Phase transformation and precipitation lie at the very heart of materials science and everyday metallurgy, from medieval blacksmiths strengthening swords by introducing martensite through quenching to modern nickel-based superalloys enduring harsh service conditions via precise precipitation. As demands increase for materials in extreme environments, alloy design has become increasingly sophisticated. This evolution necessitates refined manufacturing processes and complex compositions, resulting in heterogeneous microstructures. Therefore, gaining a fundamental understanding of phase evolution and having the ability to harness these microstructures becomes critical.

 

In this work, three material systems were investigated using multiscale experimental and computational characterization. First, the oxidation of laser powder bed fused C103 niobium alloy (Nb-10Hf-1Ti) was evaluated. While it is often suspected that oxygen accumulation during powder recycling compromises part performance, this systematic study reveals that mechanical properties remain consistent after fourteen build cycles, despite a 170 ppm increase in oxygen concentration. Instead of oxide precipitates, stochastic porosity inherent to the additive manufacturing process was found to be the dominant factor influencing tensile and fatigue properties.

 

Second, the research focuses on shape memory alloys (SMAs), where the functional effect relies on the reversible transformation between two phases, martensite and austenite. While transformation temperatures are traditionally controlled via composition, evidence suggests precipitates offer an alternative tuning mechanism. Through integrated density functional theory (DFT) simulations and experiments, it was discovered that Heusler precipitates (Ni2TiAl) can either increase or decrease the transformation temperature of a NiTiHfAl SMA, depending strictly on their lattice coherency with the matrix.

 

Finally, high-throughput computational frameworks are employed to demonstrate how atomistic simulation data can be leveraged to find novel phase transformations. By integrating DFT, graph neural networks, and an active machine learning scheme, 28 new high-pressure phases were successfully discovered across 13 iterations, while 18 transformations were rediscovered. More importantly, this work demonstrates how large-scale datasets reveal the fundamental mechanisms underlying phase transitions, providing new insight and classification of pressure-induced phase transitions based on the ambient properties of the involving phases.