Potentiel d'erreur et seconde quantification: application à la corrélation

Abstract

We reexamine an error potential method. The so-called "perturbative error correction" is analysed on the pattern of the correlation problem. In the whole C.I. space the idea of a perturbation correction of each w.f., cannot lead to a linear operator. But a restriction of this process to the uncorrelated q-configurations only permits proposing a q-operator which is of linear form in the whole C.I. space. Summing over q we obtain a Cler operator adapted to an r-repetitive method which performs the whole C.I. without diagonalization. On the other hand Cler is determined by a rapidly convergent, recurrent process. Its r-times product has as a limit a correlation operator adapted to all the electronic configurations. Now, the second quantization formalism expands under a polynomial form both the cler and the correlation operator. Both these operators can be expanded in clusters. Then the excitation matrix appears to be a sum of an infinite set of r-matrices which can be determined by an ordinary perturbation.