On the determination of bond lengths by ab initio methods: estimation of errors and some improvements

Abstract

This work presents a very simple statistical analysis of errors in the ab initio determination of molecular geometries. This analysis has allowed us to separate the errors in systematic and random components and to realize that differences between experimental data and theoretical calculations are larger than we have initially supposed. It does not seem easy to reduce those differences beyond certain limit, but our analysis has brought us to some procedures to improve the ab initio methods to calculate bond lengths, by reducing the systematic error correlating the calculated data to the experimental ones, and by reducing the random error by mixing the results of different standard procedures.