With the continuous advancement of the new energy sector, direct methanol fuel cells (DMFCs) have attracted significant research interest. However, the development of DMFCs is hindered by the reliance on platinum-based anode catalysts, which suffer from high cost, intermediate-induced poisoning, and rapid performance degradation.

Titanium dioxide (TiO₂) is an important photocatalytic material, yet its performance is limited by bottlenecks in conventional preparation methods, such as high interfacial resistance and low specific surface area. Recently, a method combining laser cladding with electrochemical dealloying, reported by Frontiers of Materials Science, was used to fabricate porous TiO₂/Cu₂O heterojunction photoanodes from a Cu–Ti precursor.

With the continuous advancement of lithiumion batteries (LIBs), increasing demands have been placed on achieving ultrafast charging capability. Graphite, the dominant commercial anode material, suffers from sluggish ion transport kinetics and interfacial instability, which pose significant challenges to the fast-charging performance of LIBs. Previous strategies, including surface coatings and conductive network construction, often address these challenges in isolation, leading to an inherent tradeoff between structural stability and electrical conductivity. Xu et al. recently reported scalable onestep pyrolysis of pitch that concurrently constructs a conformal amorphouscarbon shell and an integrated graphene network (Gr) on graphite.

A new artificial intelligence (AI) framework offers a more reliable way to restore partially hidden objects in satellite imagery. 

Infrared focal plane arrays (FPAs) are image sensors, widely used in thermal imaging systems across various fields. Future advancements in FPAs are expected to require high data rates, which can cause heat leakage from electrical interconnects into FPAs, increasing noise and raising cooling requirements. In a new study, researchers demonstrated cryogenic vertical-cavity surface-emitting laser (VCSEL)-based optical interconnects that can achieve high-speed communication while reducing thermal leakage and power consumption.

SJTU researchers propose a dynamic MCPV model and CZone mechanism, slashing memory footprint by 90%.

Waveguide-based electro-optic (E-O) frequency combs are prized for their flexible pump wavelength, tunable repetition rate and near-100% conversion efficiency, but their spectral span is usually limited to a few nanometers. Researchers now report a broadband waveguide E-O comb based on "mode circulation": a single beam is recirculated through the same phase-modulation region eight times as four distinct waveguide modes (TE0–TE3). Enabled by a Z-propagation mode multiplexer that suppresses mode hybridization in anisotropic lithium-niobate-on-insulator waveguides, the device boosts the modulation index by ~7.5×, generating 128 comb lines across 25.6 nm at 25 GHz with an effective half-wave voltage of just 1.24 V and only 7 dB on-chip loss.

A new study published in the Chinese Journal of Aeronautics （https://doi.org/10.1016/j.cja.2025.103943） tries to break this deadlock. Researchers from Northwestern Polytechnical University and Tsinghua University have developed a novel modeling framework called differential equation model long-short term memory neural network (DEM-LSTM). This hybrid approach integrates the theoretical rigor of differential equations with the nonlinear mapping capabilities of LSTM neural network, offering a solution that is both accurate and physically interpretable.