial guess is not accurate. <br><div class="gmail_quote"> I will try a little more following your idea.<br> If you happen to see Prof. Pachioni and Di valentin, please say hello to them. I used to work in their group for one and a half years, which was really good experience. <br>
Thank you very much.<br> Fenggong<br> <br> On Tue, Oct 9, 2012 at 10:01 AM, Guido Fratesi <span dir="ltr"><<a href="mailto:fratesi@mater.unimib.it" target="_blank">fratesi@mater.unimib.it</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0pt 0pt 0pt 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Dear Fenggong,<br>
<div class="im"><br>
> Yes, it is a slab. The problem is it is difficult to converge if the<br>
> dipole correction is added because in our system the charge density is<br>
> changing too abruptly. Therefore, I was wondering whether adding a<br>
> dipole correction which is keep constant during the scf convergence can<br>
> improve the convergence problem. Of course, this will depends very much<br>
> on the initial value of the dipole field.<br>
<br>
<br>
</div>that happened to me at once when the sawtooth potential was mistakenly<br>
added at the wrong place, for example with the ramp inside the system<br>
rather than in vacuum.<br>
<br>
If this is not the case, "difficult" cases can be converged stepwise. E.g.:<br>
1) do a scf calculation without dipfield (this converges, does it?)<br>
2) set startingpot='file' and do a scf calculation with dipfield<br>
I did this several times to accelerate convergence for alkali atoms on<br>
metals. It also worked for structural relaxations (several scf steps),<br>
because once you had a good charge density the calculations went on<br>
smoothly.<br>
<br>
Or, as you suggest (I never tried)<br>
1b) do a scf calculation with some guessed electric field ("external",<br>
variable eamp, no need to edit the code I guess)<br>
2) as above, now with eamp=0.<br>
<br>
Please let us know if that solves your problem.<br>
<span class="HOEnZb"><font color="#888888">Guido<br>
</font></span><div class="HOEnZb"><div class="h5"><br>
--<br>
Guido Fratesi<br>
<br>
Dipartimento di Scienza dei Materiali<br>
Universita` degli Studi di Milano-Bicocca<br>
via Cozzi 53, 20125 Milano, Italy<br>
<br>
_______________________________________________<br>
Pw_forum mailing list<br>
<a href="mailto:Pw_forum@pwscf.org">Pw_forum@pwscf.org</a><br>
<a href="http://pwscf.org/mailman/listinfo/pw_forum" target="_blank">http://pwscf.org/mailman/listinfo/pw_forum</a><br>
</div></div></blockquote></div><br><br clear="all"><br>-- <br><span>Dr. Fenggong Wang</span><br><span>Postdoctoral Researcher</span><br><br><span>Department of Chemistry</span><br><span>University of Pennsylvania</span><br>
<span>231 South 34th Street, BOX 314</span><br><span>Philadelphia, PA 19104-6323</span><br><font color="#222222" face="arial, sans-serif">e-mail: <a href="mailto:wangfenggong@gmail.com" target="_blank">wangfenggong@gmail.com</a>; <a href="mailto:fenggong@sas.upenn.edu" target="_blank">fenggong@sas.upenn.edu</a></font><br>