We investigate whether large language models can perform the creative hypothesis generation that human researchers regularly do. While the error rate is high, generative AI seems to be able to effectively structure vast amounts of scientific knowledge and provide interesting and testable hypotheses. The future scientific enterprise may include synergistic efforts with a swarm of “hypothesis machines”, challenged by automated experimentation and adversarial peer reviews.

Boron doping, combined with neutron capture in fission reactors, has been used to simulate the helium effect on fusion structural materials. However, inhomogeneous helium bubble formation was often observed due to boron segregation to grain boundaries. The excess radiation displacements due to ¹⁰B(n, α)⁷Li reaction, the high-energy lithium and helium ions, also were not accounted for, which can significantly accelerate the displacements-per-atom (dpa) accumulation alongside helium production (appm). Hereby an isotopically pure ¹⁰B doping approach is proposed to simulate the extreme environment inside fusion reactors with a high He appm-to-dpa ratio of about 10, which is about 10² × larger than in fission reactors. Computational modeling showed that ~13% of total radiation displacement was induced by ¹⁰B(n, α)⁷ Li in the case of 1 000 appm ¹⁰B doped Fe samples, which becomes even greater with increasing ¹⁰B loading. Spatially homogenous radiation damage and helium generation are predicted for grain sizes less than 1 μm, even if the boron partially formed precipitates or segregates on grain boundaries. Feasibility studies with various ¹⁰B doping (and ²³⁵U-codoping) levels in research reactors showed the estimated helium generation and radiation damage would significantly mimic fusion conditions and greatly expedite fusion materials testing, from many years down to months.

Rechargeable solid-state Li metal batteries demand ordered flows of Li-ions and electrons in and out of solid structures, with repeated waxing and waning of Li_BCC phase near contact interfaces which gives rise to various electro-chemo-mechanical challenges. There have been approaches that adopt three-dimensional (3D) nanoporous architectures consisting of mixed ion-electron conductors (MIECs) to combat these challenges. However, there has remained an issue of Li_BCC nucleation at the interfaces between different solid components (e.g., solid electrolyte/MIEC interface), which could undermine the interfacial bonding, thereby leading to the evolution of mechanical instability and the loss of ionic/electronic percolation. In this regard, the present work shows that the Li-ion and electron insulators (LEIs) that are thermodynamically stable against Li_BCC could combat such challenges by blocking transportation of charge carriers on the interfaces, analogous to dielectric layers in transistors. We searched the ab initio database and have identified 48 crystalline compounds to be LEI candidates (46 experimentally reported compounds and 2 hypothetical compounds predicted to be stable) with a band gap greater than 3 eV and vanishing Li solubility. Among these compounds, those with good adhesion to solid electrolyte and mixed ion-electron conductor of interest, but are lithiophobic, are expected to be the most useful. We also extended the search to Na or K metal compatible alkali-ion and electron insulators, and identified some crystalline compounds with a property to resist corresponding alkali-ions and electrons.

The adhesion and wetting between metal and ceramic is a basic problem in materials science and engineering. For example, past materials selection for metal-ceramic composites has relied on random trials and heuristics due to a limited understanding of their adhesion; the large chemical/structural variability that such interfaces can have hinders the identification of the governing factors. Here based on literature data, we have developed a database with ~1, 000 experimentally measured wetting angles at different temperatures and atmospheric conditions, and come up with a model for the wettability of ionocovalent ceramics (ICs) by metals using a machine learning (ML) algorithm. The random forest model uses the testing temperature and ~40 features generated based on the chemical compositions of the metal and the ceramic as predictors and exhibits strong predictive power with an R² of ~0.86. Moreover, this model and the featurization code are integrated into a single computational pipeline to enable (1) predicting metal-IC wettability of interest and (2) high-throughput searching of ICs with the desired wettability by certain metals in the entire Inorganic Crystallographic Structure Database. As a demonstration of this pipeline, the wettability of a Li-ion and electron insulator (LEI), CaO, by molten Li is estimated and compared with ab initio molecular dynamics simulation result. This ML pipeline can serve as a practical tool for methodical design of materials in systems where certain metal-ceramic wettability is desired.

Vanadium dioxide (VO₂) has emerged as a promising micro-actuator material for its large amplitude and high work density across the transition between the insulating (M1 and M2) and metallic (R) phase. Even though M2–R transition offers about 70% higher transformation stress than M1–R structural phase transition, the application of the M2 phase in the micro-actuators is hindered by the fact that previously, M2 phase can only stay stable under tensile stress. In this work, we propose and verify that by synthesizing the VO₂ nanowires under optimized oxygen-rich conditions, stoichiometry change can be introduced into the nanowires (NWs) which in turn yield a large number free-standing single-crystalline M2-phase NWs stable at room temperature. In addition, we demonstrate that the output stress of the M2-phase NWs is about 65% higher than that of the M1-phase NWs during their transition to R phase, quite close to the theoretical prediction. Our findings open new avenues towards enhancing the performance of VO₂-based actuators by using M2–R transition.

Few-layer two-dimensional (2D) materials usually have different (meta)-stable stacking patterns, which have distinct electronic and optical properties. Inspired by optical tweezers, we show that a laser with selected frequency can modify the generalized stacking-fault energy landscape of bilayer hexagonal boron nitride (BBN), by coupling to the slip-dependent dielectric response. Consequently, BBN can be reversibly and barrier-freely switched between its stacking patterns in a controllable way. We simulate the dynamics of the stacking transition with a simplified equation of motion and demonstrate that it happens at picosecond timescale. When one layer of BBN has a nearly-free surface boundary condition, BBN can be locked in its metastable stacking modes for a long time. Such a fast, reversible and non-volatile transition makes BBN a potential media for data storage and optical phase mask.

It is found that several layer-phase group-Ⅲ monochalcogenides, including GaS, GaSe, and InSe, are piezoelectric in their monolayer form. First-principles calculations reveal that the piezoelectric coefficients of monolayer GaS, GaSe, and InSe (2.06, 2.30, and 1.46 pm·V^-1) are of the same order of magnitude as previously discovered two-dimensional (2D) piezoelectric materials such as boron nitride (BN) and MoS₂ monolayers. This study therefore indicates that a strong piezoelectric response can be obtained in a wide range of two-dimensional materials with broken inversion symmetry. The co-existence of piezoelectricity and superior photo-sensitivity in these monochalcogenide monolayer semiconductors means they have the potential to allow for the integration of electromechanical and optical sensors on the same material platform.

Using nanoscale electrical-discharge-induced rapid Joule heating, we developed a method for ultrafast shape change and joining of small-volume materials. Shape change is dominated by surface-tension-driven convection in the transient liquid melt, giving an extremely high strain rate of ~10⁶ s^–1. In addition, the heat can be dissipated in small volumes within a few microseconds through thermal conduction, quenching the melt back to the solid state with cooling rates up to 10⁸ K·s^-1. We demonstrate that this approach can be utilized for the ultrafast welding of small-volume crystalline Mo (a refractory metal) and amorphous Cu₄₉Zr₅₁ without introducing obvious microstructural changes, distinguishing the process from bulk welding.

The ability to fine-tune band gap and band inversion in topological materials is highly desirable for the development of novel functional devices. Here we propose that the electronic properties of free-standing nanomembranes of the topological crystalline insulators (TCI) SnTe and Pb_1-xSn_x(Se, Te) are highly tunable by engineering elastic strain and membrane thickness, resulting in tunable band gap and giant piezoconductivity. Membrane thickness governs the hybridization of topological electronic states on opposite surfaces, while elastic strain can further modulate the hybridization strength by controlling the penetration length of surface states. We propose a frequency-resolved infrared photodetector using force-concentration induced inhomogeneous elastic strain in TCI nanomembranes with spatially varying width. The predicted tunable band gap accompanied by strong spin-textured electronic states will open new avenues for fabricating piezoresistive devices, infrared detectors and energy-efficient electronic and spintronic devices based on TCI nanomembrane.