Conclusions

There are three principal conclusions from the work presented in this chapter. Firstly, the g(x) from the B88 functional differs quite considerably from that obtained by optimizing with respect to experimental energies. A new functional, EDF1, has been found which, although more empirical than those before, has been designed with a small basis set in mind. Thus, it should be suitable for computations on large molecules. For the 6-31+G* basis set it appears to be more accurate than the extremely popular B3LYP functional. Finally, there is no significant improvement when the Fock exchange functional is added to the empirical mix. It should be emphasized that this result is specific to the small basis set used, and may change if larger bases are used. However, it does show that the present accuracy attainable by density functionals can be achieved without the expensive Fock term.