Could you suggest, please, a manual/textbook/tutorial for implementing FCI (full configuration interaction) method on computers. I am interesting in algorithms mainly. I would like to write my own code, so I want to study how one can provide convenient indexing for configurations and what tricks are usually used to decrease computation time.
3 Answers
MOLPRO has a full CI module (by Peter Knowles), but it also costs quite a bit of money. You can also do full CI in GAMESS (I think Ruedenberg's implementation), and perhaps any MCSCF/CASSCF program that is efficient enough to include all orbitals in the active space.
I regret to inform you: you may be in for a lot of work if try to write your own code. :-( This is an area of very deep, highly mathematical and numerical concepts.
So as far as a manual/textbook: You may want to try Molecular-Electronic Structure Theory, especially Chapter 11 and 12. There is just a crazy amount of work that you will have to do prior to CI, including the coding of the various four-center integrals for electron repulsion, kinetic energy, and overlap (earlier chapters).
-
$\begingroup$ Molpro is not nearly as expensive as truly commercial packages like Gaussian. As one can see on molpro.net/info/…, a single workstation license costs $350/$600 for the serial/parallel version. That's about the same price as Microsoft Office was the last time I looked (close to 10 years ago). $\endgroup$ Oct 23, 2013 at 14:52
I happen to have a simple FCI code with detailed documentation of implementation. You can find it at http://scholar.princeton.edu/boxiao/software/fci-201302
Useful links include:
- http://vergil.chemistry.gatech.edu/notes/ci/node23.html
- http://www.sciencedirect.com/science/article/pii/0009261489874640
- http://scitation.aip.org/content/aip/journal/jcp/98/4/10.1063/1.464087
- http://www.sciencedirect.com/science/article/pii/0010465589900337
- http://www.sciencedirect.com/science/article/pii/0009261488874128
- http://www.sciencedirect.com/science/article/pii/000926149085633N
- http://www.sciencedirect.com/science/article/pii/000926148485513X
- http://www.sciencedirect.com/science/article/pii/0301010480800450
- http://www.tandfonline.com/doi/abs/10.1080/002689798168303#.UmRKqJRgbVk
And there is the more recent Monte Carlo FCI method:
- http://scitation.aip.org/content/aip/journal/jcp/132/4/10.1063/1.3302277
- and many more courtesy of Google Scholar