For metal nanofilms composed of nanocrystals, the multiple deformation mechanisms will coexist and bring unique and complex elastic-plastic and fracture mechanical properties. By successfully fabricating large quantities of uniform doubly-clamped suspended gold (Au) nanobeams with different thicknesses and nanograin sizes, we obtain full-spectrum mechanical features with statistical significance by combining atomic force microscopy (AFM) nanoindentation experiments, nonlinear theoretical model, and numerical simulations. The yield and breaking strengths of the Au nanobeams have a huge increase by nearly an order of magnitude compared with bulk Au and exhibit strong nonlinear effects, and the corresponding strong-yield ratio is up to 4, demonstrating extremely high strength reserve and vibration resistance. The strong-yield ratio gradually decreases with decreasing thickness, identifying a conversion of the failure type from ductile to brittle. Interestingly, the Hall–Petch relationship has been identified to be still valid at the nanoscale, and K in the equation reaches 4.8 Gpa·nm^1/2, nearly twice of bulk nanocrystalline Au, which is ascribed to the coupling effect of nanocrystals and nanoscale thickness.

When a laser beam writes on a metallic film, it usually coarsens and deuniformizes grains because of Ostwald ripening, similar to the case of annealing. Here we show an anomalous refinement effect of metal grains: A metallic silver film with large grains melts and breaks into uniform, close-packed, and ultrafine (~ 10 nm) grains by laser direct writing with a nanoscale laser spot size and nanosecond pulse that causes localized heating and adaptive shock-cooling. This method exhibits high controllability in both grain size and uniformity, which lies in a linear relationship between the film thickness (h) and grain size (D), D ∝ h. The linear relationship is significantly different from the classical spinodal dewetting theory obeying a nonlinear relationship (D ∝ h^5/3) in common laser heating. We also demonstrate the application of such a silver film with a grain size of ~ 10.9 nm as a surface-enhanced Raman scattering chip, exhibiting superhigh spatial-uniformity and low detection limit down to 10⁻¹⁵ M. This anomalous refinement effect is general and can be extended to many other metallic films.

The atomristor (monolayer two-dimensional (2D)-material memristor) is competitive in high-speed logic computing due to its binary feature, lower energy consumption, faster switch response, and so on. Yet to date, all-atomristor logic gates used for logic computing have not been reported due to the poor consistency of different atomristors in performance. Here, by studying band structures and electron transport properties of MoS₂ atomristor, a comprehensive memristive mechanism is obtained. Guided by the simulation results, monolayer MoS₂ with moderated defect concentration has been fabricated in the experiment, which can build atomristors with high performance and good consistency. Based on this, for the first time, MoS₂ all-atomristor logic gates are realized successfully. As a demonstration, a half-adder based on the logic gates and a binary neural network (BNN) based on crossbar arrays are evaluated, indicating the applicability in various logic computing circumstances. Owing to shorter transition time and lower energy consumption, all-atomristor logic gates will open many new opportunities for next-generation logic computing and data processing.

The composition of materials in a micro-/nano-devices plays a key role in determining their mechanical, physical, and chemical properties. Especially, for devices with a compositional change on nanoscale which can often be achieved by point-by-point direct writing technology using a focused ion beam (FIB), electron beam (EB), or laser beam (LB), but so far, nanoscale composition analysis of a large-area micro/nano structures with a variation composition remains a big challenge in cost, simpleness, and flexibility. Here we present a feasible route to realize large-area composition analysis with nanoscale spatial resolution by using Raman spectroscopy. We experimentally verified the capability of this method by analyzing a complex Sn-SnO_x system of a microscale grayscale mask with nanoscale spatial resolution of composition. Further analyses using Auger electron spectroscopy, transmission electron microscopy, and atomic force microscopy indicated the effectiveness and practicality of our method. This work opens up a way to analyze the composition of a large-area complex system at a nanoscale spatial resolution, and the method can be extended to many other material systems.

In recent years, trap-related interfacial transport phenomena have received great attention owing to their potential applications in resistive switching devices and photo detectors. Not long ago, one new type of memristive interface that is composed of F-doped SnO₂ and Bi₂S₃ nano-network layers has demonstrated a bivariate-continuous-tunable resistance with a swift response comparable to the one in neuron synapses and with a brain-like memorizing capability. However, the resistive mechanism is still not clearly understood because of lack of evidence, and the limited improvement in the development of the interfacial device. By combining I-V characterization, electron energy-loss spectroscopy, and first-principle calculation, we studied in detail the macro/micro features of the memristive interface using experimental and theoretical methods, and confirmed that its atomic origin is attributed to the traps induced by O-doping. This implies that impurity-doping might be an effective strategy for improving switching features and building new interfacial memristors.

A memristor that can emulate biological synapses is a promising basic-processing unit in neural-network computation. Here we propose a new-conceptual memristor based on a memoristive interface composed of two types of non-memristive materials, successfully realizing continuously tunable resistance controlled by both voltage (current) and applied time of a single pulse with a swift response comparable with synapses. The brain-like memorizing capability of the memristor is demonstrated. The memoristive mechanism in the interface is thought to be dominated by a Schottky barrier tuned by the capture/release of the carriers in interface traps with dispersive energy.