Neuromorphic photonic computing has emerged as a competitive computing paradigm to overcome the bottlenecks of the von-Neumann architecture. Linear weighting and nonlinear spike activation are two fundamental functions of a photonic spiking neural network (PSNN). However, they are separately implemented with different photonic materials and devices, hindering the large-scale integration of PSNN. Here, we propose, fabricate and experimentally demonstrate a photonic neuro-synaptic chip enabling the simultaneous implementation of linear weighting and nonlinear spike activation based on a distributed feedback (DFB) laser with a saturable absorber (DFB-SA). A prototypical system is experimentally constructed to demonstrate the parallel weighted function and nonlinear spike activation. Furthermore, a four-channel DFB-SA laser array is fabricated for realizing matrix convolution of a spiking convolutional neural network, achieving a recognition accuracy of 87% for the MNIST dataset. The fabricated neuro-synaptic chip offers a fundamental building block to construct the large-scale integrated PSNN chip.

We propose an on-chip mode converter via two cascaded Bragg reflection processes. A forward conversion between two guided modes can be achieved with the aid of an additional mode. The proposed structure is theoretically studied and simulated via the rigorous three-dimensional finite-difference time-domain (3D-FDTD) method. The bandwidth and central wavelength of the proposed mode converter can be adjusted according to our theoretical analysis and simulation results. By applying the similar design approaches as fiber Bragg gratings, conversion spectra with different shapes can be obtained. As an example, several mode converters with bandpass and sidelobe-reduced spectra are designed. We also investigate and verify the mode conversion by experiment. Therefore, the proposed method may pave a new path for the mode converters with desired conversion spectra.

Hybrid free space optical (FSO)/radio frequency (RF) systems have been proved to be reliable links for high-data-rate wireless backhauls. In this paper, we present a unified performance analysis of the hybrid FSO/RF transmission system which transmits the identical data in both links and implements two popular diversity combining schemes, namely, selection combining (SC) and maximal ratio combining (MRC), in the receiver. Specially, for the FSO link, the Gamma-Gamma turbulence with pointing errors under heterodyne detection (HD) and intensity modulation/direction detection (IM/DD) is considered in our analysis while the general - shadowed fading which unifies popular RF fading models is employed for the analysis of the RF link. As a result, unified closed-form expressions of outage probabilities and average bit error rates for different modulation schemes are derived. Analytical and Monte Carlo simulation results are provided to characterize the performance of the hybrid FSO/RF link which is compared to the single FSO link and the single RF link. The agreement between the analytical and simulation results confirms the unification of various FSO channels and RF fading scenarios into a single closed-form expression.

Multi-wavelength distributed feedback laser array based tunable semiconductor lasers have been widely used owing to their high mode stability and simple tuning scheme. Comparing with the in-parallel laser array, the in-series one can avoid a large power loss caused by the optical combiner. However, the single-mode stability of an in-series laser array is vulnerable to the reflections of other lasers. Here we experimentally demonstrated a tunable in-series laser array with high single-mode stability and high uniformity of wavelength spacing. To reduce the impact of such reflections, the Bragg gratings in different lasers were designed in phase, and the grating phase error was reduced by utilizing the reconstruction-equivalent-chirp technique. Besides, a three-section laser structure was applied in each laser to increase the priority of the dominant lasing mode. Four lasers with small wavelength-spacings of 2.5 nm were integrated in the in-series laser array. According to the measurement results of 40 in-series laser arrays (160 lasers), wavelength deviations of 90.6% lasers were within ±0.2 nm, average wavelength spacings of 97.5% of the measured laser arrays were deviated from the design within ±0.2 nm, and SMSRs of 96.3% lasers were above 45 dB. The output power was above 25 mW and the relative intensity noise was below -130 dB/Hz. Like the in-parallel laser array, the proposed laser array can be applied in either continuous wavelength tuning or fast channel switching. For the continuous wavelength tuning applications, the 10-nm tuning range was obtained with a temperature tuning range of less than 20 °C. For the fast channel switching applications, 4 channels with 2.5-nm spacing are available, and the switching time between two channels was less than 300 ns.

We proposed a new type of TE-polarized mode-order converter based on a deeply-etched polygonal slot on a silicon-on-insulator (SOI) waveguide. Along the transverse direction of the waveguide, two irregular boundary surfaces (BSs) of the slot can introduce high-contrast index modulation on guided modes, leading to multimode interference in the slot. Therefore, when the slot is optimized, we can achieve efficient mode conversions based on the multimode interference. As examples, mode converters from the fundamental transverse electric mode $\rm (TE_0)$ to the first-order transverse electric mode $\rm (TE_1)$ and to the second-order transverse electric mode $\rm (TE_2)$ have both been demonstrated with a short device length ( 97.6%), and a low modal crosstalk (< -20.0 dB) over a broad wavelength range from 1500 to 1600 nm (∼ 100 nm). In addition, based on different polygonal slots, other types of mode conversions such as from $\rm TE_1$ to $\rm TE_3$ and from $\rm TE_2$ to $\rm TE_1$ have be realized. Fabrication error tolerance of the proposed structure is analyzed. Owing to the high efficiency and compact size, the proposed structure could be applied to the on-chip mode division multiplexing (MDM) systems for high-density integration.

A four-channel wavelength-selectable in-series DFB laser array with good single-longitudinal-mode (SLM) property is proposed and theoretically analyzed. To realize good SLM property, a linearly chirped Bragg grating with multiple phase shifts are implemented in the laser cavity. In addition, two reflector sections with chirped Bragg grating are specially designed for reflection compensation at both ends. In our simulation, the split-step time-domain model with coupled-wave equations and carrier rate equations are used. Based on the simulation results, all the channels work with low threshold currents and the SLM properties are considerably enhanced compared with the conventional in-series laser array. Notably a high precision control of grating phase is required in the practical fabrication due to the ultra-small variation of grating period. Therefore, the reconstruction-equivalent-chirp technique is discussed to provide the fabrication of high-precision grating. In addition, the complex heterogeneous integration or butt-joint technique is not required during the whole fabrication process, which simplifies the fabrication and reduces the cost. The proposed wavelength-selectable in-series laser array is deemed to be applied in the dense wavelength division multiplexing technology, especially when fast wavelength switching is needed, such as in the application of NGPON2.