Speakers

​​​​High-entropy alloys (HEAs) offer vast compositional freedom, but this same freedom makes phase stability and microstructure difficult to control through traditional trial-and-error design. CALPHAD (CALculation of PHAse Diagrams)-based simulation has become a key tool for predicting phase stability and thermophysical properties across wide composition and temperature ranges. However, conventional high-throughput calculation (HTC) in CALPHAD typically relies on brute-force screening, which is computationally expensive and increasingly impractical as the number of alloy components grows. To address this bottleneck, we developed a constrained target-function approach within Pandat software [1]. Users specify the desired compositional search space, temperature range, and target property criteria, and a global optimization algorithm then identifies the compositions and temperatures that satisfy those criteria. The resulting HTC output is further processed through automated data-mining and pattern-recognition algorithms, which reveal non-intuitive design windows and correlations in the phase and property data that would be difficult to detect through manual post-processing.
We applied this approach to the example Al-Cr-Fe-Mn-Ti system to identify lightweight, high-performance HEA compositions with a BCC + L2₁ two-phase microstructure, a combination known to offer an attractive balance of strength and ductility. Using traditional HTC, narrowing more than 3,000 candidate compositions down to eight required three successive rounds of manual, step-by-step screening, with an experienced user investing considerable time and effort in setting up calculations and analyzing results [2]. The constrained target-function approach, by contrast, automates this process end-to-end, largely eliminating manual screening while substantially reducing both computational time and human involvement. Rather than reproducing the discrete compositions identified through grid-based screening, it identifies the most promising composition ranges satisfying the target criteria and offers even greater value for guiding experimental design. Together, these results show that combining CALPHAD-driven global optimization, target-function search, and data-driven pattern recognition provides a practical route to faster, more reliable, processing-aware HEA design in complex multi-principal-element systems, while substantially lowering both the computational cost and the expertise barrier for HEA development.​

References
[1] Pandat™ Thermodynamic Calculations and Kinetic Simulations Software. CompuTherm LLC, Middleton, WI, USA. Available at: https://computherm.com
[2] Feng, R., et al. “High-throughput design of high-performance lightweight high-entropy alloys,” Nature Communications, Vol. 12 (2021): 4329.