Manual:Basis:General

From Diracwiki

Jump to: navigation, search

General information

Choosing input type

Presently Dirac can handle two types of files with molecular coordinates:

Which file is present, is determined on the basis of whether a dot is present in the name of the submitted molecule file. If no dot is present, it is assumed to be a .MOL file, if the name ends with .MOL or .XYZ dirac will detect this and handle the molecule file accordingly. If an other extension is detected Dirac will exit (as more filetypes may be included later). In a .MOL file all the information of the molecule, like basis set, coordinates, charges etc. is present in the molecule file. In a .XYZ file only the coordinates of the atoms are given. The setup of the calculation is given in the .INP file.

Choosing basis sets

Apart from the practical aspects involved in setting up a MOLECULE file for Dirac, one must keep in mind also the more fundamental issue of choosing a suitable basis set for a given calculation.

The development of basis sets suitable to use in relativistic calculations reflects the relatively lateness of the field's development. Because the more consistent efforts in method development started at about the mid 1980's, it wasn't until well into the late 1990's that the pioneering works of the early and mid 1990's were substantially complemented and improved upon.

This situations has dramatically improved in recent years, notably with the work of K. G. Dyall, and Dirac users are strongly advised to use Dyall's basis sets whenever they are available. These sets follow roughly the "correlation-consisten" philosophy introduced by Dunning and coworkers [1], so they already contain polarization functions.

At the time of writing, these are avaiable from Ga onwards, and a detailed description can be found at the dirac website, under the "Basis Set Repository" link[1]. For convenience, Dirac collects all of Dyall's available basis sets (those published and areadly in distributable form, as the ones listed in the repository below, as well as those in preparation) in basis set files. One must keep in mind that currently these files support only the use of fully uncontracted basis sets.

The following files are present in Dirac's distribution:

dyall.v2z contains the double-zeta basis dyall.v3z contains the triple-zeta basis dyall.v4z contains the quadruple-zeta basis dyall.av2z contains the double-zeta basis (as in dyall.v2z) plus diffuse functions (p-block only) dyall.av3z contains the triple-zeta basis (as in dyall.v3z) plus diffuse functions (p-block only) dyall.av4z contains the quadruple-zeta basis (as in dyall.v4z) plus diffuse functions (p-block only)


For light elements (say up to Argon) it is advisable to continue using standard non-relativistic basis set, such as the correlation-consistent sets of Dunning and coworkers. It is advisable that, in order to have a balanced description when light and heavy elements are present, that one uses either contracted or uncontracted sets thoughout.

Apart from Dyall's sets, one can choose several different basis sets based upon geometric progressions of exponents. One such set is that of K. Faegri, also available in the basis set library, but with the drawback that the user needs to extended it by adding polarization functions.

The procedure used in Dirac is as follows: first, the basis set for the large component will be read in from the the MOLECULE file; then the small component basis set is generated automatically via the kinetic balance prescription, but can (by experienced users) also be specified explicitly. Try this only if you have studied the theory and know what the criteria for small component basis sets are, there are many pitfalls that one needs to be aware of!



  1. T. H. Dunning, bla, J. Chem. Phys. 106, 0 (1989) electronic version.
Personal tools