David Moss

The rise of edge-centric 6G networks demands real-time intelligent processing of heterogeneous radio-frequency (RF) signals. Complex-valued blind source separation (CV-BSS) provides a physical-layer capability, enabling the blind extraction of source components from overlapping RF signals. Here, we demonstrate an integrated photonic vector processor (IPVP) that implements real-time CV-BSS via in-physics computation directly on raw RF signals. The IPVP operates across multiple RF bands up to 24 GHz and achieves blind, in-hardware adaptation to unknown sources and channels via in situ learning. Experiments demonstrate the efficacy of IPVP in multiple-input multiple-output (MIMO) communications and multi-band radar interference suppression. The IPVP reduces the EVMs of the two recovered signals from over 50% to 6.18% and 5.3%, respectively, while enhancing the SIR by up to 26 dB. This work demonstrates, to our knowledge, the first photonic processor capable of real-time CV-BSS at RF, highlighting the potential of integrated photonic hardware for ultralow-latency, edge-intelligent signal processing.

Microwave photonic (MWP) all-pass filters (APFs), which enable phase manipulation without altering signal amplitudes, are essential for phase engineering in high-bandwidth microwave signal processing. However, achieving a high degree of reconfigurability in the filter order and characteristic parameters of APFs remains challenging for conventional methods. Here, we propose and experimentally demonstrate MWP APFs based on a microcomb-driven finite-impulse-response (FIR) transversal filter system, enabling high reconfigurability and sharp phase transitions with high group delays. We experimentally demonstrate both first- and second-order APFs, achieving maximum amplitude variations (MAVs) down to ∼1.95 dB and maximum group delays (MGDs) up to ∼580 ps for the former, and MAVs down to ∼2.01 dB and MGDs up to ∼1740 ps for the latter. In addition, by simply programming the tap coefficients without changing any hardware, we experimentally demonstrate a high degree of filter reconfigurability with respect to filter order, group delay, center frequency, and independent control of multiple APF parameters. These results highlight the potential of our approach for implementing MWP APFs with both high performance and reconfigurability for practical applications requiring dynamic and versatile phase control.

As a rapidly expanding family of 2D materials, MXenes have attracted considerable interest for photonic applications, and their oxide derivatives offer an additional degree of freedom for tailoring their optical responses. Here, we systematically investigate how controlled oxidation regulates the nonlinear optical absorption (NOA) of Ti₃C₂T_x MXene. The MXene films are prepared using a transfer-free, layer-by-layer self-assembly method, followed by conversion into MXene oxide (MXO) and TiO₂ through a two-step oxidation process. The NOA behaviors of the MXene and corresponding MXO and TiO₂ derivatives are characterized by Z-scan measurements at ~515, ~800, and ~1030 nm, showing distinct wavelength-dependent responses. All three materials exhibit reverse saturable absorption (RSA) at ~800 nm, and progressive oxidation induces saturable absorption (SA)-to-RSA transitions at ~515 and ~1030 nm. This transition is particularly notable at ~1030 nm, where the nonlinear absorption coefficient β changes from ~−9.05 × 10⁻¹⁰ m/W for MXene to ~2.07 × 10⁻⁹ m/W for TiO₂. Intensity-dependent measurements further reveal distinct variations in β among the three materials. These results establish controlled oxidation as an effective strategy for engineering the NOA of MXene-derived materials and provide insight into the evolution of their nonlinear optical properties, offering a versatile material platform for potential nonlinear optical applications such as optical modulation and optical limiting.

Dilated convolution is pivotal for dense prediction tasks such as semantic segmentation, as it expands the receptive field to capture multi‐scale contextual information without increasing parameter complexity. However, its electronic implementation often faces bottlenecks due to non‐contiguous memory access and high computational load. Here, we report a Reconfigurable Optical Dilated Convolution Accelerator (RODCA) based on a wavelength‐dispersion co‐tuning mechanism. By dynamically controlling optical frequency comb spacing and dispersion, RODCA enables reconfigurable dilation rates while improving spectral efficiency. The system achieves an equivalent computing throughput of 13.858 Tera operations per second (TOPS). We experimentally validate the RODCA using Laplacian kernels and semantic segmentation tasks, where it achieves a pixel accuracy of 88.46% and a mean Intersection over Union (mIoU) of 76.08%, closely matching in‐silico results. This work demonstrates RODCA's capability for efficient multi‐scale feature extraction, highlighting its potential for next‐generation photonic computing in complex AI applications.

Millimetre-wave baseband combs are relevant to wireless links, sensing, and device characterisation. Direct broadband generation from an optical microcomb has remained elusive because of challenges in photoconductive conversion. Here, we demonstrate a 50 GHz repetition-rate millimetre-wave baseband comb covering the sub-THz region, generated by photoconductive conversion of a laser-cavity-soliton microcomb. The microresonator-filtered source combines low free-running phase noise and high spectral efficiency, enabling coherent conversion without additional optical amplification to single-cycle, carrier-envelope-offset-free waveforms with multiple harmonics directly observed in the converted output. The source supports free-running time-domain spectroscopy over optical path delays above 8 m, and its lines are driven by the optical repetition-rate harmonics with traceable improvement of synchronisation performance upon external locking. In addition, multisoliton operation reshapes the harmonic envelope while preserving the common line spacing, providing a simple route to source-level spectral control in the sub-THz region.

Ordinary differential equations (ODEs) are essential for modeling and governing physical phenomena and engineering systems across a wide range of scientific and engineering disciplines. Conventional photonic ODE solving systems exhibit a limited coefficient tunability, where the ODE coefficients are constrained by the device architecture. Here, an order- and coefficient-tunable microwave photonic (MWP) ODE solver based on an integrated microcomb source is demonstrated. A transversal filter structure is employed to directly synthesize the desired transfer function through convolution operations, enabling independently tuned ODE coefficients. We experimentally demonstrated simplified first-order, general first-order, and second-order MWP ODE solvers with different coefficients. For the input Gaussian pulse with a pulse width of ~0.1 ns, the measured output waveforms of the ODE solvers agree well with the calculated results, confirming the effectiveness of our approach. A highly reconfigurable MWP ODE solver with high processing accuracy has been achieved by using our approach, which offers a solution for applications in modern control systems, thermal diffusion, and biochemical reactions modelling.

Polarization control is essential for modern optical systems. Recently, two-dimensional (2D) materials featuring highly anisotropic absorption across broad wavelength regions have emerged as a promising solution for realizing optical polarizers with operation bandwidths far exceeding those achievable using bulk materials. Despite this, existing 2D-material-based optical polarizers face critical limitations in simultaneously delivering high polarizer figures of merit (FOMs), robust power endurance, and scalable fabrication, hindering their practical deployment beyond laboratory. Here, we overcome this bottleneck by demonstrating optical polarizers through integrating 2D MXene films onto silicon photonic devices and subsequently converting them in-situ into 2D titanium dioxide (TiO₂). The 2D TiO₂ exhibits significantly improved anisotropic absorption and thermal / chemical stability compared with MXene, enabling the hybrid polarizers to achieve a record high polarizer FOM of ~17.5 (among 2D-material-coated devices) and excellent power tolerance (stable under >2 W coupled average power), together with a broad operation bandwidth (>140 nm) and a high fabrication yield (~98%). We also demonstrate microring resonator polarizers achieving high polarization extinction ratios up to ~12 dB. With outstanding overall performance and strong capability for large-scale manufacturing, these polarizers represent a solid step towards industry-ready implementations of 2D-material-based optical polarizers.

Integrated nested microring resonators (NMRRs) with multiple split resonances are theoretically investigated for advanced spectral engineering. First, the spectral response of NMRRs of different orders is systematically analyzed, revealing the evolution of extinction ratio (ER), quality (Q) factor, and mode-splitting degree with device structural parameters. Next, the split resonances of NMRRs are engineered to realize different filter functions with outstanding performance, including dense wavelength division multiplexing (DWDM) and Fano filtering. The former enables reduced channel spacing by a factor of ~18 without increasing the device footprint, whereas the latter achieves high slope rates (SRs) of up to ~1337 and ~1616 dB/nm while maintaining low insertion losses (ILs) of 3 and 5 dB, respectively. Finally, fabrication tolerances and practical limitations are analyzed to guide the design and optimization of NMRRs, together with discussion on their potential for other applications. These results highlight the strong potential of NMRRs as highly versatile integrated photonic filters with flexible spectral engineering capabilities.

Microwave photonic (MWP) all-pass filters (APFs), which enable phase manipulation without altering signal amplitudes, are essential for phase engineering in high-bandwidth microwave signal processing. However, achieving a high degree of reconfigurability in the filter order and characteristic parameters of APFs remains challenging for conventional methods. Here, we propose and experimentally demonstrate a new strategy for realizing MWP APFs based on a microcomb-driven transversal filter system, enabling high reconfigurability, sharp phase transitions with high group delays, and low loss. We experimentally demonstrate both 1st- and 2nd-order APFs, achieving a maximum amplitude variation (MAV) of ~1.95 dB and a maximum group delay (MGD) of ~580 ps for the former, and a MAV of ~2.01 dB and an MGD of ~1740 ps for the latter. In addition, both independent and joint tuning of APF characteristic parameters are demonstrated by simply programming the tap coefficients without changing any hardware, enabling highly reconfigurable filter response for diverse phase-engineering functionalities. These results highlight the potential of our approach for implementing MWP APFs with both high performance and reconfigurability for practical applications requiring dynamic and versatile phase control.