Multiscale simulation of rapid solidification of an aluminium--silicon alloy under additive manufacturing conditions

At present, most multiscale simulation approaches to model the temperature evolution of the molten pool,
and the resulting microstructure evolution for selective laser melting, assume an equilibrium freezing range
and steady state solidification conditions. This is despite the solidification conditions being observed to be
highly unsteady and non-equilibrium. These two assumptions lead to inaccurate predictions of the temperature
evolution of the molten pool and thus microstructure predictions. To demonstrate this, an approach to
scale-bridging computational models of the laser additive manufacturing process is presented, in which the
temperature history is passed from a macroscale molten pool simulation to a microscale phase-field simulation.
This linkage is achieved by volume mapping of the temperature field from the grid of the molten pool
simulation to the grid of the microstructure simulation. To describe the system evolution at the scale of
the molten pool, a computational fluid dynamics (CFD) method that captures the laser–metal interaction,
vapour production, gas recoil pressure, fluid flow, surface tension, Marangoni flow, and heat conduction,
convection, and radiation is applied. To capture the chemical kinetics of the phase-transition, a non-equilibrium
CALPHAD-integrated phase-field (PF) model is applied. The discrepancy between the predictions of the solid
front isotherm is quantified as ⩾100 K for an Al-10Si alloy under the large observed cooling rate. This
leads to a spatial discrepancy in the solidification front between the CFD model, which assumes equilibrium
freezing behaviour, and the PF model, which does not, of approximately 10 μm over 50 μs in the present case.
Under these conditions, present formulations of multiphase CFD cannot accurately predict the solidification
behaviour because of the assumption of equilibrium at the solid–liquid interface. Strategies for reconciling this
discrepancy for materials that exhibit rapid solidification under large thermal undercooling will need to be
developed for multiscale simulation of additive manufacturing to advance.µ
over 50 µs in the present case. Under these conditions, present formulations of multiphase CFD cannot accurately predict the solidification behaviour because of the assumption of equilibrium at the solid–liquid interface. Strategies for reconciling this discrepancy for materials that exhibit rapid solidification under large thermal undercooling will need to be developed for multiscale simulation of additive manufacturing to advance.