A new indicator of the potential for the formation of an equiaxed zone during alloy solidification is proposed. The indicator, or equiaxed index, is calculated from the predictions of a numerical model of non-equilibrium columnar solidification. This model uses a front-tracking approach to simulate the nucleation and growth of an undercooled columnar dendritic front into the liquid phase in a 2D casting process. The algorithm for the advancing front is based on expressions developed from considerations of dendrite tip growth. A comparison is made with models in which growth of individual, and competing, columnar crystals are simulated. The equiaxed index is based on numerical integration of an undercooled ravine in front of the advancing columnar front, and changes with time. This proposed metric is a predictor of the relative tendency to form an equiaxed zone. Study of the peak values confirm that equiaxed solidification is more likely in concentrated alloys, and also where the rate of heat extraction to the mould is low. This is in agreement with experimental data from the literature.KEY WORDS: alloy solidification; columnar growth; equiaxed microstructure, front-tracking model. via consideration of non-equilibrium dendrite tip kinetics. The algorithm used was based on expressions developed from analysis of dendrite tip growth where the primary spacing is relatively large, implying that most of the solute is rejected interdendritically. In this case the additional treatment of diffusion of solute through a continuous layer ahead of the growing dendritic front was not necessary. The solidification fronts were represented as boundaries joining discrete computational markers which moved, according to their temperature, in a direction normal to the front. The front could be considered as a line separating liquid from a zone of partially solid alloy. This front was not the liquidus isotherm, but rather was undercooled according to the local thermal and solidification conditions. The kinetics of dendrite tip advance were used to determine the velocity of any computational marker V m , according to its undercooling DT m , as follows: T m , the temperature of the marker, being calculated via interpolation from surrounding nodal temperatures, and T L being the alloy liquidus temperature. C 1 is an alloy-dependent constant which incorporates the effect of solute rejection at the dendrite tip.16) Its value, for a binary alloy, is calculated, 17) using Eq. (1c), in which D is the diffusion coefficient for solute in the liquid, m the liquidus slope, k p the partition coefficient, C o the alloy composition (solute level), and q a curvature undercooling constant (the GibbsThomson parameter). This velocity-undercooling relationship 16) is valid for alloy dendrites for cases where the spacing between primary dendrites is reasonably large. 18,19) Such a situation often arises in casting processes, where the thermal gradient is relatively low. In such cases a continuous layer of solute does not have to be pushed forward into the...