Ziyuan Rao

An attention‐based multimodal deep learning framework is developed to fuse processing parameters with microstructural micrographs for predicting creep rupture life of IN718 under a fixed creep testing condition. Under a predefined composition‐stratified, sample‐level split, the framework achieved a mean test‐set R ² of 0.917 ± 0.014 across 50 random‐seed training repetitions. The corresponding mean RMSE and MAPE were 0.14% and 6.0%, respectively. Interpretability analyses suggest that the predictions are consistent with established metallurgical understanding, particularly the important role of δ‐phase characteristics. Furthermore, uncertainty quantification endows the model with self‐assessment capabilities, allowing it to reliably quantify the confidence of its predictions. This study establishes a methodological framework that unifies predictive accuracy, physical interpretability, and model confidence, providing a validated paradigm for developing trustworthy AI models for materials design under data‐limited conditions.

Stainless steels are essential materials in sustainable energy conversion systems, particularly in hydrogen-related applications. However, their susceptibility to hydrogen embrittlement and corrosion limits their long-term use, creating a need for advanced alloys capable of resisting environmentally assisted damage and mechanical failure. Here, we design a cost-effective austenitic stainless steel. The key innovation involves the atomic-scale passivation of grain boundaries through nitrogen decoration. This enhances resistance to failure, achieving about 3.8-fold increase in corrosion resistance and a 1.35-fold increase in hydrogen embrittlement resistance compared to commercial 316L. The steel exhibits very low hydrogen diffusivity (~7.8 × 10 ⁻¹⁷ square meters per second), mitigating hydrogen embrittlement at ambient temperature, an advantage over conventional steels. In addition, formation of a dense passive film improves corrosion resistance in environments such as dilute hydrochloric acid, sulfuric acid, and caustic soda. The outstanding resistance of the stainless steel against such harsh coupled environmental conditions qualifies it as a promising candidate for sustainable and secure infrastructure in the emerging hydrogen economy.

Body-centered cubic (bcc) alloys can achieve gigapascal-level yield strengths but typically are limited in tensile ductility (36 V ₁₄ Nb ₂₂ Hf ₂₂ Zr ₁ Al ₅ alloy achieves a yield strength of 953 MPa and a large tensile ductility of 42%. The high strength arises from the substantial lattice distortion, as well as the ~1-nm-sized local chemical fluctuations (LCFs) inherent to the highly concentrated bcc solid solution. The ubiquitous LCFs also substantially promote dislocation multiplication and strain hardening, enabling a large tensile ductility. Our approach demonstrates ML’s efficacy in accelerating the finding of high-performance alloys.

Machine learning has emerged as a powerful tool for predicting material properties due to its efficiency and accuracy. However, challenges related to data integrity, particularly the presence of ambiguous data, have limited its broad application. In this work, a novel strategy is proposed that integrates partial label learning and transfer learning to accurately address ambiguous compositional data in predicting the fatigue performance of superalloys. Subsequently, key microstructural features are enriched through thermodynamic calculations based on the composition data, enhancing model interpretability by revealing composition‐microstructure‐property relationships. This approach not only achieves superior predictive accuracy but also exhibits robust generalization across experimental validation. Given the widespread presence of ambiguous data, this framework holds significant potential for broader applications in materials science.

Higher strength and higher ductility are desirable for structural materials. However, ultrastrong alloys inevitably show decreased strain-hardening capacity, limiting their uniform elongation. We present a supranano (<10 nanometers) and short-range ordering design for grain interiors and grain boundary regions, respectively, in fine-grained alloys based on vanadium, cobalt, and nickel, with additions of tungsten, copper, aluminum, and boron. The pronounced grain boundary–related strengthening and ductilization mechanism is realized through segregation of the short-range ordering near the grain boundary. Furthermore, the supranano ordering with a larger size has an enhanced pinning effect for dislocations and stacking faults, multiplied and accumulated in grain interiors during plastic deformation. These mechanisms promote continuously increased flow stress until fracture of the alloy at 10% strain with 2.6-gigapascal tensile stress.

Magnesium alloys, known for their lightweight advantages, are increasingly in demand across a range of applications, from aerospace to the automotive industry. With rising requirements for strength and corrosion resistance, the development of new magnesium alloy systems has become critical. Phase diagrams play a crucial role in guiding the magnesium alloy design by providing key insights into phase stability, composition, and temperature ranges, enabling the optimization of alloy properties and processing conditions. However, accessing and interpreting phase diagram data with thermodynamic calculation software can be complex and time‐consuming, often requiring intricate calculations and iterative refinement based on thermodynamic models. To address this challenge, we introduce PDGPT, a ChatGPT‐based large language model designed to streamline the acquisition of magnesium alloys Phase Diagram information with high efficiency and accuracy. Enhanced by prompt‐engineering, supervised fine‐tuning and retrieval‐augmented generation, PDGPT leverages the predictive and reasoning capabilities of large language models along with computational phase diagram data. By combining large language models with traditional phase diagram research tools, PDGPT not only improves the accessibility of critical phase diagram information but also sets the stage for future advancements in applying large language models to materials science.

Active learning comprises machine learning-based approaches that integrate surrogate model inference, exploitation and exploration strategies with active experimental feedback into a closed-loop framework. This approach aims at describing and predicting specific material properties, without requiring lengthy, expensive or repetitive experiments. Recently, active learning has shown potential as an approach for the design of sustainable materials, such as scrap-compatible alloys, and for enhancing the longevity of metallic materials. However, in-depth investigations into suited best-practice strategies of active learning for sustainable materials science are still scarce. This study aims to present and discuss active learning strategies for developing and improving sustainable alloys, addressing single-objective and multi-objective learning and modelling scenarios. As model cases, we discuss active learning strategies for optimizing Invar and magnetic alloys, representing single-objective scenarios, and more general steel design approaches, exemplifying multi-objective optimization. We discuss the significance of finding the right balance between exploitation and exploration strategies in active learning and suggest strategies to reduce the number of iterations across diverse scenarios. This kind of research aims to find metrics for a more effective application of active learning and is used here to advance the field of sustainable alloy design. This article is part of the discussion meeting issue ‘Sustainable metals: science and systems’.

The demand for strong, compact permanent magnets essential for the energy transition drives innovation in magnet manufacturing. Additive manufacturing, particularly Powder Bed Fusion of metals using a laser beam (PBF‐LB/M), offers potential for near‐net‐shaped Nd‐Fe‐B permanent magnets but often falls short compared to conventional methods. A less explored strategy to enhance these magnets is feedstock modification with nanoparticles. It is demonstrated that modifying a Nd‐Fe‐B‐based feedstock with 1 wt.% Ag nanoparticles boost the coercivity of the magnets to a record value of 935 ± 6 kA m⁻¹ without further post‐processing or heat treatments. Suitable volumetric energy densities for the PBF‐LB/M process are determined using finite element simulations predicting melt pool behavior and part density. Microstructural analyses reveal finer grain sizes and more equiaxed nanocrystalline structures due to the modification. Atom probe tomography identifies three phases in the Ag‐modified samples, with Ag forming nanophase regions with rare‐earth elements near the amorphous Zr‐Ti‐B‐rich intergranular phase, potentially decoupling the Nd₂Fe₁₄B primary phase. The study shows that superior magnetic properties primarily result from microstructure modification rather than part density. These findings highlight inventive material design approaches via feedstock surface modification to achieve superior magnetic performance in additively manufactured Nd‐Fe‐B magnets.