We integrate GO films with silicon-on-insulator nanowires (SOI), high index doped silica glass (Hydex) and silicon nitride (SiN) waveguides and ring resonators, to demonstrate an enhanced optical nonlinearity including Kerr nonlinearity and four-wave mixing (FWM).

As a novel layered noble metal dichalcogenide material, palladium diselenide (PdSe2) has attracted wide interest due to its excellent optical and electronic properties. In this work, a strong third-order nonlinear optical response of 2D PdSe2 films is reported. We conduct both open-aperture (OA) and closed-aperture (CA) Z-scan measurements with a femtosecond pulsed laser at 800 nm to investigate the nonlinear absorption and nonlinear refraction, respectively. In the OA experiment, we observe optical limiting behaviour originating from large two photo absorption (TPA) in the PdSe2 film of β = 3.26 ×10-8 m/W. In the CA experiment, we measure a peak-valley response corresponding to a large and negative Kerr nonlinearity of n2 = -1.33×10-15 m2/W – two orders of magnitude larger than bulk silicon. In addition, the variation of n2 as a function of laser intensity is also characterized, with n2 decreasing in magnitude when increasing incident laser intensity, becoming saturated at n2 = -9.96×10-16 m2/W at high intensities. Our results show that the extraordinary third-order nonlinear optical properties of PdSe2 have strong potential for high-performance nonlinear photonic devices.

Two-dimensional layered graphene oxide (GO) films are integrated with silicon-on-insulator nanowires to experimentally demonstrate enhanced self-phase modulation, achieving high broadening factor of up to 4.14 for a device patterned with 0.4-mm-long, 10 layers of GO.

Two-dimensionallayered graphene oxide films are integrated with micro-ring resonators to experimentally demonstrate enhanced four-wave mixing, achieving up to ~7.6-dB enhancement in conversion efficiency for a uniformly coated device and ~10.3-dB for a patterned device.

We experimentally demonstrate integrated waveguide and micro-ring resonator polarizers incorporating two-dimensional layered graphene oxide films, with a high polarization dependent loss of 53.8 dB and a high polarization extinction ratio of 8.3 dB, respectively.

We theoretically investigate integrated photonic resonators formed by two mutually coupled Sagnac loop reflectors (MC-SLRs). Mode interference in the MC-SLR resonators is tailored to achieve versatile filter shapes with high performance, which enable flexible spectral engineering for diverse applications. By adjusting the reflectivity of the Sagnac loop reflectors (SLRs) as well as the coupling strength between different SLRs, we achieve optical analogues of Fano resonance with ultrahigh spectral slope rates, wavelength interleaving / non-blocking switching functions with significantly enhanced filtering flatness, and compact bandpass filters with improved roll-off. In our designs the requirements for practical applications are considered, together with detailed analyses of the impact of structural parameters and fabrication tolerances. These results highlight the strong potential of MC-SLR resonators as advanced multi-functional integrated photonic filters for flexible spectral engineering in optical communications systems.

We characterize the third-order optical nonlinearity in PdSe2 dichalcogenide films via the Z-Scan technique. A strong and negative (self-defocusing) Kerr nonlinearity (n2) of ∼ ‒ 7.65×10-16 m2/W is observed at 800 nm.

We investigate Fano resonances in integrated photonic resonators formed by coupled Sagnac loop reflectors (CSLR). Mode splitting in the CSLR resonator is tailored to achieve multiple Fano-like resonances with ultrahigh extinction ratios and slope rates.

Enhanced four-wave mixing in silicon nitride waveguides integrated with 2D graphene oxide (GO) films is experimentally demonstrated. We achieve a high conversion efficiency improvement of ~7.3 dB for a 2-cm-long waveguide with monolayer GO film.

We demonstrate broadband RF channelizers based on microcombs with 49 GHz and 200 GHz FSRs. Up to 92 parallel channels and an instantaneous bandwidth of up to 8.08 GHz are achieved, together with high-resolution RF spectral channelization and shaping. This approach is promising for high-performance integrated photonic RF receivers and processors.

We theoretically investigate advanced multi-functional integrated photonic filters formed by three waveguide coupled Sagnac loop reflectors (3WC-SLRs). By tailoring the coherent mode interference, the spectral response of the 3WC-SLR resonators is engineered to achieve diverse filtering functions with high performance. These include optical analogues of Fano resonances that yield ultrahigh spectral extinction ratios (ERs) and slope rates, resonance mode splitting with high ERs and low free spectral ranges, and classical Butterworth, Bessel, Chebyshev, and elliptic filters. A detailed analysis of the impact of the structural parameters and fabrication tolerances is provided to facilitate device design and optimization. The requirements for practical applications are also considered. These results theoretically verify the effectiveness of using 3WC-SLR resonators as multi-functional integrated photonic filters for flexible spectral engineering in diverse applications.

Convolutional neural networks (CNNs), inspired by biological visual cortex systems, are a powerful category of artificial neural networks that can extract the hierarchical features of raw data to greatly reduce the network parametric complexity and enhance the predicting accuracy. They are of significant interest for machine learning tasks such as computer vision, speech recognition, playing board games and medical diagnosis [1-7]. Optical neural networks offer the promise of dramatically accelerating computing speed to overcome the inherent bandwidth bottleneck of electronics. Here, we demonstrate a universal optical vector convolutional accelerator operating beyond 10 Tera-FLOPS (floating point operations per second), generating convolutions of images of 250,000 pixels with 8-bit resolution for 10 kernels simultaneously — enough for facial image recognition. We then use the same hardware to sequentially form a deep optical CNN with ten output neurons, achieving successful recognition of full 10 digits with 900 pixel handwritten digit images with 88% accuracy. Our results are based on simultaneously interleaving temporal, wavelength and spatial dimensions enabled by an integrated microcomb source. This approach is scalable and trainable to much more complex networks for demanding applications such as unmanned vehicle and real-time video recognition.