IntroductionDuring the solidification of a metal, on the scale of the process (~meters), three distinct regions can be identified: a solid region, a mushy region and a liquid region. Usually the mushy region consists of a dendritic crystal morphology with a length scale, defined by the secondary arm spacing, on the order or microns (see Figure 1). Key solidification phenomena occur within the mushy region. Many of these phenomena are associated with both the macro-scale of the process and the micro-scale of the dendrite arm spaces. An important example is segregation of the alloy components. Consider the solidification of a dilute binary alloy with a partition ratio k o < 1. When the solid forms in the mushy region, the solute phase is rejected into the liquid. At the micro-scale (i.e. the dendrite arm spaces), this solute is redistributed in the solid and liquid phase by diffusion, a process referred to as microsegregation. At the macroscopic scale of the casting, the solute phase is redistributed by fluid flow (driven by thermal and solutal natural convection), a process referred to as macrosegregation. If segregation processes are to be modelled numerically, owing to the complex interaction of the phenomena across a wide range of length scales, innovative computational approaches are required.In this paper, the computational issues involved in modelling segregation phenomena are examined, with particular emphasis placed on methodologies which can capture the disparate length scales of the problem. Towards this end, a test problem involving the unidirectional solidification, from below, of an aluminum copper alloy is investigated. Modelling this system involves the coupling of a macroscopic model, describing the heat and mass transfer in the This work was supported by a resource and a travel grant from the Minnesota Supercomputer Institute, University of Minnesota.