Optical fiber sensing plays a crucial role in modern measurement systems and holds significant promise for a wide range of applications. This potential, though, has been fundamentally constrained by the intrinsic latency and power limitations associated with electronic signal processing. Here, we propose an all-optical fiber sensing architecture with in-sensor computing (AOFS-IC) that achieves fully optical-domain sensing signal demodulation at the speed of light. By integrating a scattering medium with an optimized diffractive optical network, AOFS-IC enables linear mapping of physical perturbations to detected intensity, and sensing results can be directly read out without electronic processing. The proposed system maintains high accuracy across various sensing tasks, providing sub-nano strain resolution and 100% torsional angle classification accuracy, as well as multiplexed sensing of multiple physical quantities, and performing multi-degree-of-freedom robot arm monitoring. AOFS-IC eliminates computing hardware requirements while providing <3 ns demodulation delay, which is more than 2 orders of magnitude faster than conventional fiber optic sensing systems. This work demonstrates the potential of next-generation optical sensing systems empowered by all-optical computing and paves the way for expanded applications of fiber sensing through the integration of fully optical components, ultrafast measurement speed, and low power consumption.

A power splitter with a wideband arbitrary splitting ratio, which provides flexibility and adaptability in forming photonic devices such as microring resonators and Mach-Zehnder interferometers, proves to be essential in photonic integrated circuits (PICs). We designed and fabricated a directional coupler-based power splitter with a wideband arbitrary splitting ratio and a microring resonator with a wideband uniform extinction ratio (ER) based on artificial gauge field (AGF) optimization. The neural network-aided inverse design method is applied to complete the target. Less than 0.9 dB power splitting variation and 1.6 dB ER variation have been achieved experimentally over a 100-nm bandwidth. Wideband performance, design efficiency, and device compactness are obtained by utilizing this optimization, which indicates great potential and universality in PIC applications.

This paper presents the development of an 8-inch wafer-level TGV interposer, highlighting its potential for CPO applications. Fabricated using Laser-Induced Deep Etching (LIDE) and metallization, the TGV interposer demonstrated a bandwidth of 110 GHz and supported a 128-Gbaud On- Off Keying (OOK) signal interconnection, verified through both simulations and experimental methods. The research underscores the feasibility of 2.5D packaging by utilizing flip-chip technology to integrate driver and Electro- Absorption Modulated Laser (EML) chips onto the TGV interposer. The op-to electronic joint testing confirmed the EML's operational viability, achieving an optical output power of -0.07 dBm. This study illustrates advancements in TGV interposer technologies for high-performance optical communication systems.

A four-channel mode division multiplexing (MDM) transmitter is demonstrated with $4 \times 80$ Gbps signaling, assisted with a low-loss two-dimensional grating coupler over $1.85 ~\text{mm}^{2}$ footprint, exhibiting significant potential for future high-density and large-capacity optical interconnects.

All-optical signal processing has a wide range of applications in optical computing, communication, switch and so on, where intensity modulation is highly desired. In this work, we propose and demonstrate a study of all-optical intensity modulation via nonlinear tuning based on two photo absorption (TPA) between C and 2-mu m wavebands. By using multimode silicon waveguide with Euler bends for low-loss optimization, we design and fabricate a 12.3-cm silicon waveguide, achieving a minimum propagation loss of 0.23 dB/cm at C band and 0.87 dB/cm at 2-mu m waveband. The waveguide is fabricated over 220-nm silicon-on-insulator (SOI) platform. 8.9 dB extinction ratio (ER) of TPA is experimentally realized above a 400-nm spectral distance at 2-mu m waveband with a C-band pump. A mode loop structure assisted by mode-division-multiplexing (MDM) is proposed to enhance the TPA effect, with 1.6-dB extra ER improvement obtained.

We experimentally demonstrate third-order distributed Raman cascaded with lumped Raman, achieving amplification over a 102-nm bandwidth, covering the full $C$ and $L$ bands, with 30.4-dB average gain and the lowest $-7.6 d B$ effective noise figure.

Growing research interests have been directed towards the 2-μm waveband, which is considered to be another promising window for fiber and free-space optical communications. However, the signaling beyond 100 Gbit/s was difficult at this waveband due to the limited bandwidth of the available electro-optic modulators. In this study, we demonstrated a 35-GHz silicon photonic Mach-Zehnder Modulator (MZM) fabricated by a standard multi-project wafer process. Up to 100 Gb/s on-chip modulation was experimentally achieved at the wavelength of 1976 nm, which was eventually limited by the bandwidth of the receiver. The modulation efficiency is measured to be 0.82 V·cm. This is a benchmark modulation speed among existing silicon modulators around a 2 µm spectral range.

Optical switches are crucial components in integrated optics, with broad applications in data communications, sensing, and computing. We experimentally demonstrate an ultra-efficient silicon thermo-optic Mach-Zehnder switch based on waveguide superlattices with artificial gauge field. A sinusoidal modulation profile is typically applied to a binary waveguide array to achieve low-crosstalk and broadband light transmission. This design achieves the lowest power consumption of 1.21 mW without extra thermal isolation. Geometric design optimization, central to this demonstration, can be utilized to develop compact and high-efficiency devices for large-scale photonic integrated circuits. This approach paves the way for significantly reducing power consumption while maintaining scalability.

Ultra-low-noise amplification and modal crosstalk mitigation are critical for advancing mode division multiplexing (MDM) systems. We propose and experimentally demonstrate a few-mode hybrid optical amplifier that combines a few-mode distributed Raman amplifier (FM-DRA) with a single-mode phase-sensitive amplifier (PSA). The FM-DRA implemented in a 100-km few-mode fiber provides preliminary low-noise amplification, while the subsequent PSA leverages its unique phase-sensitive gain property to further enhance signal gain, improve optical signal-to-noise ratio, and suppress modal crosstalk. The system achieves average on-off gains of 20.3 dB and 14.8 dB across a 35-nm bandwidth (1530 to 1565 nm) for the LP01 and LP02 modes, respectively. Remarkably, the effective noise figures (NFs) for the LP01 and LP02 modes are below -2.8 dB and -3.7 dB, respectively, with the lowest effective NFs reaching -5.9 dB and -6.8 dB, outperforming the standalone FM-DRA. Additionally, the PSA demonstrates significant modal crosstalk reduction for each mode across the 35-nm bandwidth. These results demonstrate the potential of the proposed scheme to significantly enhance the performance of MDM systems.

Transmission with few-mode fiber (FMF) is one of the most promising methods for increasing single fiber capacity, with urgent demands for low-noise and wideband amplification. In this work, we experimentally demonstrate a few-mode hybrid Raman amplifier (FM-HRA) based on a second-order few-mode distributed Raman amplifier (FM-DRA) cascaded with a lumped Raman amplifier (LRA). The few-mode hybrid amplification with a 100-nm bandwidth (1520 to 1620 nm), covering the S + C + L bands, is realized using a weakly-coupled FMF with a transmission length of 100 km and a dispersion compensating fiber (DCF). The maximum on-off gains for three modes ( LP 01 , LP11, and LP21) are 22.6, 25.1, and 25.5 dB, and the average on-off gains over 100-nm bandwidth are 19.6, 21.4, and 21.3 dB, respectively. The differential modal gain (DMG) is less than 2.9 dB. The effective noise figure (NF) of the three modes is lower than - 1.1 dB over 100-nm bandwidth, and the lowest effective NF is - 6.8 dB, which is close to the effective NF of FM-DRA. This means HRA can fully utilize the advantages of ultralow noise of DRA and high gain of LRA. The MIMO-less transmission experiment demonstrates the highest data rate of 56 Gbps for LP01 mode. This work features pure Raman amplification with high gain, ultralow noise, and wideband. The configuration of high-order DRA in FMF and first-order LRA in DCF also offers the benefits of better pump power utilization and dispersion compensation.

Dense waveguides are the basic building blocks for photonic integrated circuits(PICs).Due to the rapidly increasing scale of PIC chips,high-density integration of waveguide arrays working with low crosstalk over broadband wavelength range is highly desired.However,the subwavelength regime of such structures has not been adequately explored in practice.We propose a waveguide superlattice design leveraging the artificial gauge field mechanism,corresponding to the quantum analog of field-induced n-"photon"resonances in semiconductor superlattices.This approach experimentally achieves-24 dB crosstalk suppression with an ultrabroad transmission bandwidth more than 500 nm for dual polarizations on the Si3N4 platform.The fabricated waveguide superlattices support high-speed signal transmission of 112 Gbit/s with high-fidelity signal-to-noise ratio profiles and bit error rates.This design,featuring a silica upper cladding,is compatible with standard metal back-end-of-the-line processes.Based on such a fundamental structure,which is readily transferable to other platforms,passive and active devices over versatile platforms can be realized with a significantly shrunk on-chip footprint,thus it holds great promise for significant reduction of the power consumption and cost in PICs.

In this paper, a four-channel silicon photonic mode division multiplexing (MDM) transmitter chiplet is proposed over a 1.85-mm2 footprint, utilizing micro-ring modulators and a two-dimensional grating coupler (2D GC) with side-distributed Bragg reflectors. Through genetic optimization approach, the optimized 2D GC can simultaneously couple the two orthogonal polarizations of the on-chip TE0 and TE1 modes to the four LP modes in the FMF, all exhibiting high coupling efficiency beyond -3.8 dB, so as to support the MDM optical fiber interface. High data rate up to 4 × 80-Gbps signaling is experimentally demonstrated with a bit error rate below the 7% hard-decision forward-error-correction threshold. The proposed transmitter chiplet features single fiber and single-wavelength applications of quad small form-factor pluggable (QSFP) transceivers, pave the way toward future high-density and large-capacity optical interconnects scenarios.

Barium titanate (BaTiO3, BTO) thin films, with their exceptionally high Pockels coefficients, present a promising alternative to lithium niobate (LiNbO3, LN) for integrated photonic devices. BTO's compatibility with complementary metal–oxide–semiconductor (CMOS) technology further enhances its appeal, contingent on the development of low-temperature growth processes. This study investigates the impact of growth temperature on the electro-optic (EO) performance of BTO films, revealing a clear correlation between lower growth temperatures and reduced EO coefficients. Notably, BTO films grown at 400 °C maintain significant EO coefficients of approximately 51.6 pm/V. These findings underscore the potential of low-temperature grown BTO films for high-performance EO applications. By elucidating the relationship between growth temperature, crystallinity, and EO performance, this research provides critical guidelines for fabricating high-performing BTO films compatible with CMOS technology, facilitating the advancement of next-generation photonic devices.

In this work, a reflective metasurface SDM connector is demonstrated with an insertion loss of 7 dB (experiment) and 2.2 dB (theory) over 80-nm bandwidth, featuring MCF FIFO applications with massive production potential.

A triple-waveguide-coupling based reconfigurable optical power splitter is demonstrated for 1-to-15 channel-scalable optical interconnects leveraging only four switching elements, exhibiting no power wasting, low crosstalk and indicating significant potential for large-scale capacity reconfiguration scenarios.

We propose an integrated photonic three-dimensional tensor convolution accelerator by time-wavelength interleaved frequency synthesis technology, achieving the extraction of three-dimensional data edge information with 1 K resolution @ 15 FPS and a PSNR of $24 d B$.

Recently, the 2 mu m wavelength band has attracted increasing interest because of its potential prospects for free-space and fiber optical communications. Silicon photonic integrated circuits can operate in this band, but a major challenge exists for integrated photodetectors because the photon energies in this wavelength band are less than the energy bandgap of germanium which is the material used for in silicon photonic photodetectors for the conventional optical communications bands. While it is possible to have weak optical absorption at the 2-mu m wavelength band using strained germanium and defect states, the need for a long absorption length can degrade the resistance-capacitance (RC) limited bandwidth. To address this problem, we proposed a compact 3D waveguide loop to improve both the responsivity and the bandwidth. A closed loop for light-matter interaction enhancement is demonstrated, enabling the responsivity of the detector to reach 0.11 A W-1 at wavelengths of 1955 nm. The measured 3 dB bandwidth of the photodetector is 36 GHz and is sufficient for high-speed detection of 112 Gb s-1 PAM-4 signals. The device is fabricated in a commercial foundry and extends the high-speed communications capabilities of silicon photonics to the 2 mu m wavelength band.

Broadband optical amplifiers are essential for fully utilizing the low-loss transmission window of single-mode fiber (SMF), thereby enabling ultra-wideband and high-capacity optical transmission. In this work, we experimentally demonstrate a third-order hybrid Raman amplifier (HRA) that consists of a third-order distributed Raman amplifier (DRA) cascaded with a lumped Raman amplifier (LRA). The proposed HRA is implemented over a 101.5-km SMF EX2000 and a 15.7-km dispersion-compensating fiber, achieving a 102-nm (1524-1626 nm) amplification bandwidth that covers the full C and L bands. The HRA provides a maximum on-off gain of 33.7 dB and an average on-off gain of 30.4 dB. The average effective noise figure (NF) is-5.3 dB, and the lowest effective NF is-7.6 dB, indicating that the HRA takes full advantage of the ultralow effective NF of the DRA and the high gain provided by the LRA. Using pure Raman amplification, we demonstrate intensity-modulation direct-detection (IM-DD) transmission over 18 wavelength-division multiplexing (WDM) channels at two data rates. The 56 Gbps/lambda PAM-4 signals meet the 7 % overhead hard-decision forward error correction (HD-FEC) threshold, while the 112 Gbps/lambda PAM-4 signals satisfy the 20 % overhead soft-decision forward error correction (SD-FEC) threshold.

We propose a 4D-photonic convolution tensor core leveraged by the micro-ring mesh that enabling parallel batch processing and parallel multi-kernel computation. The equivalent computing power is up to 2.24 Tera-operations per second (TOPs) per second with over 90% recognition accuracy.