Conventional vision systems suffer from severe power and latency bottlenecks due to the physical separation of sensing and processing, which are challenges that in sensor computing seeks to resolve. However, existing neuromorphic hardware remains hindered by high programming energy and unidirectional photocurrent polarity. This work presents an energy efficient reconfigurable photodiode based on ferroelectric Hf_xZr_1-xO₂ (HZO) and ambipolar semiconductor WSe₂, achieving energy-efficient in-sensor computing.

Researchers from Xinjiang University and East China University of Science and Technology have developed a novel iron-based single-atom catalyst (ZnS@Fe-NSC) that overcomes the long-standing trade-off between high activity and long-term stability in electrochemical CO₂ reduction. By integrating sulfur doping and zinc sulfide (ZnS) nanoparticles within the catalyst architecture, the team achieved exceptional CO production efficiency with over 98% selectivity and maintained stable operation for 30 hours, paving the way for more durable and efficient energy conversion technologies.

Structures exposed to wind, waves or flowing water can be difficult to model and monitor, especially when reliable measurements are scarce. A research team has developed FSI-PINNs, a physics-informed neural network framework that links fluid pressure with structural motion. Using no more than 1% of a simulated dataset, the method predicted flow fields and beam displacement and identified unknown structural density, offering a possible new tool for engineering problems in data-scarce environments.

We have developed a facile dip-coating method for fabricating ordered two-dimensional (2D) self-assembled block copolymer (BCP) micelle templates on substrates. Through point-to-point adsorption on these micelle templates, 2D non-close-packed, short-range ordered gold nanoparticle (AuNP) arrays were successfully obtained. The study elucidates the governing principles of critical processing parameters: the effects of withdrawal speed and micelle concentration on template morphology, as well as the influence of nanoparticle size and micelle corona thickness on the electrostatic adsorption process, thereby enabling the controllable fabrication of large-area 2D non-close-packed nanoparticle arrays. When employed as nanoscale floating gates in non-volatile memory devices, these BCP@AuNP composite arrays demonstrate superior performance—specifically, the incorporation of AuNPs broadens the memory window and enhances device operational stability compared to pure BCP micelle template layers. This strategy offers distinct advantages including operational simplicity, high efficiency, broad applicability, and cost-effectiveness, without requiring additional pre- or post-treatment steps. The entire process demonstrates substantial potential for industrial-scale fabrication of controllable 2D non-close-packed nanoparticle arrays.

Since 2014, the hot oscillatory pressure has improved density and mechanical properties in various composites, yet the synergy between frequency and amplitude in modifying microstructure remains unresolved. Here, we integrate densification kinetics, metal‑fatigue theory, and viscoelasticity to systematically dissect the cooperative effects of oscillatory parameters during sintering. Our approach achieves near‑full density (>99.5% relative density) and increases flexural strength by 21% over hot pressing, while elucidating the governing mechanisms. These findings offer a rational framework for optimizing oscillation cycles, directly benefiting the production of high‑performance ceramics.