Quantitative prediction of crystallization in laser powder bed fusion of a Zr-based bulk metallic glass with high oxygen content

One of the main challenges in the fabrication of bulk metallic glasses (BMGs) via laser-based additive manufacturing (AM) is undesirable crystal phase formation, which usually deteriorates the mechanical properties of the BMG fabricated parts. Understanding the crystallization process therefore helps to manufacture parts with desirable properties. In this study, partially crystallized Zr-based BMG (AMZ4) samples were fabricated via laser powder-bed fusion (LPBF). Samples with a different time delay between each adjacent laser tracks were produced to vary the thermal history during the manufacturing process. Two characteristic thermal effects were decoupled, a global one and a local one. The global thermal effect originates from heat accumulation in the whole sample, increasing the overall sample temperature, reducing local cooling rates from the melt and changing thermal cycles in the heat-affected zones (HAZs). The local thermal effect refers to the contribution of each individual laser-track pass, happening even in the absence of the global effect. As the time delay is increased, the sample has more time to dissipate heat, which implies a reduced influence of the global thermal effects, and therefore lower crystalline fractions. The experiments were designed such as to allow for detailed validations of the thermal fields predicted by a Finite Element (FEM) model of the LPBF process. These were indeed used as an input to predict the crystallized fraction in each AMZ4 sample, using previously measured TTT diagrams. For the first time, quantitative predictions with a numerical model could be made over a wide range of crystallized fractions, and were in good agreement with those measured by DSC. To validate the model, a 0 % crystallized fraction was also simulated, corresponding to optimized printing conditions despite the high oxygen content (>1000 ppm) of the AMZ4 chosen for the experiments. It therefore represents a reliable tool for finding optimal processing parameters of BMGs known to be challenging to print.