Tensor operations dominate modern computational workloads, yet their further acceleration demands hardware platforms with greater parallelism. Although photonic computing provides a compelling route for parallel processing, fully exploiting all native multiplexing dimensions of optical fields is impeded by the challenges in routing and programming light in all dimensions simultaneously. Here we introduce FieldCore, a fully multiplexed photonic tensor core that jointly harnesses wavelength, radio-frequency, guided-mode, time and space dimensions, thereby enabling parallelism to scale multiplicatively within a single optical field. Enabled by inverse-designed silicon photonics, FieldCore preserves a uniform programmed computation across all multiplexed channels in parallel. Experimentally, we validate and benchmark its performance from ultra-high-baudrate arithmetic operations to high-fidelity image convolution and parallel handwritten-digit recognition. We further use FieldCore to unlock applications that naturally require high-dimensional data processing, such as high-dimensional hyperspectral classification and massively parallel mechanical fault diagnosis. Our FieldCore supports an estimated aggregate compute throughput of 69.12 tera operations per second (TOPS) and accommodates up to 1,800 parallel input streams within a single core, establishing a scalable paradigm for fully multiplexed photonic tensor computing and AI inference.

The escalating demands of compute-intensive applications urgently necessitate the adoption of optical interconnect technologies to overcome bottlenecks in scaling computing systems. This requires fully exploiting the inherent parallelism of light across scalable dimensions for data loading. Here we experimentally demonstrate a synergy of wavelength- and mode- multiplexing combined with high-order modulation formats to achieve multi-tens-of-terabits-per-second optical interconnects using foundry-compatible silicon photonic circuits. Implementing an edge-guided analog-and-digital optimization method that integrates high efficiency with fabrication robustness, we achieve the inverse design of mode multiplexers based on digital metamaterial waveguides. Furthermore, we employ a packaged five-mode multiplexing chip, achieving a single-wavelength interconnect capacity of 1.62 Tbit s-1 and a record-setting multi-dimensional interconnect capacity of 38.2 Tbit s-1 across 5 modes and 88 wavelength channels, with high-order formats up to 8-ary pulse-amplitude-modulation (PAM). This study highlights the transformative potential of optical interconnect technologies to surmount the constraints of electronic links, thus setting the stage for next-generation datacenter and optical compute interconnects.

The escalating capacity limitations of conventional near-infrared telecommunication bands have spurred urgent investigations into wide-band optical communication systems spanning from the near-infrared to mid-infrared regimes. This has motivated the development of optical components combining broadband bandwidth with high-speed operation. The state-of-the-art modulators face challenges in achieving broad operational bandwidth due to waveguide dispersion and velocity mismatch. Here we demonstrate a thin-film lithium niobate (TFLN) electro-optic (EO) modulator with an unprecedented 800-nm operational bandwidth, covering the full O-U telecom bands and extending into the 2-μm regime. The TFLN modulator exhibits >67 GHz EO bandwidth across O-U bands (~100 GHz at O-/S-/C-/L-bands) and >50 GHz at 2-μm band. It enables single-lane exceeding 240 Gbps PAM-4 transmission across O-U bands and a record 170 Gbps PAM-4 transmission at 2-μm band. This breakthrough establishes TFLN as a compelling platform for multispectral photonics, bridging conventional telecom infrastructure with emerging 2-μm technologies for next-generation optical communications.

We demonstrate the world's first field-trial of bidirectional 200G TFDM coherent PON enabled by real-time FPGA-based reception with two subcarriers, achieving - 34-dBm downstream sensitivity and - 31-dBm upstream sensitivity at 4.5-dB subcarriers power-difference with 21-dB dynamic range.

The growing demands of emerging services are driving wireless communication systems toward higher data rates and capacity. However, the increasingly congested radio frequency (RF) spectrum presents a major challenge. Millimeter wave (mmWave) in the W-band (75110 GHz) offers promising solutions due to its wide available bandwidth and low atmospheric attenuation, making it an ideal candidates for future high-capacity wireless links. Direct full-photonic-aided mmWave systems, leveraging photonic up- and down-conversion techniques, enable the generation of high-quality, broadband mmWave signals and allow seamless integration with optical fiber networks. Such fiberradiofiber bridge architectures provide flexible extensions of fiber networks and various wireless applications. In this work, we present a comprehensive analysis of recent advances in photonics-aided mmWave systems and demonstrate a transparent polarization-division multiplexed (PDM) fiberradiofiber signaling bridge operating in the W-band based on direct full-photonic conversion technology. A single-channel system achieves a maximum data rate of 140 Gbit/s using 35-Gbaud/s 16-quadrature amplitude modulation (16QAM) and 155 Gbit/s using 31-Gbaud/s 32QAM over a 20-km transmission fiber, a 1-m wireless link, and a 10-km reception fiber. The bit error rate (BER) threshold is set to 2x10(-2) , compatible with soft-decision forward error correction (SD-FEC) with 15% overhead (OH). Furthermore, by employing optical polarization-division multiplexing and a 2x2 multiple input multiple output (MIMO) antenna array, we realize the full-coherent fiberwirelessfiber transmission of PDM signals. Under the same transmission configuration, the system achieves up to 66-Gbaud/s 16QAM, yielding a total data rate of 264 Gbit/s (33 Gbaud/s x two polarizations x4 bits/symbol). This demonstration surpasses a total capacity of 250 Gbit/s, significantly expanding the application potential of full-photonic mmWave wireless transmission architectures.

As the global push toward the next frontier of wireless communications accelerates, the envisioned next generation network is expected to exploit millimeter-wave (mmWave) spectra, such as the W-band, to satisfy the surging demand for ultra-high data rates and massive capacity. To address the requirements for high-capacity transmission and ultra-dense deployments, a pivotal enabler is the mmWave fiber-wireless integrated architecture, which modulates a mmWave carrier immediately after photodetection, thereby streamlining the active antenna unit (AAU) in the radio access network. Nevertheless, existing bidirectional mmWave demonstrations still operate in lower frequency ranges, exhibit intricate architectures, and show limited synergy with optical fiber fronthaul. To address these limitations, we propose a W-band bidirectional seamless fiber-wireless integration system that implements full-photonic up-/down-conversion at the fiber optic network side. Leveraging the intrinsic ultra-wide bandwidth of photonic devices, the scheme enables direct generation and reception of broadband mmWave signals, eliminating electronic mixers and RF oscillators in the AAU, thereby reducing hardware complexity and mitigating bandwidth bottlenecks as well as electromagnetic-compatibility concerns. A centralized light-source combined with wavelength reusing further removes the need for additional lasers in the AAU. Proof-of-concept experiments demonstrate high data-rate transmission at 45.2 Gbps uplink and 128 Gbps downlink, underscoring the potential of the proposed photonic, bidirectional, seamlessly integrated fiber-wireless architecture to enhance capacity and enable flexible deployment of mmWave links in next-generation mobile networks.

Coherent passive optical network (PON) is a promising architecture that offers high sensitivity and support for advanced modulation formats. However, fully coherent systems have high complexity and cost, making them unsuitable for cost-sensitive user-end applications. Additionally, in coherent access architectures, the lasers of optical network units (ONUs) in the upstream are typically kept on rather than being switched off to avoid wavelength drift, with the Mach-Zehnder modulators (MZMs) biased at the null point for modulation. However, the limited extinction ratio (ER) of the MZM introduces direct current (DC) leakage, which degrades system performance and limits the number of users the network can support. To address this issue, we design and fabricate the first silicon-photonics-based ultimate-simplified coherent-optics chip for ONU. The simplified transmitter employs a single MZM combined with a high-speed optical switch (OS), while the simplified receiver adopts a hybrid and a standalone photodetector (PD) for coherent detection with Alamouti coding. By integrating the OS, the upstream achieves a maximum ER of 63.93 dB, while the OS switching speed meets the guard interval requirements of PON systems. Compared to a traditional single MZM with a 26 dB ER, our design achieves an additional 37.93 dB improvement, which significantly increases the user capacity in PON systems. Comprehensive experiments were conducted to validate the proposed design. Ultimately, leveraging this ultimate-simplified coherent-optics chip, a 100 Gbps bidirectional transmission was successfully demonstrated with a power budget exceeding 29 dBm, verifying its feasibility and cost-effectiveness for next-generation coherent optical access networks. (c) 2026 Chinese Laser Press

Digital-analog radio-over-fiber (DA-RoF) has emerged as a promising fronthaul solution that combines the high spectral efficiency of analog transmission with the robustness of digital transmission. However, the performance of DA-RoF critically depends on several tightly coupled parameters, including the rounding factor (RF), scaling factor (SF), geometric shaping (GS) factor, and pre-equalization taps coefficients, which jointly affect quantization noise, nonlinear distortion, and bandwidth-induced inter-symbol interference (ISI). Conventional grid search-based optimization is computationally prohibitive and impractical for optical communication. In this work, we propose a reinforcement-learning (RL)-enabled DA-RoF fronthaul agent architecture, capable of autonomously learning optimal transmitter parameters from end-to-end signal-to-noise ratio (SNR) feedback without a differentiable channel model. Experimental results demonstrate that the trained agent steadily improves SNR through sequential decision making and outperforms baseline, achieving 2.7-dB SNR improvement for 1- to 4-order DA-RoF transmission, reaching final SNR of 35.8 dB, 42.9 dB, 53.8 dB, and 63.2 dB and supporting 1024-, 4096-, 16384-, 65536-quadrature amplitude modulation (QAM) format, respectively. These results validate that the proposed RL-enabled framework provides online, scalable, and hardware-efficient parameter optimization for DA-RoF fronthaul systems, paving the way toward high-order modulation format and intelligent next-generation radio access networks.

Forecasting vehicle behavior within complex traffic environments is pivotal within Intelligent Transportation Systems (ITS). Though this technology plays a significant role in alleviating the prevalent operational difficulties in logistics and transportation systems, the precise prediction of vehicle trajectories still poses a substantial challenge. To address this, our study introduces the Spatio Temporal Attention-based methodology for Target Vehicle Trajectory Prediction (STATVTPred). This approach integrates Global Positioning System(GPS) localization technology to track target movement and dynamically predict the vehicle's future path using comprehensive spatio-temporal trajectory data. We map the vehicle trajectory onto a directed graph, after which spatial attributes are extracted via a Graph Attention Networks (GATs). The Transformer technology is employed to yield temporal features from the sequence. These elements are then amalgamated with local road network structure maps to filter and deliver a smooth trajectory sequence, resulting in precise vehicle trajectory prediction. This study validates our proposed STATVTPred method on T-Drive and Chengdu taxi-trajectory datasets. STATVTPred achieves an AMR of 73.07% on the Beijing dataset, surpassing the Transformer by 6.38% and the LSTM Encoder-Decoder by 37.45%, while also reducing Distance Error (DE) by 26.93% and 20.95% in Beijing and Chengdu, respectively, also much lower than the baseline results. This is expected to establish STATVTPred as a new approach for handling trajectory prediction of targets in logistics and transportation scenarios, thereby enhancing prediction accuracy. Note to Practitioners-This article is motivated by the need for a high-precision trajectory prediction method for unmanned aerial vehicles (UAVs) in complex traffic scenarios. Especially in scenarios such as urban canyons or when tracking a target vehicle is lost, predicting the movement trajectory of the UAV with high precision and speed generates significant interest. In practice, as the trajectory of a target movement isn't always known, and solely relying on satellite positioning is limited by the application scenario. It presents a challenge to determine the movement trajectory of the lost UAV in a short period of time. In response to these issues, we integrated GPS positioning technology, extracted spatial attributes through a graph attention network, and used Transformer technology to extract time features from a sequence. By combining these two kinds of attributes, we obtained comprehensive spatiotemporal trajectory data, thereby dynamically predicting the future path of the UAV. The target trajectory prediction technology developed in this paper can be expanded to multiple UAV trajectory prediction scenarios to achieve coordinated path optimization, thereby improving overall performance. This paper presents an innovative approach that would benefit practitioners in the field by providing a solution to the challenging problem of high precision trajectory prediction in complex traffic scenarios.

We propose two compact (6x8 mu m(2)) silicon nitride wavelength demultiplexers bridging 600/1550 nm and 600/1310 nm bands via inverse design. The devices exhibit insertion losses below 0.75 /0.68 dB and crosstalks below -18.6 /-20.6 dB at their center wavelengths, with a minimum feature size of 80 nm.

The coherent passive optical network (PON) is a cost-effective point-to-multipoint solution to accommodate the ever-increasing traffic demands in the future optical access network. Recently, time and frequency division multiplexing (TFDM) coherent PON have attracted increasing interest, due to its advantages of providing low latency access and flexible bandwidth allocation. One critical challenge for TFDM-PON system is the upstream burst-mode digital signal processing (DSP). DSP should converge fast in the specific time slot to ensure low latency for subcarrier allocation. The conventional subcarrier recognition and demultiplexing are realized in the frequency domain. The process of time-frequency conversion introduces extra computational complexity if the burst-mode DSP use time-domain algorithms. In this paper, we propose and experimentally demonstrate a data-aided DSP scheme for burst-mode detection in coherent TFDM-PON. Subcarrier demultiplexing is implemented in the time domain by using coarse and fine filtering. Coarse and fine filtering are achieved by convolving the signal with finite impulse response (FIR) pulse-shaping filters. Least square (LS) algorithm is used for channel estimation achieving fast convergence. With a 79.36-ns training sequence, -37.5-dBm and -29-dBm sensitivities are achieved for QPSK and 16QAM signal-based 200G TFDM-PON with four subcarriers after 20-km fiber transmission.

We propose and demonstrate a bidirectional W-band seamless fiber-wireless integrated system, utilizing all-optical up-&down-conversions at fiber optic network side, achieving the access data rates of 45.2 and 128 Gbps in UL and DL, respectively.

Towards higher capacity for the next generation Ethernet in datacenters, transmitting high-baud-rate signal is necessary. This work presents a band-interleaving receiver with subcarrier multiplexing (SCM) signaling for systems with excessive transmitter bandwidth and severely limited receiver bandwidth. In the proposed scheme, an ultra-high-bandwidth SCM signal is divided into two sub-bands, and modulated onto the optical carrier through an external modulator. The two sub-bands of the SCM signal are extracted by a wavelength selective switch, and then detected by a direct detection (DD) receiver and a lite-coherent receiver, respectively. This approach halves the receiver-side bandwidth requirement. As a proof of concept, we experimentally achieve the transmission of a 100-GHz SCM QAM signal over 500-m single-mode optical fiber (SSMF), attaining a highest net data rate of 428 Gbps above the 0.8798 NGMI threshold. To our knowledge, this represents a successful demonstration of signal transmission with a Nyquist bandwidth of up to 100 GHz and a net data rate exceeding 400 Gbps per wavelength in a system with severely limited receiver bandwidth of about 61 GHz.

The ever-increasing demand for improved speeds, expanded bandwidth, and reduced latency in emerging applications has not only driven the recent advanced algorithm breakthroughs but also fostered closer integration of metro and access networks. This work proposes a novel all-optical metro-access integration network (MAIN) enabled by coherent digital subcarrier multiplexing technology. This specially designed architecture can effectively eliminate the latency and jitters caused by the optical-electrical-optical conversion in the conventional scheme and save wavelength resources at the same time. For experimental validation, we successfully demonstrate 400G bidirectional coherent transmission within our proposed architecture, involving three nodes in the metro network and the access network as an example. Also, the related problems during the transmission are extensively discussed with the experimental results. Finally, we reach an aggregation rate of 4 x 100 Gbps using a DP-16QAM signal. By integrating a SOA into the transmitter of the ONU, a power budget of 32.6 dB with a 29 dB dynamic range is achieved. In addition, the experimental results also show that the proposed all-optical MAIN architecture can be smoothly combined with the TFDM scheme to further enhance the flexibility, which also paves the way for further research on the next-generation coherent metro and access network.

The E/W band millimeter-wave (MMW) emerges as a promising frequency band with many advantages such as the possession of abundant spectrum resources, low atmospheric attenuation, and good directionality. Meanwhile, full-photonic based MMW systems offer the potential to further reduce the complexity of electronic down-conversion and enhance overall system bandwidth, and allow seamless integration with the fiber-optic network. In this paper, we proposed an E/W band MMW communication architecture based on full-photonic up- and down-conversions, utilizing a frequency- and phase-locked optical two-tone generator at both the transmitting and receiving end. We successfully demonstrated a long-distance field trial of E/W-Band wireless transmission, achieving 26.8-km wireless delivery of 32-Gb/s quadrature phase shift keying (QPSK) signal, and the largest net-rate-distance-product up to 745.74 Gb/skm per antenna per channel. The long-distance transmission performance of different modulation formats was also studied. We also achieved a 22-Gb/s QPSK transmission over a longer distance of 30.4 km, with a net-rate and distance product reaching 581.56 Gb/skm. Moreover, this paper analyzes the complex sea surface channel and experimentally verifies the influence of the sea surface environment on the channel, which is instructive for long-distance MMW sea surface communications.

We report a fast transceiver online IQ-skew estimation method with special designed preamble in upstream burst-mode DSP, achieving transceiver estimation errors within ±0.3 ps after average operation in the 400G coherent TFDM-PON upstream.

The rapid advancement of generative artificial intelligence (AI) in recent years has profoundly reshaped modern lifestyles, necessitating a revolutionary architecture to support the growing demands for computational power. Cloud computing has become the driving force behind this transformation. However, it consumes significant power and faces computation security risks due to the reliance on extensive data centers and servers in the cloud. Reducing power consumption while enhancing computational scale remains persistent challenges in cloud computing. Here, we propose and experimentally demonstrate an optical cloud computing system that can be seamlessly deployed across edge-metro network. By modulating inputs and models into light, a wide range of edge nodes can directly access the optical computing center via the edge-metro network. The experimental validations show an energy efficiency of 118.6 mW/TOPs (tera operations per second), reducing energy consumption by two orders of magnitude compared to traditional electronic-based cloud computing solutions. Furthermore, it is experimentally validated that this architecture can perform various complex generative AI models through parallel computing to achieve image generation tasks.

We demonstrate an all-optical metro-access-mobile integrated network enabled by digital subcarrier multiplexing, achieving 400 Gbps aggregation rate by 4SCs of 12.5 GBaud/SC DP-16QAM, with 31.5-dB and 24-dB power budgets for PON and mobile RAN, respectively.