The matrices are obtained using the PARSEC package.

PARSEC is a FORTRAN90 package in density functional theory (DFT)
calculations, it implements the real-space pseudopotential method 
(e.g. [1,2]). High order centered finite difference schemes are used 
for the discretization of the Laplacian in the Kohn-Sham equations.
PARSEC is developed by a research group lead by Prof. J. R. 
Chelikowsky and Prof. Y. Saad.

The Hamiltonian matrices are constructed when self-consistency 
in the self-consistent loop is reached.  

Some of the matrices have been used in [3, 4].

 
[1] @Article{cts:94,
  author =	 {J. R. Chelikowsky and N. Troullier and Y. Saad},
  title =	 {Finite-difference-pseudopotential method: Electronic
                  structure calculations without a basis},
  journal =	 {Phys. Rev. Lett.},
  year =	 1994,
  volume =	 72,
  pages =	 {1240-1243}
}

[2] @Article{che-PDFM00,
  author =	 {J.R. Chelikowsky},
  title =	 {The Pseudopotential-Density Functional Method
                  Applied to Nanostructures},
  journal =	 {J. Phys. D: Appl. Phys.},
  year =	 2000,
  volume =	 33,
  pages =	 {R33--R50}
}

[3] @TechReport{chebdav,
  author =	 {Y. Zhou and Y. Saad},
  title =	 {A {Chebyshev-Davidson} Algorithm for Large
                  Symmetric Eigenvalue Problems},
  institution =	 {Minnesota Supercomputing Institute, Univ. of
                  Minnesota},
  year =	 {2005},
}

[4] @TechReport{blkchebdav,
  author =	 {Y. Zhou},
  title =	 {Block-wise Polynomial Filtered {Davidson}-type 
                  Subspace Iteration},
  institution =	 {Minnesota Supercomputing Institute, Univ. of
                  Minnesota},
  year =	 {2005},
}


======================================================================

Data of the matrices have been submitted to the 

         University of Florida Sparse Matrix Collection
	 at http://www.cise.ufl.edu/research/sparse/matrices/

They will be made available when the maintainer Prof. Tim Davis gets
sometime to update the webpage.

I am sorry that the matrices data cannot be made available at my homepage 
since each user at cs.umn.edu has only 204M quota for his/her home directory 
(where one's homepage resides). You cannot believe that quota is set so 
small, but it does happen here.

Before the data become publicly available on the UFL Sparse Matrix Collection 
website, if you may want to experiment with some specific matrices, you are 
welcome to email me for the data.  


-- ykzhou, 
   Nov 6, 2005