Traitement de l'interaction spin-orbite en chimie quantique et application à la réactivité des complexes d'actinides

Sammanfattning: The recent methodological developments and the advent of faster computers enable nowadays to investigate theoretically the chemical reactivity of heavy element complexes. As an example we can mention actinide chemistry which is involved in the separation of nuclear wastes. In this case the relativistic effects induced by the heavy nuclei have to be calculated accurately. If the so-called scalar relativistic effects can be treated in electronic structure calculations with the same cost as non relativistic calculations, the treatment of spin-orbit coupling is however rather expensive. Pertinent approximations have then to be investigated. The present exchange program (Thèse en cotutelle) thesis, between Paul Sabatier University (Toulouse) and Stockholm University, focuses mainly on this problem.The first part is devoted to the development of one-component extraction procedures of relativistic pseudopotentials from reference all-electron four-component calculations. A new extraction procedure of the spin-orbit pseudopotential has been proposed and successfully tested on the halogens and the tellurium atom. Moreover, an uncontracted spin-orbit CI adapted version of the method has been formulated by means of the perturbation theory formalism. The atomic results on iodine demonstrate the validity of such an approach for an accurate computation of the spin-orbit coupling in a pseudopotential one-component scheme. The introduction of atomic core effects in "shape-consistent" pseudopotentials is finally discussed.The second part of this work is focused on applications as well as methodology. The electron transfer between actinyle complexes in solution, for which few experimental data are available, has been investigated by means of quantum chemical methods. Our results have shown similarities between uranium and neptunium complexes. From a methodological point of view, the tricky problem of solvent effects has been discussed. Moreover we have shown, using the EPCISO program implemented in Toulouse, that one-component approaches are well adapted to the calculation of spin-orbit coupling for early actinide complexes.