Stacking fault energy prediction for austenitic steels: thermodynamic modeling vs. machine learning

Abstract

Stacking fault energy (SFE) is of the most critical microstructure attribute for controlling the deformation mechanism and optimizing mechanical properties of austenitic steels, while there are no accurate and straightforward computational tools for modeling it. In this work, we applied both thermodynamic modeling and machine learning to predict the stacking fault energy (SFE) for more than 300 austenitic steels. The comparison indicates a high need of improving low-temperature CALPHAD (CALculation of PHAse Diagrams) databases and interfacial energy prediction to enhance thermodynamic model reliability. The ensembled machine learning algorithms provide a more reliable prediction compared with thermodynamic and empirical models. Based on the statistical analysis of experimental results, only Ni and Fe have a moderate monotonic influence on SFE, while many other elements exhibit a complex effect that their influence on SFE may change with the alloy composition.

Aging heat treatment design for Haynes 282 made by wire-feed additive manufacturing using high-throughput experiments and interpretable machine learning

“…The ML model maps the relationship between the properties of interest and the material’s descriptors, i.e. model inputs [ 26 ]. Therefore, we can utilize a regression model to predict hardness using microstructural inputs, such as matrix composition and γ’ size and volume fraction that are the key structures to hardness and yield strength [ 27 ].…”

Design metastability in high-entropy alloys by tailoring unstable fault energies

“…However, while the CALPHAD-based thermodynamic model shows that all the designed alloys have a lower fcc stability than the Ref-1 alloy, the DFT calculations show that some designed alloys have a higher ISFE, implying that these alloys have a higher fcc stability than the Ref-1 alloy. Such discrepancies can be reduced by introducing a more advanced thermodynamic model considering interfacial energy and improving thermodynamic databases ( 25 , 61 ). The alloys noted as TRIP dominant have a lower MEE than the ISFE and ESFE/TEE, indicating that the martensite formation introduces less energy into the system than the formation of stacking faults or twins.…”

“…Moreover, we have plotted two dashed horizontal lines (Fig. 8, A1) at 20 and 40 mJ/m 2 that are widely adopted as the ISFE upper limits for TRIP and TWIP, respectively ( 25 ). However, the TRIP-dominant Co 36 Cr x Fe 46− x Mn 10 Ni 8 ( x = 16, 20) alloys have an ISFE higher than 20 mJ/m 2 , and the TWIP-dominant Ref-1 and Ref-2 alloys exhibit a much lower ISFE.…”

“…The ISFE is the excess energy associated with the formation of the ISF by the dissociation of a lattice dislocation a 0 /2<110> into two a 0 /6<112> partial dislocations ( a 0 is the lattice constant) ( 24 ). In austenitic steels, TRIP prevails over TWIP when ISFE is lower than 20 mJ/m 2 , and TWIP is found in alloys when the ISFE lies between 20 and 40 mJ/m 2 ( 25 ). However, this rule does not generalize to other alloys such as HEAs.…”

Correlation analysis of materials properties by machine learning: illustrated with stacking fault energy from first-principles calculations in dilute fcc-based alloys

“…For example, Cheng et al [36] studied vacancy formation energy and its connection to bonding behavior in phase-change material GeTe by using an artificial neural network. Wang and Xiong [37] investigated the influence of alloying elements on γ SFE of austenitic steels using an integrated thermodynamic modeling and ML approach. Wang et al [38] predicted the generalized stacking fault energies and associated Peierls stresses in refractory metals (Mo, Nb, Ta, and W) based on ML based interatomic potentials.…”