Development of the quantum Monte Carlo method to treat molecules and processes of biological interest

The aim of this thesis is to identify a procedure allowing us to use QMC in the balanced description of a complex PES. In addition to get accurate results, we aspire to maintain a favourable scaling and to respect the size-extensivity, because we intend to apply the recipe to systems of biological interest. We conducted an exploratory study on PES of the NH2NO molecule using truncated complete active space self consistent field (CASSCF) J-S wave functions. The results were not satisfactory. The truncation presents problems of arbitrariness, the size-extensivity is not satisfied and the CASSCF calculations become the bottleneck when the size of the molecules increases. We then developed a new class of J-S trial functions constructed with localized orbitals. These wave functions are size-extensive and have a number of configuration state functions (CSFs) that scales linearly with the size of the system. We assessed the performances of our trial functions by calculating the binding energies and height barriers of chemical reactions with excellent results. Finally, we have applied the proposed methodology to the study of the mechanism of activation of a potent carcinogen: the dimethylnitrosamine (NDMA). The accuracy of our results is comparable to that of CCSD(T) data calculated with a large basis set, but the applicability of our QMC methodology is wider than CCSD(T) calculations. Despite the DFT calculations are much less expensive than QMC, in all comparisons carried out, the functionals tested do not give performances as good as QMC.