YWYilun Wang

Yilun Wang

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Yilun Wang has a total of 137 co-authors. Here are the top co-authors by number of shared publications:

  • Xinliang Zhang

    Huazhong University of Science and Technology

    23 shared publications
  • Lei Ye

    Huazhong University of Science and Technology

    8 shared publications
  • Xinyu Huang

    School of Integrated Circuits and Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, Wuhan, China

    7 shared publications
  • Wentao Deng

    Wuhan National Laboratory for Optoelectronics and School of Optical and Electronic Information, Huazhong University of Science and Technology, Wuhan, China

    6 shared publications
  • Jianji Dong

    Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, Wuhan, China

    5 shared publications
  • Xiaoyan Gao

    Wuhan National Laboratory for Optoelectronics & School of Optical and Electronic Information, Huazhong University of Science and Technology, Wuhan, China

    5 shared publications

Latest Publications

  • Probabilistic photonic computing based on chaotic Kerr optical frequency combs

    by Qi Liang, Qiuquan Yan, Hao Ouyang, Chao Zhou, Hao Hao, Licheng Wang, Yilun Wang, Yuhua Tang, Jun Zhang, Tian Jiang

    Artificial neural networks (ANNs) have become powerful tools for deciphering complex data patterns. However, conventional ANNs can be viewed as point estimators that fail to capture predictive uncertainty, leading to overconfident predictions. Probabilistic models grounded in Bayesian inference offer a principled framework for characterizing uncertainty, yet they are generally ill-suited for conventional deterministic computing architectures. Here, we propose and experimentally demonstrate a probabilistic photonic computing link based on chaotic Kerr optical frequency combs (CK-PPC). The intrinsic randomness of chaotic Kerr optical frequency combs is exploited as a native entropy source, and a cascade of Mach–Zehnder modulators with a micro-ring resonator (MRR) array is employed to realize the probabilistic photonic computing link. Building upon this platform, a hybrid Bayesian convolutional network (HBNN) is constructed and applied to blood-cell microscopic image classification and out-of-distribution (OOD) detection. Compared with conventional convolutional neural networks, our approach improves OOD detection by 23% while maintaining high recognition accuracy. This scheme paves the way toward integrable, high-speed probabilistic optical processors for trustworthy artificial intelligence in real-world scenarios.

  • Open Access

    Femtojoule optical Kerr switching with milliwatt-peak-power in silicon-organic hybrid nanocavity

    by Yizheng Chen, Xiaoyan Gao, Gaoneng Dong, Wentao Gu, Jianhua Ning, Wentao Ye, Yilun Wang, Wenchan Dong, Lei Lei, Jing Xu, et al.

    All-optical Kerr switches enable ultrafast switching speeds, essential for next-generation communication and computing. However, their practical deployment is hindered by the intrinsically weak optical nonlinearity, which necessitates high switching energies. Although ultrashort pulses with Watt-level peak powers and extremely low duty cycles can alleviate this requirement, they are incompatible with real-world systems that demand high-duty-cycle data. In this work, we overcome this long-standing challenge by experimentally demonstrating an optical Kerr switch compatible with mainstream communication signals, while maintaining femtojoule-level switching energy. The breakthrough is achieved through a combination of exceptionally tight optical field confinement and a high-nonlinearity polymer material in a silicon-organic hybrid slot nanobeam cavity. This design reduces the required peak power to milliwatt-level, approximately two orders of magnitude lower than previous Kerr switches, enabling error-free switching of a 33%-duty-cycle 40-Gbit/s signal with 7.5 mW peak power and 63 fJ/bit switching energy. These achievements pave the way for chip-scale all-optical switches for ultrafast photonic signal processing.


  • Open Access

    Optical logic convolutional neural network

    by Wenkai Zhang, Jingcheng Li, Shiji Zhang, Bo Wu, Yilun Wang, Hailong Zhou, Jianji Dong, Xinliang Zhang

    Optical computing presents a promising avenue to meet the escalating computational demands. However, optical analog computing is susceptible to environmental perturbations, relies heavily on digital-to-analog converters and analog-to-digital converters, and requires electronic or photonic nonlinear operations. While optical digital computing mitigates some issues, its reliance on manual, task-specific configuration hinders broader applications like inference. Here, we propose the concept of an optical logic convolutional neural network (OLCNN). We demonstrate a 1-by-3 optical logic convolutional operator (OLCO) for pattern generation and validate its high-speed computing capacity at 20 Gbit/s. A 2-by-2 OLCO is then implemented to perform three types of image edge extraction. By scaling up, a 3-by-3 OLCO is constructed for an OLCNN to achieve four-class classification on the MNIST dataset with an average test accuracy of 95.1%. By synergizing optical logic devices with neural networks, this work pioneers a logic-driven paradigm for high-speed, energy-efficient optical hardware in artificial intelligence.


  • On-chip input-hidden-layer-degenerate optical diffractive nonlinear neural network

    by Wenguang Xu, Bo Wu, Shiji Zhang, Hailong Zhou, Yilun Wang, Jianji Dong, Xinliang Zhang

    Integrated photonic computing offers compelling advantages, including ultrahigh bandwidth, ultralow latency, and inherently low power consumption, positioning it as a promising platform for artificial intelligence (AI) applications. However, the input scale of current photonic computing chips remains fundamentally constrained by their one-dimensional interface. Here, we propose and experimentally demonstrate an input-hidden-layer-degenerate optical diffractive nonlinear neural network (IHD-ODN 3 ), which expands the input dimension of on-chip optical networks from N to N 2 within a compact footprint. By structurally merging the input layer with linear diffractive hidden layers, input data are directly encoded into optical neurons. Leveraging the intrinsic properties of diffraction, multilayer diffraction interactions among these neurons promote sufficient data mixing, thereby executing an equivalent nonlinear function and mapping the data into a high-dimensional space with a footprint of 0.5mm 2 . We validate the nonlinear computational capability of IHD-ODN 3 through four-class spiral and handwritten digit classification tasks, achieving accuracies of 88% and 98%, respectively. Furthermore, we demonstrate high-performance image compression and reconstruction, with a compression ratio of up to 256:1. This input-hidden-layer-degenerate architecture not only unlocks N 2 dimensional input encoding, but also delivers low-power, highly integrated and programmable optical nonlinear computing with linear photonic hardware. Our approach is fully compatible with mainstream linear optical neural network architectures, paving the way for next generation of large-scale, highly integrated nonlinear optical deep neural networks.


  • Open Access

    Sub-terahertz plasmonic lithium niobate phase modulator

    by Jihao Zhao, Yilun Wang, Xiaoyan Gao, Jian Cheng, Dingshan Gao, Wentao Gu, Wenchan Dong, Qizhi Yan, Liao Chen, Yu Yu, et al.

    Ultrahigh-bandwidth, high-stability, and ultra-compact electro-optic (EO) modulators are the key components for next-generation high-speed optical communications and computing systems. Surface plasmon polaritons (SPPs) have attracted much attention for reducing the device size and accelerating the device response owing to their nanoscale field confinement below the diffraction limit. However, existing plasmonic EO modulators still have challenges in either bandwidth or stability due to the frequency-dependent permittivity or their limited thermal and photochemical stability. These limitations lead to a modulation efficiency drop-off at low frequencies due to the decrease in dielectric constant and obstruct reliable long-term operation. Here, by integrating SPPs with the lithium niobate (LN) platform, we develop an EO phase modulator with ultrahigh bandwidth, high thermal stability, and an ultra-compact footprint. This plasmonic LN phase modulator exhibits a measured ultra-flat bandwidth exceeding 110 GHz with an ultralow half-wave voltage length product of 0.081 V cm and a compact modulation length of 15 μm. Further theoretical analysis validates a calculated 3-dB bandwidth of 184 GHz and indicates a potential 3-dB bandwidth beyond 760 GHz through optimizing the electrode pad size. Moreover, under a thermal shock of 260 °C, the variation of the EO modulation response is below 0.2 dB, evidencing excellent thermal reliability. This study offers a new solution for photonic integrated circuits that demand ultrahigh bandwidth, long-term reliability, and extreme integration density, with potential applications in next-generation high-speed optical communications.


  • Open Access

    Scaling up for end-to-end on-chip photonic neural network inference

    by Bo Wu, Chaoran Huang, Jialong Zhang, Hailong Zhou, Yilun Wang, Jianji Dong, Xinliang Zhang

    Optical neural networks are emerging as a competitive alternative to their electronic counterparts, offering distinct advantages in bandwidth and energy efficiency. Despite these benefits, scaling up on-chip optical neural networks for end-to-end inference is facing significant challenges. First, network depth is constrained by the weak cascadability of optical nonlinear activation functions. Second, the input size is constrained by the scale of the optical matrix. Herein, we propose a scaling up strategy called partially coherent deep optical neural networks (PDONNs). By leveraging an on-chip nonlinear activation function based on opto-electro-opto conversion, PDONN enables network depth expansion with positive net gain. Additionally, convolutional layers achieve rapid dimensionality reduction, thereby allowing for an increase in the accommodated input size. The use of a partially coherent optical source significantly reduces reliance on narrow-linewidth laser diodes and coherent detection. Owing to their broader spectral characteristics and simpler implementation, such sources are more accessible and compatible with scalable integration. Benefiting from these innovations, we designed and fabricated a monolithically integrated optical neural network with the largest input size and the deepest network depth, comprising an input layer with a size of 64, two convolutional layers, and two fully connected layers. We successfully demonstrate end-to-end two-class classification of fashion images and four-class classification of handwritten digits with accuracies of 96% and 94%, respectively, using an in-situ training method. Notably, performance is well maintained with partially coherent illumination. This proposed architecture represents a critical step toward realizing energy-efficient, scalable, and widely accessible optical computing.


  • An Ultra‐Efficient Integrated Plasmonic Lithium Niobate Electro‐Optic Mach‐Zehnder Modulator

    by Jihao Zhao, Yilun Wang, Xiaoyan Gao, Wei Chen, Dingshan Gao, Liao Chen, Chi Zhang, Yu Yu, Jianji Dong, Xinliang Zhang

    Electro‐optic (EO) Mach‐Zehnder modulators (MZMs) featuring high modulation efficiency and wide bandwidth are essential for large‑capacity optical communication systems. To date, thin‐film lithium niobate (TFLN) MZMs have emerged as a promising solution owing to the exceptional EO bandwidth and compactness, and thus significantly reduced capacitance. However, integrated TFLN MZMs are still several millimeters to centimeters in path length due to the limited modulation efficiency determined by the suboptimal confinement of the electric and optical fields. This limitation hinders large‐scale integration for parallelization or multiplexing and prevents the economically efficient co‐integration with compact electronics. Here, this challenge is overcome by merging sub‐wavelength plasmonic slot waveguides with the TFLN platform to form a strong field confinement below the diffraction limit of optics, and enhance EO overlap and light‐matter interactions. A record‐high modulation efficiency of 0.070 V cm is demonstrated with an ultra‐short length of 15 µm. This plasmonic TFLN MZM exhibits a 3‐dB EO bandwidth exceeding 110 GHz that allows for transmitting 110 Gbaud binary phase‐shift keying signals with a bit error ratio of 2.5 × 10 −5 . This demonstrated plasmonic TFLN MZM provides a promising solution to future ultra‐high‐speed and large‐scale photonic integrated systems for optical interconnections, optical computing, and optical sensing functions.


  • Open Access

    2D MoS2-based reconfigurable analog hardware

    by Xinyu Huang, Lei Tong, Langlang Xu, Wenhao Shi, Zhuiri Peng, Zheng Li, Xiangxiang Yu, Wei Li, Yilun Wang, Xinliang Zhang, et al.

    Biological neural circuits demonstrate exceptional adaptability to diverse tasks by dynamically adjusting neural connections to efficiently process information. However, current two-dimension materials-based neuromorphic hardware mainly focuses on specific devices to individually mimic artificial synapse or heterosynapse or soma and encoding the inner neural states to realize corresponding mock object function. Recent advancements suggest that integrating multiple two-dimension material devices to realize brain-like functions including the inter-mutual connecting assembly engineering has become a new research trend. In this work, we demonstrate a two-dimension MoS2-based reconfigurable analog hardware that emulate synaptic, heterosynaptic, and somatic functionalities. The inner-states and inter-connections of all modules co-encode versatile functions such as analog-to-digital/digital-to-analog conversion, and linear/nonlinear computations including integration, vector-matrix multiplication, convolution, to name a few. By assembling the functions to fit with different environment-interactive demanding tasks, this hardware experimentally achieves the reconstruction and image sharpening of medical images for diagnosis as well as circuit-level imitation of attention-switching and visual residual mechanisms for smart perception. This innovative hardware promotes the development of future general-purpose computing machines with high adaptability and flexibility to multiple tasks.


  • High-performance compact vertical germanium photodiodes enabled by silicon corner reflectors

    by Wei Chen, Yang Shi, Yilun Wang, Mingjie Zou, Wen Zhang, Yu Yu, Xinliang Zhang

    High-performance germanium photodiodes are critical components in silicon photonic systems for high-capacity data communications. By reducing the length of the photodiodes, a smaller resistance–capacitance product can be achieved, leading to a larger bandwidth and lower dark current. However, this also leads to diminished responsivity due to insufficient light absorption. Here, we introduce a silicon corner reflector (SCR) to alleviate this issue by reflecting and recycling the unabsorbed light. The process of evanescent coupling between the silicon and germanium layers is elaborately engineered to optimize the efficiency of light absorption. Experimentally, a responsivity of 0.96 A/W, which is a 21% increase compared to the one without SCR, is achieved at 1550 nm with a germanium length of 4.8 μm. Simultaneously, a remarkably low dark current of 0.76 nA and a large bandwidth of 100 GHz are achieved. Open eye diagrams of 140 Gb/s on–off keying and 240 Gb/s four-level pulse amplitude signals are obtained. To the best of our knowledge, this work achieves the lowest dark current density and noise equivalent power to date and offers a promising solution for low-cost, high-performance optical detection.


  • Open Access

    Programmable nonlinear optical neuromorphic computing with bare 2D material MoS2

    by Lei Tong, Yali Bi, Yilun Wang, Kai Peng, Xinyu Huang, Wei Ju, Zhuiri Peng, Zheng Li, Langlang Xu, Runfeng Lin, et al.

    Nonlinear optical responses in two-dimensional (2D) materials can build free-space optical neuromorphic computing systems. Ensuring the high performance and the tunability of the system is essential to encode diverse functions. However, common strategies, including the integration of external electrode arrays or photonic structures with 2D materials, and barely patterned 2D materials, exhibit a contradiction between performance and tunability. Because the unique band dispersions of 2D materials can provide hidden paths to boost nonlinear responses independently, here we introduced a new free-space optical computing concept within a bare molybdenum disulfide array. This system can preserve high modulation performance with fast speed, low energy consumption, and high signal-to-noise ratio. Due to the freedom from the restrictions of fixed photonic structures, the tunability is also enhanced through the synergistic encodings of the 2D cells and the excitation pulses. The computing mechanism of transition from two-photon absorption to synergistic excited states absorption intrinsically improved the modulation capability of nonlinear optical responses, revealed from the relative transmittance modulated by a pump-probe-control strategy. Optical artificial neural network (ANN) and digital processing were demonstrated, revealing the feasibility of the free-space optical computing based on bare 2D materials toward neuromorphic applications.