Leveraging the power of machine learning, we introduce a breakthrough approach in high-volume manufacturing of photonics chips for advanced applications. Despite the transformative potential of photonics in many industries, its widespread adoption has been hindered by multiple challenges in the fabrication of complex integrated chips. We deployed machine learning models with diverse architectures at every stage of our manufacturing process to overcome these challenges. Inevitable variations in the fabrication process often lead to performance variability among photonics chips on a single wafer and across different wafers. We effectively overcome this challenge by employing a deep neural network to study the variability in the performance of individual chips, enabling us to predict the precise optimizations necessary to compensate for inevitable process variations. We describe our selection of the deep neural network architecture that addresses this challenge, our methodology for obtaining a high-quality dataset for training, and the enhancements in performance uniformity achieved through machine learning-enhanced production masks. Moreover, our use of machine learning has allowed us to bypass the time-consuming and labour-intensive process of optical chip testing, which significantly limits the scalability of photonic deployments in high-volume applications. As a powerful alternative to such testing, we developed a new technology that relies on a wafer probe that collects metrology data from multitude of locations on an undiced wafer. Utilizing a support vector machine (SVM), we analyze this metrology data and employ nonlinear binary classification to accurately predict the performance of hundreds of chips on a wafer across various metrics. We describe the approach employed for data collection to train the model, the trade-offs involved in hyperparameter tuning, and our methodology for evaluating the predictive quality of the binary classifiers. Additionally, we highlight the new capability of in-situ monitoring of wafer fabrication, which enables high-volume production and deployment of photonic solutions.
We introduce cutting-edge monolithically integrated photonic designs realized in a high-performance silica-on-silicon planar lightwave circuit (PLC) platform. Systems-on-chip require integration from a dozen to a few hundred optical functions, necessitating component and wafer level optimizations. Our closed-loop feedback framework enables us to achieve low propagation losses (<0.009 dB/cm), efficient fiber coupling (0.5 dB/facet), temperature stability (< 10 pm/°C), wavelength-independent operation, as well as tight polarization and phase control. Due to the lack of two-photon absorption, low scattering, and negligible absorption, our silica-on-silicon platform is well-suited for high-pump power applications in LiDAR and accelerated computing. We discuss how these characteristics allow us to monolithically integrate high-performance optical building blocks such as K-clocks, cascaded lattice filters, polarization-beam splitters, and optical hybrid components into systems-on-chip for advanced photonics applications. We demonstrate the versatility and robustness of the platform by discussing examples of monolithically integrated chips used in AI/computing accelerators, and advanced vision applications based on LiDAR. Based on exceptional optical characteristics achieved through our platform, our systems-on-chip have emerged as high-performance and scalable solutions, capable of meeting the rigorous demands imposed by a wide range of applications.
We present advanced photonic architectures based on a state-of-the-art silica-on-silicon planar lightwave circuit (PLC) platform. The industry-leading performance characteristics of the platform with respect to fiber coupling, propagation losses, and polarization control have led to its mass deployments in telecom and datacom applications. Recently, a broad range of architectures has emerged that take advantage of the unmatched phase control in silica-on-silicon PLCs for demanding applications. We review some exceptional capabilities of our PLC platform and discuss how ultra-low propagation loss of <0.009 dB/cm is achieved concurrently with polarization-insensitive operation with relatively high confinement that allows loss-free waveguide bends with a 1 mm radius of curvature. Combined with fiber-matched mode converters and temperature-stable operation (< 10 pm/°C), consistent performance is achieved across entire optical communication bands. End-to-end optimizations allow us to reach high performance in advanced optical building blocks such as cascaded lattice filters, polarization-beam splitters, arrayed waveguide gratings (AWGs), and coherent systems. The robustness and versatility of the platform are demonstrated by a survey of mature designs that encompass multiple classes of applications. We discuss multi-channel (de-)multiplexer designs that address the challenging requirements of today’s datacom and telecom deployments, the realization of 10+ meter long delay lines and K-clocks, the utilization of the platform in optical coherence tomography (OCT) systems, and PLC-based solutions for automotive manufacturers of LiDAR systems. Finally, we discuss how our design, manufacturing, and testing processes are controlled with machine learning, allowing in-situ monitoring of wafer fabrication, real-time process adjustments, and wafer-level predictions of device performance across a wide range of performance metrics.
Artificial intelligence (AI) and machine learning (ML) have tremendous potential for increasing the scale and reach of the photonics industry. We present how the use of AI/ML has revolutionized the field of photonic integrated circuit design and manufacturing, and resulted in mass deployments of high-performance optical chips for multiple classes of datacom and telecom applications. First, we discuss our use of a deep neural network multivariate regression model to optimize the individual design parameters of hundreds of optical chips on a given mask. This work successfully addresses the systematic processing variations within a wafer, resulting in an unprecedented homogeneity of performance of optical chips in a high-volume production environment. Second, we present our approach of using ML to predict the performance of optical devices by wafer probing. This novel approach eliminates the expensive and time-consuming process of optical chip testing and instead relies on a wafer probe measurement to infer the performance of hundreds of chips on a wafer. We discuss the complexity of the problem of predicting the performance in multi-dimensional parameter space, the inherent challenges that cannot be overcome by traditional methods, and the reasons why ML is an essential tool to solve this problem. The support vector machine (SVM) that we developed performs nonlinear binary classification based on a regression from the probe measurement, allowing unprecedented control over our process, including in-situ monitoring of wafer fabrication and real-time process adjustments, and thus achieving consistently high performance of optical chips at high production volumes.
Planar lightwave circuits (PLCs) provide economical, high-capacity solutions for systems using wavelength-division multiplexing. To accommodate a higher volume of optical integration in a smaller footprint, PLC technology has trended towards higher refractive index contrast platforms resulting in tighter optical confinement. Further progress in the densification of photonic functionality, especially for multi-stage interferometric configurations, must rely on the development of advanced architectures to increase the density of functional units. We present a breakthrough approach to the synthesis of ultra-dense interferometric chains, reaching packing density of waveguides close to theoretical limits. The proposed framework is well suited for mixed parallel and sequential interferometric structures in low- or high-refractive index contrast platforms. The new methodology allows the addition of stages to an interferometric chain without appreciable increase in device footprint, thus creating a highly-optimized ultra-dense waveguide layout. To validate this approach, we designed and fabricated a 4-λ LAN multiplexer that comprises 7 interferometric stages in a silica-on-silicon platform with a refractive index contract of Δn = 2.0%. Despite the relatively low refractive index contrast, the device was realized in a footprint of only 0.15 cm2. The multiplexer exhibits exceptional optical performance, including on-chip loss of 0.2 dB, negligible polarization-dependent loss, and a remarkably flat single-mode spectral response with no insertion loss penalty. This ultra-compact implementation, combined with the state-of-the-art optical performance characteristics, led to a wide deployment of the multiplexer in data center applications, and provided a rapidly-advancing roadmap for unprecedented densification of optical functionality in PLCs in any refractive index platform.
By leveraging advanced wafer processing and flip-chip bonding techniques, we have succeeded in hybrid integrating a
myriad of active optical components, including photodetectors and laser diodes, with our planar lightwave circuit (PLC)
platform. We have combined hybrid integration of active components with monolithic integration of other critical
functions, such as diffraction gratings, on-chip mirrors, mode-converters, and thermo-optic elements. Further process
development has led to the integration of polarization controlling functionality. Most recently, all these technological
advancements have been combined to create large-scale planar lightwave circuits that comprise hundreds of optical
elements integrated on chips less than a square inch in size.
We present innovations in Planar Lightwave Circuits (PLCs) that make them ideally suited for use in advanced defense
and aerospace applications. We discuss PLCs that contain no micro-optic components, no moving parts, pose no spark
or fire hazard, are extremely small and lightweight, and are capable of transporting and processing a range of optical
signals with exceptionally high performance. This PLC platform is designed for on-chip integration of active
components such as lasers and detectors, along with transimpedance amplifiers and other electronics. These active
components are hybridly integrated with our silica-on-silicon PLCs using fully-automated robotics and image
recognition technology. This PLC approach has been successfully applied to the design and fabrication of multi-channel
transceivers for aerospace applications. The chips contain hybrid DFB lasers and high-efficiency detectors, each capable
of running over 10 Gb/s, with mixed digital and analog traffic multiplexed to a single optical fiber. This highlyintegrated
functionality is combined onto a silicon chip smaller than 4 x 10 mm, weighing < 5 grams. These chip-based
transceivers have been measured to withstand harsh g-forces, including sinusoidal vibrations with amplitude of 20 g
acceleration, followed by mechanical shock of 500 g acceleration. The components operate over a wide range of
temperatures, with no device failures after extreme temperature cycling through a range of > 125 degC, and more than
2,000 hours operating at 95 degC ambient air temperature. We believe that these recent advancements in planar
lightwave circuits are poised to revolutionize optical communications and interconnects in the aerospace and defense
industries.
We report on recent progress in simulations, physical layout, fabrication and hybridization of planar grating-based transceivers for passive optical networks (PONs). Until recently, PON transceivers have been manufactured using bulk micro-optical components. Today, advancements in modeling and simulation techniques has made it possible to design complex elements in the same silica-on silicon PLC platform and create an alternative platform for manufacturing of bi-directional transceivers. In our chips we simulated an integrated chip that monolithically combined planar reflective gratings and cascaded Mach-Zehnder interferometers. We used a combination of the finite element method and beam propagation method to model cascaded interferometers with enhanced coupling coefficients. Our simulations show that low-diffraction order planar reflective gratings, designed for small incidence and reflection angles, possess the required dispersion strength to meet the PON specifications. Subsequently, we created structures for passive alignment and hybridized photodetectors and lasers. We believe that advancements in simulation of planar lightwave circuits with embedded planar reflective gratings will result in displacement of the thin-film filters (TFFs) technology in many applications that require a high degree of monolithic and hybrid integration.
Recent deployments of fiber-to-the-home (FTTH) represent the fastest growing sector of the telecommunication
industry. The emergence of the silicon-on-insulator (SOI) photonics presents an opportunity to exploit the wide
availability of silicon foundries and high-quality low-cost substrates for addressing the FTTH market. We have now
demonstrated that a monolithically integrated FTTH demultiplexer can be built using the SOI platform. The SOI filter
comprises a monolithically integrated planar reflective grating and a multi-stage Mach-Zehnder interferometer that were
fabricated using a CMOS-compatible SOI process with the core thickness of 3.0 ?m and optically insulating layer of
silica with a thickness of 0.375 ?m. The Mach-Zehnder interferometer was used to coarsely separate the 1310 nm
channel from 1490 and 1550 nm channels. Subsequently, a planar reflective grating was used to demultiplex the 1490
and 1550 nm channels. The manufactured device showed the 1-dB bandwidth of 110 nm for the 1310 nm channel. For
the 1490 nm and 1550 nm channels, the 1-dB bandwidth was measured to be 30 nm. The adjacent channel isolation
between the 1490 nm and 1550 nm channels was better than 32 dB. The optical isolation between the 1310 nm and
1490 and 1550 nm channels was better than 45 dB. Applications of the planar reflective gratings in the FTTH networks are discussed.
The deployment of Passive Optical Networks (PON) for Fiber-to-the-Home (FTTH) applications currently represents
the fastest growing sector of the telecommunication industry. Traditionally, FTTH transceivers have been
manufactured using commodity bulk optics subcomponents, such as thin film filters (TFFs), micro-optic collimating
lenses, TO-packaged lasers, and photodetectors. Assembling these subcomponents into a single housing requires active
alignment and labor-intensive techniques. Today, the majority of cost reducing strategies using bulk subcomponents
has been implemented making future reductions in the price of manufacturing FTTH transceivers unlikely. Future
success of large scale deployments of FTTH depends on further cost reductions of transceivers. Realizing the necessity
of a radically new packaging approach for assembly of photonic components and interconnects, we designed a novel
way of hybridizing active and passive elements into a planar lightwave circuit (PLC) platform. In our approach, all the
filtering components were monolithically integrated into the chip using advancements in planar reflective gratings.
Subsequently, active components were passively hybridized with the chip using fully-automated high-capacity flip-chip
bonders. In this approach, the assembly of the transceiver package required no active alignment and was readily
suitable for large-scale production. This paper describes the monolithic integration of filters and hybridization of active
components in both silica-on-silicon and silicon-on-insulator PLCs.
Recent progress in the development of planar reflective gratings has resulted in the demonstration of multiplexers, comb filters, interleavers, power monitors, and receivers for long-haul and metro-area networks. Until recently, all of these devices were based on a single-grating architecture. We have now successfully designed, fabricated, and tested optical chips that are composed of cascaded planar reflective gratings. The chips have been realized in both additive and subtractive dispersion configurations. The versatility of cascaded gratings was utilized to produce a variety of optical responses, including single-mode transmission of wide bands (> 100 nm) with simultaneous demultiplexing of narrow optical channels with Gaussian and box-like responses. We have further demonstrated that cascaded gratings can be used to suppress optical noise and improve isolation. The devices were fabricated using a standard silica-on-silicon process with a refractive index contrast of 0.82% and have a remarkably small footprint of less than 0.3 sq. cm. We discuss the potential for tailoring of cascaded planar reflective gratings for applications in biophotonics, spectroscopy, and telecommunications.
Optical add/drop multiplexers (OADMs) have emerged as the key enabling components for building long-haul and metro-area networks. The wide-spread deployment of OADMs in the access market will depend on the availability of cost-effective integrated solutions. We have successfully fabricated a fully-integrated OADM based on planar reflective gratings. The device uses a combination of two grating elements arranged in a subtractive dispersion configuration. The first grating demultiplexes a 300-nm-wide band and drops optical channels at 1490 nm and 1550 nm, commonly used by service providers to send information to the end user. The second grating completely counter-balances the dispersion properties of the first grating and ultimately yields zero dispersion in the output waveguide. Such a configuration allows the transmission of optical signals though the OADM in an ultra-wide band spanning 1250 to 1410 nm. This ultra-wide 'through' band is a critical step allowing the use of low-cost lasers, without temperature stabilization, for sending data to a service provider. The OADM was manufactured using an industry standard silica-on-silicon process which was augmented with grating facet formation and metallization. In spite of using low refractive index contrast waveguides (0.82%), the device had a remarkably low footprint of only 0.25 square centimeters. Applications of the OADM in access market networks is discussed.
A new approach for constructing devices of various free spectral ranges (FSRs) is described. We show that devices with different FSRs can be built around the same aberration-free architecture based on elliptical grating facets. Elliptical facets, combined with double astigmatic point design, are demonstrated to lead to dramatic improvements in reflective grating performance compared to traditional flat facet designs. A discussion on the proper selection of the grating order for devices with various FSRs is given. The proposed theory was applied to manufacture devices with various FSRs. A standard silica-on-silicon process was used to fabricate interleavers with narrow FSR of 0.8 and 1.6 nm. Subsequently, we show how the above methodology can be used to scale the reflective grating design to devices with wide FSR. We applied the theory to produce coarse wavelength division multiplexing filters with FSR in excess of 500 nm. The filters exhibited insertion losses of 2.5 dB and polarization dependent losses of less than 0.2 dB. Applications of wide FSR devices in metro edge and access networks are discussed.
We report the results of an experimental study on near- threshold gain mechanism in optically pumped GaN epilayers and GaN/AlGaN separate confinement heterostructures (SCHs) over the temperature range of 10 to 300 K. We show that in GaN epilayers the near-threshold gain mechanism is inelastic exciton-exciton scattering for temperatures below approximately 150 K, whereas at elevated temperatures an electron-hole plasma is the dominant gain mechanism. An analysis of the relative shift between the spontaneous emission and lasing peaks in SCH samples, combined with the temperature dependence of the lasing threshold, reveals that exciton-exciton scattering is the dominant gain mechanism leading to low-threshold ultraviolet lasing in the GaN/AlGaN SCH over the entire temperature range studied. Strongly polarized (TE:TM > 300:1) lasing peaks were observed in a wavelength range of 358 - 367 nm. We found that high finesse lasing modes originated from self-formed microcavities in the AlGaN and GaN layers. The lasing threshold was measured to be as low as 15 kW/cm2 at 10 K and 105 kW/cm2 at room temperature. Based on our results we suggest ways for the realization of GaN-active-medium UV laser diodes.
Stimulated Emission and Pump-Probe studies were performed in GaN, InGaN, and AlGaN epilayers as well as GaN/AlGaN separate confinement heterostructures. We show that in GaN epilayers the near-threshold gain mechanism is inelastic exciton-exciton scattering for temperatures below approximately 150 K, whereas at elevated temperatures electron-hole plasma is the dominant gain mechanism. An analysis of the relative shift between the spontaneous emission and lasing peaks in SCH samples, combined with the temperature dependence of the lasing threshold, reveals that exciton-exciton scattering is the dominant gain mechanisms leading to low-threshold UV lasing in the GaN/AlGaN SCH over the entire temperature range studied. We further performed optical pumping of AlGaN epilayers at different temperatures. Stimulated emission has been observed in AlxGa1-xN thin films for Al concentrations as high as x = 0.26, with a resultant stimulated emission wavelength as low as 328 nm at room temperature. This result indicated that AlGaN-based structures are suitable not only for use in deep-UV detectors, but also as a potential source of deep-UV laser radiation. The interband optical transitions in GaN and InGaN have also been studied at 10 K and room temperature using nondegenerate nanosecond optical pump-probe techniques. At low temperatures, strong, well- resolved features were seen in the absorption and reflection spectra corresponding to the 1s A and B exciton transitions. Broadening and decrease in intensity of these features were studied as the function of excitation pump density. We found that values of induced transparency and induced absorption are extremely large in GaN epilayers. The pump-probe results in GaN epilayers were directly compared to ones obtained from InGaN films. Significant differences in near-bandedge absorption changes were clearly observed between the two materials.
Edge and surface-emitted stimulated emission (SE) in optically pumped GaN thin films was studied in the temperature range of 20 K to 700 K. The single-crystal GaN films used in this work were grown by MOCVD on sapphire and 6H-SiC substrates. We have observed that the SE peak shifts from 360 nm at 20 K to 412 nm at 700 K, which is the highest temperature at which SE has been reported for this material. The temperature sensitivity of the SE threshold was studied over the entire temperature range. The characteristics temperature was found to be about 170 K over the temperature range of 300 K to 700 K for samples grown on both sapphire and SiC substrates. The energy position of the SE and spontaneous emission peaks were shown to shift linearly to longer wavelengths with increasing temperature and empirical expressions for this shift are given. We demonstrate that the energy separation between the spontaneous and SE peaks gradually increases from 90 meV at 300 K to 200 meV at 700 K indicating that an electron-hole plasma is responsible for the SE mechanisms above room temperature (RT). We demonstrate that the surface-emitted SE in GaN epilayers comes from cracks, burn spots, and other imperfections, and is due to the scattering centers and, under strong optical excitation, become points of origin for burning of the sample surface. This study shows that GaN has an extremely low temperature sensitivity compared to other semiconductors and is suitable for the development of light emitting devices that can operate significantly above RT.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.