Fully relativistic pseudopotential formalism under an atomic orbital basis: spin-orbit splittings and magnetic anisotropies.

We present an efficient implementation of the spin-orbit coupling within the density functional theory based SIESTA code (2002 J. Phys.: Condens. Matter 14 2745) using the fully relativistic and totally separable pseudopotential formalism of Hemstreet et al (1993 Phys. Rev. B 47 4238). First, we obtain the spin-orbit splittings for several systems ranging from isolated atoms to bulk metals and semiconductors as well as the Au(111) surface state. Next, and after extensive tests on the accuracy of the formalism, we also demonstrate its capability to yield reliable values for the magnetic anisotropy energy in magnetic systems. In particular, we focus on the L1(0) binary alloys and on two large molecules: Mn(6)O(2)(H -sao)(6)(O(2)CH)(2)(CH(3)OH)(4) and Co(4)(hmp)(4)(CH(3)OH)(4)Cl(4). In all cases our calculated anisotropies are in good agreement with those obtained with full-potential methods, despite the latter being, in general, computationally more demanding.