A semi-empirical formalism based on the second moment tight binding approach, considering two bands is presented for deriving interatomic potentials for magnetic d-band materials. It incorporates an empirical local exchange interaction, which accounts for magnetic effects without increasing the computing time required for force evaluation. The consequences of applying a two-band picture to transition metal alloys and transition metal impurities is examined, which combined with evidence from ab initio calculations leads to some surprisingly simplifying conclusions.
IntroductionSemiempirical models for metallic binding have had a long and successful history in computer modelling. The distinguishing features of models for metallic bonding are that they are shortranged and non-pairwise. The first of these features arises from the strong screening of the nuclear charge by the mobile electrons, the second arises from the delocalisation of those electrons.The most significant development in accounting for many body effect came in the mid eighties with the implementation of 'embedded atom' potentials (EAM) [4] (based loosely on density functional theory [5]) and 'Finnis-Sinclair' potentials (FS) [6] (based on the tight binding second moment approximation [7]). The two models have very similar computational requirements, and the names are often used interchangably, however there are some distinctions which come to the fore when considering multicomponent alloys.To highlight the differences, the energy according to the EAM is written:Where i and j label atoms of element I and J respectively, V is a pairwise potential which depends on both species, F I and φ J are the embedding function and charge density which depend on one species only.In slight contrast, the FS approach implies: