This PDF file contains the front matter associated with SPIE Proceedings Volume 12007, including the Title Page, Copyright information, Table of Contents, and Conference Committee listings.

The exponential growth in data center traffic is driving an increase in demand for cost and energy efficient interface bandwidth scaling. This creates a need for a paradigm shift in I/O technology that meets future connectivity requirements within data centers. Silicon Photonics (SiPh) based optical interfaces significantly improve I/O density by optimizing solutions along three vectors: Packaging density, speed per lane, and number of wavelengths per channel. Each vector must be independently optimized. This paper recommends a SiPh based optical I/O platform that merges mature silicon chiplet packaging and fiber connectivity to achieve the highest I/O efficiency for networking applications.

During the past years, the processor-memory performance gap, also known as “Memory Wall” problem, has forced designers to allocate more than 50% of the chip real-estate for caching purposes to alleviate limited memory bandwidth. Optical technology holds the credentials of delivering high-bandwidth and energy-efficient photonic integrated memories that could revisit the traditional computing architectures. The migration, however, to fully functional and practical optical RAMs will require the exploitation of wavelength dimension as well as seamless cooperation between storage and peripheral decoding units, for efficient RAM architectural layouts. In this paper we present the first demonstration of an all-optical 8-bit RAM storage unit comprising WDM-enabled 2×4 Row and 1×4 Column Decoders and a 2×4-bit optical RAM Bank for storing a 20Gb/s 4-bit WDM-formatted optical data word per row. The proposed scheme incorporates a shared multi-λ SOA-MZI Access Gate (AG) per Word Line (WL) for granting access-control to the appropriate word line, WL “00” or WL “01”, and a passive Column Decoder that directs the incoming WDM-formatted data words to the respective RAM cells. Each RAM cell is in turn based on an elementary monolithically integrated InP photonic Flip-Flop (FF). The proposed architecture is experimentally verified for successful Write operation of a 4-bit WDM word to a selected 4-bit RAM row at 20Gb/s RAM throughput and a peak power penalty within the range of [7.8-10.7] dB, promising a 4× speed-up in memory-access throughput and paving the way for high-bandwidth multi-bit optical RAM-architectures that may relax the memory-bottleneck of computing architectures.

This paper reviews the current use of optical connection in top-level supercomputers and describes future expectations. The inter-node interconnects of seven world’s No.1 systems of the last decade are introduced focusing on the use of optical connection. The interconnect and system configuration of the current No.1 system Fugaku is also explained in detail. This paper also discusses the use of optical connection in future supercomputers. The system configuration and interconnect of Frontier, which is expected to become the world's No.1 supercomputer, is introduced. Towards the late 2020s, expectations for higher transmission speeds of 100 to 200 Gbps and the use of CPO are described. The impact of additional latency caused by FEC on parallel computing is also discussed.

We demonstrate direct bandwidth measurement of multimode polymer waveguides based on an optical sampling technique. The pulse shape can be completely recovered after transmission due to the advantages of optical sampling technology in the field of ultrashort pulse measurement. A reduction in averaged bandwidth (Bandwidth-length products) from 241 GHz (27 GHz·m) to 180 GHz (20 GHz·m) of 11 cm-long straight waveguides is observed when using mode scramblers (MS) to fully stimulate the higher-order modes. The effects such as bending, crossing, and twisting of both the rigid and flexible waveguides on the bandwidth are also investigated. The proposed method is effective for measuring the bandwidth and dispersions of waveguides, fibers and optical devices with a short length for varies applications.

The emergence of demanding machine learning and AI workloads in modern computational systems and Data Centers (DC) has fueled a drive towards custom hardware, designed to accelerate Multiply-Accumulate (MAC) operations. In this context, neuromorphic photonics have recently attracted attention as a promising technological candidate, that can transfer photonics low-power, high bandwidth credentials in neuromorphic hardware implementations. However, the deployment of such systems necessitates progress in both the underlying constituent building blocks as well as the development of deep learning training models that can take into account the physical properties of the employed photonic components and compensate for their non-ideal performance. Herein, we present an overview of our progress in photonic neuromorphic computing based on coherent layouts, that exploits the phase of the light traversing the photonic circuitry both for sign representation and matrix manipulation. Our approach breaks-through the direct trade-off of insertion loss and modulation bandwidth of State-Of-The-Art coherent architectures and allows high-speed operation in reasonable energy envelopes. We present a silicon-integrated coherent linear neuron (COLN) that relies on electro-absorption modulators (EAM) both for its on-chip data generation and weighting, demonstrating a record-high 32 GMAC/sec/axon compute linerate and an experimentally obtained accuracy of 95.91% in the MNIST classification task. Moreover, we present our progress on component specific neuromorphic circuitry training, considering both the photonic link thermal noise and its channel response. Finally, we present our roadmap on scaling our architecture using a novel optical crossbar design towards a 32×32 layout that can offer >;32 GMAC/sec/axon computational power in ~0.09 pJ/MAC.

A direct optical interface on ICs has the potential to ease the bandwidth bottlenecks in transferring data between advanced ICs. We demonstrate high speed microLEDs transferred onto silicon CMOS circuitry together with the integrated drivers for the LEDs to send data from the IC. On the receive side of the same IC, we demonstrate integrated CMOS detectors with trans-impedance and limiting amplifiers. These chips are shown to operate at Gb/s speeds and can be interfaced with multicore fibers to make simple low-cost data-paths between standard silicon ASICs. Compared to SERDES based interfaces where only a few lanes are run at very high speeds, these wide parallel optical interfaces can be considerably lower power and offer much higher overall bandwidth and bandwidth density. We demonstrate these links using 130nm CMOS process on SOI substrates, with <2pJ per bit and show their superior performance compared to FP lasers in terms of BER and mode partition noise.

The commoditization of photonics would be possible only with the development of photonic integrated circuits and appropriate volume applications that require them. As such an application, a light detection and ranging(LiDAR) sensor has recently been in strong demand from various applications including autonomous driving. In terms of technology, as silicon photonics enters an industrial phase and begins to utilize the existing CMOS infrastructure, photonic integrated circuits are also expected to enter a virtuous cycle of volume and cost. This work outlines the current status of LiDAR research using the silicon photonics platform in Samsung. Based on the III/V-on-Si technology, Samsung's platform enables the development of chip-scale LiDAR that integrates all photonic devices such as wavelength-tunable laser, semiconductor optical amplifier, and custom optical phased array. With the LiDAR chip in the core, a palm-top LiDAR module prototype including control and signal processing circuits is also presented. Then, initial application-level attempts in autonomous driving are presented in the hope of pathfinding towards the LiDAR commoditization, and more broadly, commoditization of photonics.

The fundamental vibrational-rotational absorption signature of almost all the chemical compounds lies in the Mid- IR spectrum (λ=3-15μm) thus offering superior light-analyte interaction in this regime. In particular, the successful inscription of infrared-spectroscopy in a multi-pass cell has significantly boosted its use mainly in the gas-sensing application at the sub-ppm/ppb level. However, the requirement of bulky, alignment sensitive, and need of expertise-hands makes it inappropriate for many fields especially in portable applications like stand-alone environment monitoring, detection of chemical-warfare-agent in the battlefield, Astro-biological applications, etc. An external disruption-free handheld device (i.e., unaffected from any external vibration, physical stress, and thermal variations) with high specificity and selectivity are still prerequisites for such in-situ applications. The advancements in photonics have shown enormous possibilities to miniaturize all spectroscopic components to a single chip. In this context, the slow light-assisted engineered photonic structure on a QCL/QCD (quantum confined laser and detector) is most promising to replace bulky multi-pass cell optics. In principle, it slows down the light with several folds to enhance the light-analyte interaction and thus open an avenue for an on-chip sensing platform. Most efficient QCLs demonstration explored in the InP platform, also a selection of InP-InGaAs eliminates the requirement of the costly wafer-bonding process. In this paper, we consider slow-light assisted and wavelength-tunable periodic photonic structures. The device is designed such that it supports transverse magnetically polarized mode directly emitted from QCLs. It eliminates the use of any additional polarization-rotator (conversion from TM to TE mode) which reduces fabrication complexity and additional space on the chip.

In this paper, in order to apply to optical connection components such as fan-in-fan-out (FIFO) devices for multicore fibers (MCFs), we design polymer optical waveguides with circular cores satisfying the single-mode condition, which are fabricated using the Mosquito method, and experimentally demonstrate that the optimum S-shape bent core waveguide can be fabricated as designed. The Mosquito method is a technique we have developed to form circular cores in both multimode and single-mode waveguides using a commercially available microdispenser and multi-axis syringe scanning robot. We have demonstrated that single-mode waveguides fabricated by the Mosquito method can be coupled with SMF with low loss. However, in order to realize a waveguide type FIFO device for MCFs by the Mosquito method, it is necessary to maintain micron-order positioning accuracy for the cores while three dimensionally aligning multiple cores in FIFO patterns. Since the pitch between the cores on the MCF side is very narrow, the core position is likely to deviate from the design value due to the liquid monomer flow caused by needle scan. We statistically confirm the possible amount of deviation from the design position by needle scan and feed it back to the program of needle scan path for the FIFO. We successfully fabricate a four-core FIFO with a desired core pitch. It is experimentally confirmed that the insertion loss is 1.4 dB on average. From these investigations, we confirm that the Mosquito method has an ability to fabricate small and low-loss FIFO devices.

We give an overview of our recent progress on interfacing components for short-reach optical interconnects fabricated through two-photon polymerization-based laser direct writing. We show mode field conversion tapers printed on single-mode optical fibers for easy and efficient interfacing to various photonic integrated circuits, circular and square planar waveguide structures with V-groove inspired alignment structures for easy coupling to fibers and fan-out diffractive optical elements. For all these components, we present the process flow from optical design and simulation over laser direct writing fabrication and metrology to proof-of-concept demonstration.

Printing optical waveguides is an approach to the high volume implementation of optical data transmission in conventional electronic systems. Flexographic printing enables the manufacturing of circular segment-shaped polymer waveguides on planar substrates, which show great potential as economic Gbit/s-capable short-range networks. This work describes a process chain to manufacture and integrate a printed optical data transmission path in conventional printed circuit boards (PCB). This sequence of processes gives an outlook on up-scaling utilizing printed optical waveguides to mass manufacturing. Since the significant challenge in integration is achieving sufficient optical coupling, geometrical tolerances are investigated using raytracing simulation. Relevant degrees of freedom of the laser diode and waveguide are varied and validated by measuring alignment profiles. As a result, the mechanical interface provided by the PCB is presented and validated by confocal measurements. An innovative pick and place tool assembles the separated flexible waveguide to realize a demonstration system. As a validation, Fast Ethernet data transmission is presented over a flexible optical connection. In further steps, a miniaturization of the system is the goal to achieve a standardized system for applications like galvanic isolation.

Current developments are pushing the integration of optical technologies deeper into the architecture of data centers,1 a trend in which co-packaging figures prominently due to its many inherent advantages2, 3. Several materials are used as a basis for these co-packaged platforms, but glass stands out for its many positive properties, such as high thermal and dimensional stability, great optical transparency, excellent high-frequency properties for electric circuits, and extremely low cost. To seize these advantages, we pursued an approach called electro-optical circuit board (EOCB), in which optical and electrical interconnections are realized by glass-integrated optical waveguides and electrical circuits on both sides of the glass board. An ion-exchange technique was developed to integrate low-loss optical single-mode waveguides into large-sized glass boards (457 mm x 303 mm). In the reported work, the next milestone in developing this process was achieved by reducing the diffusion metal mask opening’s width from 6 μm to 3 μm by mask-less laser patterning. These smaller mask opening allow for optical waveguides with a more circular modal field shape resulting in smaller coupling losses to optical fibers. Additionally, the reduction of propagation losses of multi-mode waveguides for wavelengths down to the visible range was achieved. This opens up the field of sensing and quantum application to EOCBs.

Optical fiber components have the potential of enabling interconnections in compact systems because they provide reliable and efficient manipulation of light in application fields such as telecommunication, sensing and high power. A variety of glasses and fiber components including tapers, tips, bundles and couplers are typically fabricated using hydrofluoric acidbased etching processes. However, such a standard approach has some limitations related to the generation of surface defects (e.g., roughness and microcracks), poor process control and high chemical disposal costs. We propose an innovative glass etching process based on molten salts that will overcome these limitations. Molten salts can be thermally activated to etch glass materials with high precision. Initial plant development and industrial manufacturing capabilities are demonstrated on a modular etching system through a research collaboration. This system also has the advantage of managing a set of fibers simultaneously with an automatic process control. First results of etched glasses and especially, biconical fiber tapers show extremely smooth surfaces, good homogeneity, high reproducibility and potential scalability for further processing of fiber couplers. With respect to the fabrication tolerances, a value of ± 1 μm over a length of 10 mm has been found for the case of etched multimode tapers. The use of molten salts as an etching tool can be extended to economically create microstructures in glass panels for optical or fluidic purposes.

We present inter-chip optical link based on direct optical wire (DOW) bonding technology fabricated by meniscus-guided polymerization in open-air. The arch shape DOW structure is formed in a single procedure for directly linking silicon photonic chips, where grating couplers are integrated to out-couple guided optical modes. Although a typical grating coupler is employed, the inter-chip DOW link supports a low insertion loss of 6 dB in total with a wavelength-insensitive operation in the measured wavelength range of 1520 nm to 1590 nm. The half-arch shape DOW for linking chip-to-fiber is also shown to verify the feasibility of hybrid integration with edge coupling devices. DOW bonding technology can provide a convenient route to enable direct optical link capable of agile and high-throughput manufacturing for inter-chip optical interconnection.

Flexographic printing enables the rapid production of complex planar optical networks at low costs. Only waveguides without an upper cladding have been fabricated using flexographic printing. In these systems, the waveguide core is guided in contact with air. The Polymer Poly(methyl methacrylate) (PMMA) has been used as the lower cladding in previous systems. In this paper, a process for a flexographically printed waveguide in a complete cladding of UV-curable polymer will be demonstrated for the first time. Straight waveguides with cladding are compared to waveguides with without an upper cladding. Attenuation is determined to be 1.85 dB cm⁻¹ and 2.1 dB cm⁻¹ , respectively. In addition, s-bends are manufactured. For these, the cutoff radius is 22 mm (not cladded) and 30 - 45 mm (cladded). The results show that it is possible to manufacture waveguides suitable for industrial applications flexographically.

New circuit architectures and technologies for high-speed electronic and photonic integrated circuits are essential to realize optical interconnects with higher symbol rate. As a consequence of the increasing speeds, close integration and co-design of photonic and electronic chips have become a necessity to realize high-performance transceivers with novel packaging approaches. Extensive co-design also enables the design of new electro-optic architectures to create and process optical signals more efficiently. This paper and presentation will illustrate a number of recent developments of application-specific high-speed electro-optic transceiver circuits including e.g. broadband driver amplifiers, transimpedance amplifiers, analog equalizers and multiplexer circuits for signal generation and reception at 100 Gbaud and beyond. The basic concepts and architectures, technological aspects, design challenges and trade-offs will be discussed.

Increasing the throughput of a transmitter by scaling out to multiple wavelengths in high bandwidth density links, such as those found in data centers or high performance compute clusters, is beginning to gain traction for two reasons. First, increasing the data rate per wavelength by both increasing the baud rate (>50G) and by increasing the number of bits per symbol (PAM-4) consumes more power and increases latency, from having to use a powerful FEC. Second, a technological advantage, in achieving tight integration of lasers with silicon based photonics, has reduced laser coupling losses and tighter control on laser wavelengths, which in turn allows for greater utilization of the available optical bandwidth. In this talk, we will review our current and past efforts in realizing multi-wavelength laser sources heterogeneously integrated on silicon as a means to generate fully integrated transmitters with bandwidth capacity in excess of 1Tbps over 40 channels. We will also discuss some of the recent results from our MOSCAP microring modulators and Si-Ge/ quantum dot based avalanche photodetectors that enable a fully integrated heater-free compact transceiver. The estimated power consumption for the optical components in such a transmitter is around 1pJ/bit, which is roughly a 10x reduction compared to the state-of-the-art.

An example of continue breakthrough in Silicon Photonics (SiPh) is heterogeneous integration of active devices at wafer level not just to overcome the natural band-gap limit of the Silicon, but more importantly to exploit its high level of integration, significantly reducing packaging costs while driving down the cost of optical communications. In this paper, we describe a powerful combination and coupling of integrated 45°up-reflecting mirrors with longwavelength InP vertical cavity surface emitting lasers (VCSELs) used to develop a TX module with aggregated capacity up to 2-Tb/s capacity. The Photonic Integrated Circuit (PIC) designed and developed under the H2020 research project PASSION, heterogeneously embeds 40 VCSELs covering the C band on a single 3μm -thick Silicon-On-Insulator (SOI) multiplexer chip. The PIC exit-waveguide is also terminated with an up-reflective mirror and coupled with a fiber optic– based periscope, minimizing the form-factor while improving mechanical reliability of the overall packaged module. VCSELs are directly modulated with Discrete Multi-Tone (DMT) modulation format allowing 50 Gb/s rate per VCSEL. The PIC dimension is about 20 x 20 sqmm and power consumption < 5 pJ/bit at 2Tb/s. A Land Grid Array (LGA) interposer hosting the PIC sorrounded by 40 (flip-chip bonded) linear VCSEL drivers, providing electrical and thermal decoupling to the PIC is also described, achieving a compact and a thermally efficient packaging solution. Conveniently, a modular approach is pursued using the same identical 2-Tb/s TX module when building up a supermodule enabling an aggregating capacity up to 16 Tb/s on a single polarization state.

We describe the assembly of a 5G transceiver leveraging photonics for the generation, emission and detection of THz wireless signals. The transceiver and all associated control electronics and power supplies are designed for mounting in a mobile aerial unit. A photonics motherboard concept that brings together polymer, III-V and SiNbased photonic platforms and provides optical fiber connectivity is used for the assembly. In addition, scalable integration of 3D components, in this case an antenna rod or rod array, is demonstrated. Thermal considerations arising from the dense integration of photonic and electronic components and the resulting concentrated heat load are also discussed.

Silicon photonics combined with complementary metal–oxide–semiconductor (CMOS) electronics leveraging wavelength-division multiplexing (WDM) are of interest for AI, optical computing, and high-speed Optical IO applications [1,2]. To power these applications, multi-wavelength light sources based on laser arrays [3] or mode locked lasers (MLL) have been proposed and demonstrated [4]. As optical sources mature, the CW-WDM multisource agreement (MSA) has emerged to define a set of wavelength grids and power levels so different applications can leverage a common set of laser technologies [5]. In this paper we demonstrate the first multi-wavelength optical source compliant with the CW-WDM MSA standard that operates from room temperatures through 100°C. The SuperNovaTM outputs 8 wavelengths across 8 fibers for a total of 64 optical carriers and complies with the 8+1 MSA wavelength plan (1 optional wavelength) with channels spaced at 400+/-100 GHz and output power within the Type 2 power class. The optical source is mode hop free with >40dB SMSR, <145 dB/Hz RIN, and <20 MHz linewidth across all channels and all operating conditions.

We report on glass-molded micro-optical interposers for single-mode fiber-to-PIC coupling fabricated in parallel by isothermal molding of 1-inch glass plates yielding over 100 arrays of 8 lenses each. Excess losses between PIC and singlemode fiber are below 1 dB. In addition to allowing a narrow package footprint, beam transformation maps mode profiles between fibers and surface couplers, and, in case of grating couplers, can adapt the light incidence angle on a wavelengthspecific basis, facilitating packaging of PICs in CWDM and LAN-WDM modules. The interposers can be further extended to support polarization management and isolation by coating with polarization selective thin film stacks in MacNeille configuration, as well as wavelength multiplexing by coating with dichroic stacks. Using this technique, over 256 bidirectional transceiver channels can be packaged in the footprint of a single reticle.

We present a C-band grating coupler for the 800 nm silicon nitride platform with fully etched grates and suppressed back-reflection. By interrupting the first few grates in the center of the beam incoming from the waveguide and gradually increasing their span, waveguide-to-waveguide back-reflection is reduced by 5-12 dB in the C-band, at which the proposed grating coupler features an extra 2.7 dB insertion losses compared to the 5.8 dB of a standard reference design. Such grating couplers are expected to be particularly useful for applications sensitive to optical feedback, such as external cavity lasers.

The recently emerged inverse design methods have become a widespread tool to address some of the bottlenecks in photonic devices. Consequently, increasing the functionality and efficiency of many designs are feasible with the judicious formulation of the inverse design. In the present work, different versions of inverse design methods are applied to obtain an integrated on-chip lens to focus broadband incident light between 1300 nm and 1750 nm. In another study, 1×2 polarization-insensitive wavelength selective structure made of a low-refractive-index material is investigated. The successful separation of 1300 nm and 1550 nm irrespective of the polarization type (either transverse-electric, TE, or transverse-magnetic TM) is achieved. Taking advantage of the design with a low refractive index material, the structural dimensions are scaled and 3D printed versions are used in the experiment to prove the operating principles of the designs and demonstrate the good agreement between the numerical and experimental data. Efficient integrated light couplers and polarization-independent wavelength selective devices hold great potential to be an integrated part of photonic devices. Inverse design approaches will continue to improve the efficiency of conventional building blocks of integrated photonics devices and help the realization of novel nanophotonic structures.

Microfabricated lenses are a key-enabling technology for Datacom and Telecom applications, especially to implement co-packaged optics (CPO) for optical transceivers based on photonic integrated circuits (PIC). The manufacturing of such microlenses needs to be compatible with high-volume production, without sacrificing the optical quality and performances. For most applications, the output of an astigmatic beam from a laser diode needs to be collimated in the x- and y-directions: this correction cannot be performed by a rotationally-symmetric lens. A compact approach can be adopted by means of only one elliptical microlens, which has a different focal length in both directions. Here, we propose an analytical model, which can be applied for designing and fabricating elliptical microlenses by means of the reflow process, which is fully based on the optical description of the lens, in terms of RoC and conic constants. The model is demonstrated through a wafer-level fabrication of such lenses.

Based on a hypothetical 51.2 Tbps switch using co-packaged optics, we discuss key optical connectivity considerations and lay out different cabling options. Means are discussed to relax the length-accuracy requirements for fiber jumpers while maintaining a crossing-free cable layout for ease of assembly and serviceability. A key consideration is the mechanical stress in the fiber terminations which can lead to a degradation in the polarization integrity of external laser sources.

Optical communication roadmaps have been moved forward 3-4 years by the critical need for remote working during the pandemic. The development of 800 Gb/s and 1.6 Tb/s technologies exceeds original projections with the first co-packaged optical modules anticipated in 2023. Furthermore, advances in quantum computing and quantum communication are creating a demand for new “quantum grade” optical fiber interconnect to support quantum networks and connectivity to quantum photonic integrated circuits. In this paper, we report on the latest advances in optical interconnect required to enable the proliferation of co-packaged optics, quantum networks and quantum photonic integrated circuits.

We propose a novel high-density and low-profile multi-fiber passive assembly technology to polymer waveguides. We developed a small-sized ferrule with a footprint of only 5 × 5 mm and a thickness of 0.5 mm. An embedded 12 v-groove silicon chip was employed as a receptacle, and we formed the facet of waveguides on the v-groove chip. The optical properties for the proposed structure were experimentally demonstrated. The propagation loss of the single mode waveguides was as small as 0.5 dB/cm for the wavelength of 1310 nm and the average connecting loss of 3.06 dB was obtained.

Reflecting codes are frequently employed to reduce erroneous output from electromechanical/optical switches and to assist error correction in today’s communication systems such as digital terrestrial television and some cable TV systems. In this work, all optical reflective code is designed and simulated using a nonlinear optical effect inside metal-insulator-metal plasmonic waveguides based on Mach–Zehnder interferometer. Finite-difference time-domain (FDTD) method is used to analyze the performance of proposed structure and results are verified with MATLAB simulation.

We propose to leverage a silicon-organic hybrid integration structure to implement an integrated entangled photon pair source with high photon pair generation rate. This method combines the weak nonlinear absorption of organic materials with the high light confinement of silicon waveguides. Due to the supression of TPA and FCA, the pump power saturation threshold of the hybrid waveguide can be greatly increased. Therefore, with the high nonlinear coefficient and strong pump power, this hybrid integrated structure can achieve high photon pair generation rate. This work shows that silicon-organic hybrid integration could be a competitive platfrom for quantum photonic circuits.

We demonstrated a large aperture 1×16 silicon photonics OPA leveraging one-millimeter-long subwavelength grating antennas (SGA). The SGA is implemented by placing subwavelength silicon segments in the vicinity of the conventional strip waveguide so that they only interact with the evanescent field of the strip waveguide. The strength of the SGA can be conveniently controlled by tuning the location and size of the silicon segments. With the 1×16 OPA, a light beam of 0.1°×1.8° with a sidelobe suppression ratio > 10 dB is achieved.