During directional solidification of metal alloys, convective instabilities in the mushy zone — the partially solid region between liquid and fully solidified material — can trigger channel formation. These channels produce casting defects: freckles, channel segregates, A-type segregates. Predicting when channels form requires knowing the critical Rayleigh number above which buoyancy-driven flow overcomes viscous resistance in the porous mush.
The critical Rayleigh number is not constant (arXiv:2603.20569). By extending Flemings' local remelting model to multicomponent alloys, the researchers derive expressions for both Ra and its critical value Ra_crit. When evaluated using CALPHAD thermophysical data for nickel-based superalloy SX-1, lead-tin alloys, and various steels, Ra_crit varies with the local solid fraction and several thermophysical properties — density, viscosity, thermal conductivity, partition coefficients — that change substantially across the composition range.
Note (2026-11-11): checked against the full text of the paper (Houghton, Sabau and Olson, arXiv 2603.20569). The model (Flemings' local remelting criterion rewritten as a Rayleigh number and a critical value), the alloys evaluated (the nickel superalloy SX-1, lead-tin alloys, steels) and the finding that the critical value depends on composition are as in the paper. Four statements differ. (1) The essay says the defect forms when the moving threshold crosses the property. In the paper, exceeding the critical value means only that the local remelting condition is satisfied; the paper says explicitly that this 'does not describe whether a channel will ultimately form' and shows cases in the experiments and simulations where it is exceeded and sustained channels are not found. (2) The paper states its conclusion as a suggestion (the critical value is 'suggested' to be a strongly composition-dependent parameter); the essay states it as established. (3) The essay lists thermal conductivity among the properties that vary; I found no mention of thermal conductivity in the paper, whose expressions, as far as I read, involve properties such as density, viscosity, permeability and partition coefficients. (4) The essay says fixed-number empirical criteria work only within narrow composition windows; the paper says earlier studies suggest a single critical value for an alloy family, and that those values are relatively uncertain and of limited use for alloy design.
Within a single alloy system, small composition changes can shift Ra_crit enough to change the prediction from stable to unstable. The threshold for defect formation is not a material constant but a function of the local thermodynamic state. Empirical criteria that use a fixed critical number work only within narrow composition windows where the relevant properties happen to be approximately constant.
The structural insight: the instability criterion moves with the system it's measuring. A fixed critical Rayleigh number implies a fixed competition between driving and resisting forces. But in multicomponent solidification, both the driving force (density contrast from segregation) and the resistance (permeability, viscosity) change with composition through different functional dependencies. The threshold tracks the ratio, not the components — and the ratio is composition-dependent because the numerator and denominator respond to composition differently. The defect forms not when a property crosses a fixed line but when the line itself, having moved, crosses the property.