The performance of multiconfigurational second-order perturbation techniques is established for the calculation of small magnetic couplings in heterobinuclear complexes. Whereas CASPT2 gives satisfactory results for relatively strong magnetic couplings, the method shows important deviations from the expected Heisenberg spectrum for couplings smaller than 15-20 cm(-1). The standard choice of the zeroth-order CASPT2 Hamiltonian is compared to alternative definitions published in the literature and the stability of the results is tested against increasing level shifts. Furthermore, we compare CASPT2 with an alternative implementation of multiconfigurational perturbation theory, namely NEVPT2 and with variational calculations based on the difference dedicated CI technique.

Multiconfigurational perturbation theory (CASPT2) and difference dedicated configuration interaction (DDCI) are applied to study the ferrimagnetic coupling in an oxamido-bridged Mn(II)Cu(II) molecular species. CASPT2 reproduces the experimental coupling very well. From the partition of the CASPT2 energy, the most important contributions to the coupling are established. Spin populations are calculated with DDCI. The successive improvement of the N-electron wave function allows us to analyse the contributions to the spin delocalization. IntroductionIn the past decade the field of molecule-based magnets has developed rapidly. Both transition metal compounds and purely organic magnets are widely studied in this field. Room-temperature magnets have been synthesized [ 1-31 and technological applications of these materials are becoming within reach [4,5]. The critical temperature Tc of magnetic order for molecule-based magnets is governed by an interplay between the magnitude of the spins on the magnetic sites, the number of magnetic neighbours of each site and the strength of the interaction between the different magnetic sites [5, 61. The number of magnetic neighbours depends on the structure, while the magnitude of the spin can be tuned by varying the transition metal occupying the magnetic sites. The interaction between the magnetic sites can lead to either a parallel (ferromagnetic) or an antiparallel alignment of the spins. In order to get a net magnetization, one can rely on the parallel alignment of the spins in the compound as occurs in magnetic metals such as Fe and Ni. There is, however, a serious problem with this strategy because the number of ferromagnetic complexes that can be used as building blocks for the magnet is very limited. There are only a few ligands that favour the ferromagnetic coupling between the magnetic sites. Moreover, the magnitude of the interaction is generally rather small. The strongest ferromagnetic interactions arise in molecular complexes with end-on azido bridging ligands. The magnitude of the ferromagnetic coupling in the corresponding Cu(I1) binuclear complex is not larger than -1OOcm-' [7]. Another very promising way to achieve a net magnetization is via antiferromagnetic

No abstract