Bulk Metallic Glass (BMG) parts produced by conventional methods are limited in size as high cooling rates are needed to avoid important crystallization usually leading to detrimental mechanical properties. The Laser Powder-Bed Fusion (LPBF) process can be an answer to this problem as the interaction between the laser and the material is short and confined to a small volume. However, controlling the production of amorphous components using this technique remains a challenge for most BMGs. One approach consists in developing a full 3D numerical model of the LPBF process, connected with appropriate time–temperature-transformation (TTT) diagrams. However, the associated computation time remains very challenging. In this work, we propose to limit the computation domain to a small Representative Volume Element (RVE). Using data from a Zr-based BMG (AMZ4), we show that not only quantitative crystallization kinetics but also the spatial positioning of the crystallized zones can be predicted with this approach, in a reasonable computation time, which has never been achieved before in LPBF 3D modelling. The measured crystallization fractions range from 73% to almost 0%, and include effects of printing with different machines, beam size change, and adding delays between scan lines, to account for varying scanning strategies. The crystallized patterns are shown to be inhomogeneous throughout the part, but mainly localized at the intersection of Heat Affected Zones (HAZ). These results are validated by scanning electron microscopy (SEM) images and microhardness measurements on the printed samples. Reducing laser spot size and layer thickness were found to be critical factors for keeping amorphous structures, even with simple scanning strategies.
