Search for: All records

Data-driven approaches to materials exploration and discovery are building momentum due to emerging advances in machine learning. However, parsimonious representations of crystals for navigating the vast materials search space remain limited. To address this limitation, we introduce a materials discovery framework that utilizes natural language embeddings from language models as representations of compositional and structural features. The contextual knowledge encoded in these language representations conveys information about material properties and structures, enabling both similarity analysis to recall relevant candidates based on a query material and multi-task learning to share information across related properties. Applying this framework to thermoelectrics, we demonstrate diversified recommendations of prototype crystal structures and identify under-studied material spaces. Validation through first-principles calculations and experiments confirms the potential of the recommended materials as high-performance thermoelectrics. Language-based frameworks offer versatile and adaptable embedding structures for effective materials exploration and discovery, applicable across diverse material systems.

 

Strain engineering in two-dimensional (2D) materials is a powerful but difficult to control approach to tailor material properties. Across applications, there is a need for device-compatible techniques to design strain within 2D materials. This work explores how process-induced strain engineering, commonly used by the semiconductor industry to enhance transistor performance, can be used to pattern complex strain profiles in monolayer MoS2 and 2D heterostructures. A traction–separation model is identified to predict strain profiles and extract the interfacial traction coefficient of 1.3 ± 0.7 MPa/μm and the damage initiation threshold of 16 ± 5 nm. This work demonstrates the utility to (1) spatially pattern the optical band gap with a tuning rate of 91 ± 1 meV/% strain and (2) induce interlayer heterostrain in MoS2–WSe2 heterobilayers. These results provide a CMOS-compatible approach to design complex strain patterns in 2D materials with important applications in 2D heterogeneous integration into CMOS technologies, moiré engineering, and confining quantum systems. 

Computation-guided selection of dopants enables the transformation of Hg₂GeTe₄from intrinsic to degenerate carrier concentrations and the thermoelectric performance is assessed experimentally.

 

In oxide materials, an increase in oxygen vacancy concentration often results in lattice expansion, a phenomenon known as chemical expansion that can introduce detrimental stresses and lead to potential device failure. One factor often implicated in the chemical expansion of materials is the degree of localization of the multivalent cation electronic states. When an oxygen is removed from the lattice and a vacancy forms, it is believed that the two released electrons reduce multivalent cations and expand the lattice, with more localized cation states resulting in larger expansion. In this work, we computationally and experimentally studied the chemical expansion of two Pr-based perovskites that exhibit ultra-low chemical expansion, PrGa 1− x Mg x O 3− δ and BaPr 1− x Y x O 3− δ , and their parent compounds PrGaO 3− δ and BaPrO 3− δ . Using density functional theory, the degree of localization of the Pr-4f electrons was varied by adjusting the Hubbard U parameter. We find that the relationship between Pr-4f electron localization and chemical expansion exhibits more complexity than previously established. This relationship depends on the nature of the states filled by the two electrons, which may not necessarily involve the reduction of Pr. F ′-center defects can form if the reduction of Pr is unfavorable, leading to smaller chemical expansions. If hole states are present in the material, the states filled by the electrons can be Pr-4f and/or O-2p hole states depending on the degree of Pr-4f localization. The O-2p holes are more delocalized than the Pr-4f holes, resulting in smaller chemical expansions when the O-2p holes are filled. X-ray photoelectron spectroscopy reveals low concentrations of Pr 4+ in PrGa 0.9 Mg 0.1 O 3− δ and BaPr 0.9 Y 0.1 O 3− δ , supporting the possible role of O-2p holes in the low chemical expansions exhibited by these materials. This work highlights the non-trivial effects of electron localization on chemical expansion, particularly when hole states are present, pointing to design strategies to tune the chemical expansion of materials. 

Successful dopability in AgInTe₂requires careful navigation of the compensating intrinsic defects to maximize dopant solubility and efficiency.

 

While machine learning has emerged in recent years as a useful tool for the rapid prediction of materials properties, generating sufficient data to reliably train models without overfitting is often impractical. Towards overcoming this limitation, we present a general framework for leveraging complementary information across different models and datasets for accurate prediction of data-scarce materials properties. Our approach, based on a machine learning paradigm called mixture of experts, outperforms pairwise transfer learning on 14 of 19 materials property regression tasks, performing comparably on four of the remaining five. The approach is interpretable, model-agnostic, and scalable to combining an arbitrary number of pre-trained models and datasets to any downstream property prediction task. We anticipate the performance of our framework will further improve as better model architectures, new pre-training tasks, and larger materials datasets are developed by the community.