Deep neural networks (DNNs) have shown their superiority in a variety of complicated machine learning tasks. However, large-scale DNNs are computation- and memory-intensive, and significant efforts have been made to improve the efficiency of DNNs through the use of better hardware accelerators as well as software training algorithms. The optical neural network (ONN) is a promising candidate as a next-generation neurocomputing platform due to its high parallelism, low latency, and low energy consumption. Here, we devise a hardware-efficient optical neural network architecture named optical subspace neural network (OSNN), which targets lower optical component usage, area cost, and energy consumption of previous ONN architectures with comparable task performance. Additionally, a hardware-aware training framework is provided to minimize the required control precision, lessen the chip area, and boost the noise robustness.

A complete chip-based quantum communication system is demonstrated. The core functions of quantum transmitter, quantum receiver and quantum random number generator are implemented onto photonic integrated circuits (PICs) of different materials. For the first time, these PICs are all interfaced in a compact optoelectronic assembly where they operate synchronously to distribute information theoretically secure encryption keys in real-time. After reviewing the challenges of system integration for quantum photonic circuits, we present our development of plug-and-play quantum communication modules that are practical, scalable, power efficient and perform with high stability under real-life conditions.

High-speed silicon micro-ring modulators (MRMs) have recently emerged as an attractive alternative to traditional modulators due to their compact size and low power consumption. We demonstrate an MRM is a promising candidate for the next generation short-reach optical networks beyond 100 Gbaud per channel optical interconnects.

Silicon microring resonators play pivotal roles in future optical interconnect systems due to their ultra-compact sizes and low energy consumption. The typical silicon microring resonator is operated by the reversed p-n junction to achieve electro-optic (E-O) modulation, but it suffers from the relatively low E-O tuning efficiency. Transparent conducting oxides (TCOs) have attracted attention due to the dramatic change of their refractive indices owing to the strong plasma dispersion effect. This talk will provide an overview of our recent research progress in the heterogeneously integrated silicon microring resonator based on embedded metal-oxide-semiconductor capacitors using TCOs to enhance E-O tuning efficiency.

Large-scale, electronically reconfigurable photonic integrated circuits (PICs) can enable programmable gate array (PGA) to realize extremely fast, arbitrary linear operations, with potential applications in classical and quantum optical information technology. The basic building blocks of existing PGAs are thermally tunable broadband Mach-Zehnder-Interferometers, which pose several limitations in terms of size, power, and scalability. Phase change materials (PCMs), exhibiting large nonvolatile change in the refractive index, can potentially transform these devices, providing at least one order of magnitude reduction in the device size, zero static energy consumption, and minimal cross-talk. In this talk, I will discuss different PCMs that can be used in conjunction with silicon and silicon nitride photonics, to create reconfigurable optical switches for visible and infrared wavelengths. I will also talk about different heaters that are needed to actuate the phase transitions on-chip.

Heterogeneous integration of III-V on silicon lasers eliminates some constraints of chip-to-chip alignment, but the optical coupling between the two media remains of importance for repeatable performances. First, we present a processing enhancement of the bonding oxide thickness uniformity across the wafer, improving the cross-section reproducibility. Next optimized tapering of the III-V/Si waveguides, offering a design agnostic to the number of quantum wells, will be shown. Finally, the yield of III-V on Silicon tunable lasers was evaluated by mean of wafer level measurements, using a yield oriented tuning of each cavity, so that lasers characteristics can be fairly compared.

We route the single photons from a trapped barium ion in a nanophotonic circuit. For this routing, we first generate C-band telecom single photons from barium ion which makes them compatible with the silicon-nitride photonic foundry. Then using the thermo-optic property of silicon-nitride, we switch the single photons in a Mach-Zehnder interferometer controlling the current of the phase-shifter. These results could enable a new generation of compact and reconfigurable integrated photonic devices that can serve as efficient quantum interconnects for quantum computers and sensors.

Optical interposers are promising as a robust, reliable, and scalable technology for high-density coupling between the dissimilar platforms of optical fiber and silicon photonics (SiP) chips. To extend this concept, femtosecond laser micro-structuring was harnessed to develop a multi-level, mirror-waveguide optical circuit platform in fused silica glass. The flexible laser writing facilitated compact, low-profile vertical interconnection between multi-core fibers and SiP circuits, exploiting total internal reflection mirrors and vertical grating couplers. Various design strategies of laying out 3D waveguide fanouts, multi-core fiber sockets, and turn-mirrors were explored in 40 channel systems. The flexible interposer technology is scalable to higher channel counts, while maintaining a small footprint, thus offering a broad solution to challenges in areas of optical interconnects and photonic packaging.

At PhotonFirst we develop integrated photonics-based sensing solutions for a broad range of applications and markets, with an emphasis on Fiber Bragg Grating (FBG) monitoring applications. The way in which photonic building blocks are combined and connected is vital for the performance, reliability, footprint, manufacturability, and applicability range of the solution. Targeting a broad market range, requires understanding the commonalities and differences of the specific needs and impact on the product. In our contribution we highlight and demonstrate several examples from our capabilities and solutions and discuss focus areas for future developments of specific interest for sensing solution development.

Silicon nitride waveguide based photonic integrated circuits (PICs) are intensively investigated for a wide range of sensing applications in the visible to sub 1-µm near-infrared spectral region. The monolithic co-integration of silicon photodiodes and read-out electronics offers additional benefits in terms of performance and miniaturization. We discuss challenging aspects related to the efficient coupling and routing of light to, from, and within PICs and present interfacing photonic building blocks offering potential solutions. We demonstrate the suitability of these interfacing building blocks by using them for the realization of a PIC-based multi-channel optical coherence tomography concept at 840 nm.

We have proposed a low-noise graded-index plastic optical fiber (GI POF) with the microscopic heterogeneous strictures in the fiber core material. The microscopic heterogeneities in the low-noise GI POF provide strong mode coupling and stabilize data transmission through the reduction in interferometric noise in a multimode fiber link. Here, we present our latest work in which error-free data transmission employing four-level pulse-amplitude modulation (PAM-4) at a data rate of 53 Gb/s has been achieved using the low-noise GI POF without the use of forward error correction techniques in short-reach links. We also present the detailed analysis of the distinctive microscopic heterogeneities in the low-noise GI POF that achieves error-free PAM-4 transmission.