Search for: All records

Abstract One compelling vision of the future of materials discovery and design involves the use of machine learning (ML) models to predict materials properties and then rapidly find materials tailored for specific applications. However, realizing this vision requires both providing detailed uncertainty quantification (model prediction errors and domain of applicability) and making models readily usable. At present, it is common practice in the community to assess ML model performance only in terms of prediction accuracy (e.g. mean absolute error), while neglecting detailed uncertainty quantification and robust model accessibility and usability. Here, we demonstrate a practical method for realizing both uncertainty and accessibility features with a large set of models. We develop random forest ML models for 33 materials properties spanning an array of data sources (computational and experimental) and property types (electrical, mechanical, thermodynamic, etc). All models have calibrated ensemble error bars to quantify prediction uncertainty and domain of applicability guidance enabled by kernel-density-estimate-based feature distance measures. All data and models are publicly hosted on the Garden-AI infrastructure, which provides an easy-to-use, persistent interface for model dissemination that permits models to be invoked with only a few lines of Python code. We demonstrate the power of this approach by using our models to conduct a fully ML-based materials discovery exercise to search for new stable, highly active perovskite oxide catalyst materials. 

Abstract In recent years,T_dtransition metal dichalcogenides have been heavily explored for their type‐II Weyl topology, gate‐tunable superconductivity, and nontrivial edge states in the monolayer limit. Here, the Fermi surface characteristics and fundamental transport properties of similarly structured 2M‐WSe₂bulk single crystals are investigated. The measurements of the angular dependent Shubnikov–de Haas oscillations, with support from first‐principles calculations, reveal multiple three‐ and two‐dimensional Fermi pockets, one of which exhibits a nontrivial Berry's phase. In addition, it is shown that the electronic properties of 2M‐WSe₂are similar to those of orthorhombic MoTe₂and WTe₂, having a single dominant carrier type at high temperatures that evolves into coexisting electron and hole pockets with near compensation at temperatures below 100 K, suggesting the existence of a Lifshitz transition. Altogether, the observations provide evidence towards the topologically nontrivial electronic properties of 2M‐WSe₂and motivate further investigation on the topological properties of 2Mtransition metal dichalcogenides in the atomically thin limit. 

Abstract The emergence of memristive behavior in amorphous–crystalline 2D oxide heterostructures, which are synthesized by atomic layer deposition (ALD) of a few‐nanometer amorphous Al₂O₃layers onto atomically thin single‐crystalline ZnO nanosheets, is demonstrated. The conduction mechanism is identified based on classic oxygen vacancy conductive channels. ZnO nanosheets provide a 2D host for oxygen vacancies, while the amorphous Al₂O₃facilitates the generation and stabilization of the oxygen vacancies. The conduction mechanism in the high‐resistance state follows Poole–Frenkel emission, and in the the low‐resistance state is fitted by the Mott–Gurney law. From the slope of the fitting curve, the mobility in the low‐resistance state is estimated to be ≈2400 cm²V⁻¹s⁻¹, which is the highest value reported in semiconductor oxides. When annealed at high temperature to eliminate oxygen vacancies, Al is doped into the ZnO nanosheet, and the memristive behavior disappears, further confirming the oxygen vacancies as being responsible for the memristive behavior. The 2D heterointerface offers opportunities for new design of high‐performance memristor devices.