Memory and logic in the same device: Future artificial intelligence applications and data-intensive computations require the development of neuromorphic systems beyond traditional heterogeneous device architectures. Physical separation between a peripheral signal-processing unit and a memory-operating unit is one of the main bottlenecks of heterogeneous architectures, blocking further improvements in efficient resistance matching, energy consumption, and integration compatibility. Tong et al . present a transistor-memory architecture based on a homogeneous tungsten selenide-on-lithium niobate device array (see the Perspective by Rao and Tao). Analog peripheral signal preprocessing and nonvolatile memory were possible within the same device structure, promising diverse neuromorphic functionalities and offering potential improvements in neuromorphic systems on-chip. —YS

Emerging wearable devices would benefit from integrating ductile photovoltaic light-harvesting power sources. In this work, we report a small-molecule acceptor (SMA), also known as a non–fullerene acceptor (NFA), designed for stretchable organic solar cell ( s -OSC) blends with large mechanical compliance and performance. Blends of the organosilane-functionalized SMA BTP-Si4 with the polymer donor PNTB6-Cl achieved a power conversion efficiency (PCE) of >16% and ultimate strain (ε _u ) of >95%. Typical SMAs suppress OSC blend ductility, but the addition of BTP-Si4 enhances it. Although BTP-Si4 is less crystalline than other SMAs, it retains considerable electron mobility and is highly miscible with PNTB6-Cl and is essential for enhancing ε _u . Thus, s -OSCs with PCE > 14% and operating normally under various deformations (>80% PCE retention under an 80% strain) were demonstrated. Analysis of several SMA-polymer blends revealed general molecular structure–miscibility–stretchability relationships for designing ductile blends.

Microwave photonic filter (MPF) is one of the key fundamental subsystems in microwave photonics, and it shows great potentiality in numbers of applications such as optoelectronic oscillator, microwave frequency measurement, and so forth. A narrowband bandpass MPF is highly desirable with its high selectivity in microwave photonic applications. However, a resonator with a high quality factor (>10 ⁶ ) is very difficult to fabricate on the mature silicon photonics platform without optimization, thus restraining the applications of the MPF. In this paper, we propose and experimentally demonstrate a tunable sub-gigahertz bandwidth MPF based on an ultrahigh quality factor silicon microring resonator. Most performance aspects of the MPF are well-balanced. A full width at half-maximum bandwidth of 170 MHz is achieved thanks to the ultrahigh total quality factor as 1.14 × 10 ⁶ of the microring resonator, and the average waveguide loss and the intrinsic Q factor of the microring resonator are calculated as 0.25 dB/cm and 2.67 × 10 ⁶ , respectively. The corresponding rejection ratio of the bandpass filter is 26.5 dB. Besides, the central frequency of the filter could be continuously tuned from 2.0 to 18.4 GHz with a microheater fabricated upon the microring, and a 16.4 GHz tuning range is achieved with the maximum power consumption of 14.4 mW. The device area is ~0.05 mm ² .

Silicon photonics is promising for artificial neural networks computing owing to its superior interconnect bandwidth, low energy consumption and scalable fabrication. However, the lack of silicon-integrated and monitorable optical neurons limits its revolution in large-scale artificial neural networks. Here, we highlight nonlinear germanium-silicon photodiodes to construct on-chip optical neurons and a self-monitored all-optical neural network. With specifically engineered optical-to-optical and optical-to-electrical responses, the proposed neuron merges the all-optical activation and non-intrusive monitoring functions in a compact footprint of 4.3 × 8 μm². Experimentally, a scalable three-layer photonic neural network enables in situ training and learning in object classification and semantic segmentation tasks. The performance of this neuron implemented in a deep-scale neural network is further confirmed via handwriting recognition, achieving a high accuracy of 97.3%. We believe this work will enable future large-scale photonic intelligent processors with more functionalities but simplified architecture.