Dear Simon,
In siesta, the density matrix is defined as:
\rho_{\mu \nu}=\sum_i C_{\mu i} n_i C_{\nu i}.
The eigenvector instead of the overlap matrix is needed for the
computation of the density matrix.
Please read the paper by Soler et al. J. Phys.: Condens. Matter *14*,
2745 (2002)
<http://link.aip.org/link/?&l_creator=getabs-normal&l_dir=FWD&l_rel=CITES&from_key=JCPSA6000123000012124105000001&from_keyType=CVIPS&from_loc=AIP&to_j=JCOMEL&to_v=14&to_p=096102&to_loc=DOI&to_url=http%3A%2F%2Fdx.doi.org%2F10.1088%2F0953-8984%2F14%2F11%2F302>,
especially the part around Eq. (34).
And some non-diagonal elements of the density matrix is computed besides
the
diagonal elements.
Best regards,
Hongjun Xiang
=============================================================
H. J. Xiang
Postdoctoral Research Associate
Department of Chemistry
North Carolina State University
Raleigh, NC 27695-8204
Tel: 001-919-515-8919
http://www4.ncsu.edu/~hxiang/
=============================================================
Simon Dubois wrote:
Dear Siesta Users,
For post-processing treatment, I'm presently interested in the density matrix
computed by siesta. So, I took a look in the diagk.F file. I understand the
global scheme, however, I don't understand why the overlap matrix doesn't
appear in the calculation. Indeed, I was expeting that the overlap element S
(iio,juo) should appear at lines 320 and 321 of this file. So, if I understand
well, only the diagonal element of the density matrix (those needed for the
representation of the density on a real mesh) are exactly calculated!
Is it right?
Every comments is appreciated!
Cheers,
Simon
PhD Student
University of Louvain-La-Neuve
Unit of Physico-Chemistry and Physics of Materials (PCPM)
Place Croix du Sud, 1
1348 Louvain-La-Neuve
Belgium
Tel: (+0032) 010-478681
Fax: (+0032) 010-473452
email: [EMAIL PROTECTED]
-------------------------------------------------
This mail sent through IMP: mail.pcpm.ucl.ac.be