An on-chip photonic deep neural network for image classification.

Deep neural networks with applications from computer vision to medical diagnosis1-5 are commonly implemented using clock-based processors6-14, in which computation speed is mainly limited by the clock frequency and the memory access time. In the optical domain, despite advances in photonic computation15-17, the lack of scalable on-chip optical non-linearity and the loss of photonic devices limit the scalability of optical deep networks. Here we report an integrated end-to-end photonic deep neural network (PDNN) that performs sub-nanosecond image classification through direct processing of the optical waves impinging on the on-chip pixel array as they propagate through layers of neurons. In each neuron, linear computation is performed optically and the non-linear activation function is realized opto-electronically, allowing a classification time of under 570 ps, which is comparable with a single clock cycle of state-of-the-art digital platforms. A uniformly distributed supply light provides the same per-neuron optical output range, allowing scalability to large-scale PDNNs. Two-class and four-class classification of handwritten letters with accuracies higher than 93.8% and 89.8%, respectively, is demonstrated. Direct, clock-less processing of optical data eliminates analogue-to-digital conversion and the requirement for a large memory module, allowing faster and more energy efficient neural networks for the next generations of deep learning systems.

Monolithic optical PAM-4 transmitter with autonomous carrier tracking.

We present two single channel optical PAM-4 transmitters, one based on a novel 3-section PN-capacitive micro-ring modulator with on-chip low-power driver and a near-zero power capacitive wavelength locking system and another one based on a 2-section thermally tuned PN micro-ring modulator of the similar size with the same modulator driver. The maximum error-free data-rate of 16 Gb/s and 22 Gb/s at the energy efficiency of 200 fJ/b and 430 fJ/b for the former and the latter transmitters are measured, respectively, and the design trade-offs are discussed. The chips are fabricated in the GlobalFoundries 90 nm CMOS SOI process.

Learning active sensing strategies using a sensory brain-machine interface.

Diverse organisms, from insects to humans, actively seek out sensory information that best informs goal-directed actions. Efficient active sensing requires congruity between sensor properties and motor strategies, as typically honed through evolution. However, it has been difficult to study whether active sensing strategies are also modified with experience. Here, we used a sensory brain-machine interface paradigm, permitting both free behavior and experimental manipulation of sensory feedback, to study learning of active sensing strategies. Rats performed a searching task in a water maze in which the only task-relevant sensory feedback was provided by intracortical microstimulation (ICMS) encoding egocentric bearing to the hidden goal location. The rats learned to use the artificial goal direction sense to find the platform with the same proficiency as natural vision. Manipulation of the acuity of the ICMS feedback revealed distinct search strategy adaptations. Using an optimization model, the different strategies were found to minimize the effort required to extract the most salient task-relevant information. The results demonstrate that animals can adjust motor strategies to match novel sensor properties for efficient goal-directed behavior.

Integrated coherent optical receiver with feed-forward carrier recovery.

Coherent optical communication provides optical links with a high spectral efficiency and sensitivity. An essential feature of a coherent optical receiver is to phase lock the optical local oscillator to the carrier of the incoming signal. In this work, we propose and demonstrate, for the first time, a novel coherent optical receiver, where the relative instantaneous phase between the incoming optical carrier and a semiconductor laser (SCL), serving as the optical local oscillator, is first detected using a balanced photodiode, filtered, and used in a feed-forward scheme to modify the phase of the optical local oscillator, effectively recovering the input carrier, which is then used for data recovery. The proposed architecture leverages high-performance on-chip photonic devices to realize a low-power coherent optical receiver without utilizing a phase-locked loop and eliminates the required high data-rate ADC, lowering the complexity of the backend DSP. The photonic part of the implemented prototype was integrated on a 180 nm silicon-on-insulator photonic process within a footprint of 1.0 mm × 0.8 mm. Clock and data recovery at 10 GBaud/s with bit-error-rates better than 10-6 and 10-3 for optical BPSK and QPSK have been demonstrated, respectively.

N × N optical phased array with 2N phase shifters.

Two-dimensional (2-D) integrated optical phased arrays (OPA) have many applications from optical imaging to LiDAR. Conventionally, 2-D beam-steering in an N × N OPA requires N2 phase shifters placed within the phased array aperture, resulting in a high per-element power consumption while limiting the minimum achievable element-to-element spacing. In this paper, we report an OPA architecture, where for 2-D beam-steering in an N × N OPA, only 2N phase shifters outside of the array aperture are used, which significantly reduces the total OPA power consumption and eliminates electrical routing within the aperture. As a proof of concept, an 8 × 8 OPA is implemented that uses 16 phase shifters to perform 2-D beam-steering without tuning the wavelength. Using the aperture size of 77 µm × 77 µm for the implemented OPA transmitter, far-field beam-steering over a range of about 7° is demonstrated.

Frequency locking of semiconductor lasers to RF oscillators using hybrid-integrated opto-electronic oscillators with dispersive delay lines.

Direct frequency locking of lasers to RF oscillators has many applications such as high resolution optical frequency synthesis, coherent optical communication, spectroscopy, sensing, and imaging. Here we present a hybrid-integrated opto-electronic loop that directly frequency locks a semiconductor laser to an RF synthesized source using an opto-electronic oscillator with a dispersive optical delay line. Cascaded ring filters, operating near the resonance frequency, provide an enhanced chromatic dispersion with a compact footprint. The electronic chip is integrated in the GlobalFoundries 180 nm CMOS SOI technology and the photonic chip is integrated in the IME 180 nm SOI technology. A tracking range of 0.5 GHz is achieved while consuming 33 mW power. The proposed scheme is used to frequency lock a commercially available DFB laser, reducing the laser frequency fluctuations by an order of magnitude compared to the free-running case.

Integrated electro-optical phase-locked loop for high resolution optical synthesis.

Electrical frequency synthesizers have been in existence for several decades and are an integral part of almost every communication and sensing system. In the optical domain, however, despite promising bench-top demonstration of frequency synthesizers, large size, high-power consumption, and high-cost have significantly limited their large deployment compared to their electrical counterparts. Here we report an integrated electro-optical phase locked loop (EOPLL) as the core of an optical synthesizer where photonic and electronic devices are integrated in a standard silicon-on-insulator (SOI) process. A sophisticated integrated electronic-photonic architecture is proposed enabling reliable, low-cost, and high resolution optical synthesis. The small on-chip optical delay and electronically assisted frequency detection and acquisition provide tunable phase and frequency locking. The integrated EOPLL consumes 28.5 mW with total chip area of 2.4 mm2 making it comparable with electrical synthesizers enabling large-scale deployment in applications such as low-cost optical spectroscopy, detection, sensing, and optical communication.

Self-equalizing photodiodes, a hybrid electro-optical approach to tackle bandwidth limitation in high-speed signaling.

In this paper we provide the design details of self-equalizing photodetectors which enable higher data rate transmission by improving the overall bandwidth of the bandwidth limited transmission link, through a hybrid electro-optical solution. Two different self-equalizing photodiodes, one having fixed equalization and the other being programmable are presented as proof of concept.

Nanophotonic projection system.

Low-power integrated projection technology can play a key role in development of low-cost mobile devices with built-in high-resolution projectors. Low-cost 3D imaging and holography systems are also among applications of such a technology. In this paper, an integrated projection system based on a two-dimensional optical phased array with fast beam steering capability is reported. Forward biased p-i-n phase modulators with 200MHz bandwidth are used per each array element for rapid phase control. An optimization algorithm is implemented to compensate for the phase dependent attenuation of the p-i-n modulators. Using rapid vector scanning technique, images were formed and recorded within a single snapshot of the IR camera.

Nanophotonic coherent imager.

An integrated silicon nanophotonic coherent imager (NCI), with a 4 × 4 array of coherent pixels is reported. In the proposed NCI, on-chip optical processing determines the intensity and depth of each point on the imaged object based on the instantaneous phase and amplitude of the optical wave incident on each pixel. The NCI operates based on a modified time-domain frequency modulated continuous wave (FMCW) ranging scheme, where concurrent time-domain measurements of both period and the zero-crossing time of each electrical output of the nanophotonic chip allows the NCI to overcome the traditional resolution limits of frequency domain detection. The detection of both intensity and relative delay enables applications such as high-resolution 3D reflective and transmissive imaging as well as index contrast imaging. We demonstrate 3D imaging with 15µm depth resolution and 50µm lateral resolution (limited by the pixel spacing) at up to 0.5-meter range. The reported NCI is also capable of detecting a 1% equivalent refractive index contrast at 1mm thickness.

Frequency tunable magnetostatic wave filters with zero static power magnetic biasing circuitry.

A single tunable filter simplifies complexity, reduces insertion loss, and minimizes size compared to frequency switchable filter banks commonly used for radio frequency (RF) band selection. Magnetostatic wave (MSW) filters stand out for their wide, continuous frequency tuning and high-quality factor. However, MSW filters employing electromagnets for tuning consume excessive power and space, unsuitable for consumer wireless applications. Here, we demonstrate miniature and high selectivity MSW tunable filters with zero static power consumption, occupying less than 2 cc. The center frequency is continuously tunable from 3.4 GHz to 11.1 GHz via current pulses of sub-millisecond duration applied to a small and nonvolatile magnetic bias assembly. This assembly is limited in the area over which it can achieve a large and uniform magnetic field, necessitating filters realized from small resonant cavities micromachined in thin films of Yttrium Iron Garnet. Filter insertion loss of 3.2 dB to 5.1 dB and out-of-band third order input intercept point greater than 41 dBm are achieved. The filter's broad frequency range, compact size, low insertion loss, high out-of-band linearity, and zero static power consumption are essential for protecting RF transceivers from interference, thus facilitating their use in mobile applications like IoT and 6 G networks.

A BIOCOMPATIBLE GLASS-ENCAPSULATED TRIAXIAL FORCE SENSOR FOR IMPLANTABLE TACTILE SENSING APPLICATIONS.

This paper reports a microfabricated triaxial capacitive force sensor. The sensor is fully encapsulated with inert and biocompatible glass (fused silica) material. The sensor comprises two glass plates, on which four capacitors are located. The sensor is intended for subdermal implantation in fingertips and palms and providing tactile sensing capabilities for patients with paralyzed hands. Additional electronic components, such as passives and IC chips, can also be integrated with the sensor in a hermetic glass package to achieve an implantable tactile sensing system. Through attachment to a human palm, the sensor has been shown to respond appropriately to typical hand actions, such as squeezing or picking up a bottle.

An integrated photonic-assisted phased array transmitter for direct fiber to mm-wave links.

Millimeter-wave (mm-wave) phased arrays can realize multi-Gb/s communication links but face challenges such as signal distribution and higher power consumption hindering their widespread deployment. Hybrid photonic mm-wave solutions combined with fiber-optics can address some of these bottlenecks. Here, we report an integrated photonic-assisted phased array transmitter applicable for low-power, compact radio heads in fiber to mm-wave fronthaul links. The transmitter utilizes optical heterodyning within an electronically controlled photonic network for mm-wave generation, beamforming, and steering. A photonic matrix phase adjustment architecture reduces the number of phase-shift elements from M × N to M + N lowering area and power requirements. A proof-of-concept 2 × 8 phased array transmitter is implemented that can operate from 24-29 GHz, has a steering range of 40°, and achieves 5 dBm EIRP at an optical power of 55 mW without using active mm-wave electronics. Data streams at 2.5 Gb/s are transmitted over 3.6 km of optical fiber and wirelessly transmitted attaining bit-error rates better than 10-11.

The next generation of hybrid microfluidic/integrated circuit chips: recent and upcoming advances in high-speed, high-throughput, and multifunctional lab-on-IC systems.

Since the field's inception, pioneers in microfluidics have made significant progress towards realizing complete lab-on-chip systems capable of sophisticated sample analysis and processing. One avenue towards this goal has been to join forces with the related field of microelectronics, using integrated circuits (ICs) to perform on-chip actuation and sensing. While early demonstrations focused on using microfluidic-IC hybrid chips to miniaturize benchtop instruments, steady advancements in the field have enabled a new generation of devices that expand past miniaturization into high-performance applications that would not be possible without IC hybrid integration. In this review, we identify recent examples of labs-on-chip that use high-resolution, high-speed, and multifunctional electronic and photonic chips to expand the capabilities of conventional sample analysis. We focus on three particularly active areas: a) high-throughput integrated flow cytometers; b) large-scale microelectrode arrays for stimulation and multimodal sensing of cells over a wide field of view; c) high-speed biosensors for studying molecules with high temporal resolution. We also discuss recent advancements in IC technology, including on-chip data processing techniques and lens-free optics based on integrated photonics, that are poised to further advance microfluidic-IC hybrid chips.

Highly stable integration of graphene Hall sensors on a microfluidic platform for magnetic sensing in whole blood.

The detection and analysis of rare cells in complex media such as blood is increasingly important in biomedical research and clinical diagnostics. Micro-Hall detectors (µHD) for magnetic detection in blood have previously demonstrated ultrahigh sensitivity to rare cells. This sensitivity originates from the minimal magnetic background in blood, obviating cumbersome and detrimental sample preparation. However, the translation of this technology to clinical applications has been limited by inherently low throughput (<1 mL/h), susceptibility to clogging, and incompatibility with commercial CMOS foundry processing. To help overcome these challenges, we have developed CMOS-compatible graphene Hall sensors for integration with PDMS microfluidics for magnetic sensing in blood. We demonstrate that these graphene µHDs can match the performance of the best published µHDs, can be passivated for robust use with whole blood, and can be integrated with microfluidics and sensing electronics for in-flow detection of magnetic beads. We show a proof-of-concept validation of our system on a silicon substrate and detect magnetic agarose beads, as a model for cells, demonstrating promise for future integration in clinical applications with a custom CMOS chip.

An implantable wireless tactile sensing system.

The sense of touch is critical to dexterous use of the hands and thus an essential component to efforts to restore hand function after amputation or paralysis. Prosthetic systems have focused on wearable tactile sensors. But wearable sensors are suboptimal for neuroprosthetic systems designed to reanimate a patient's own paralyzed hand. Here, we developed an implantable tactile sensing system intended for subdermal placement. The system is composed of a microfabricated capacitive force sensor, a custom integrated circuit supporting wireless powering and data transmission, and a laser-fused hermetic silica package. The miniature device was validated through simulations, benchtop testing, and ex vivo testing in a primate hand. The sensor implanted in the fingertip accurately measured skin forces with a resolution of 4.3 mN. The output from this novel sensor could be encoded in the brain with microstimulation to provide tactile feedback. More broadly, the materials, system design, and fabrication approach establish new foundational capabilities for various applications of implantable sensing systems.

An implantable, wireless, battery-free system for tactile pressure sensing.

The sense of touch is critical to dexterous use of the hands and thus an essential component of efforts to restore hand function after amputation or paralysis. Prosthetic systems have addressed this goal with wearable tactile sensors. However, such wearable sensors are suboptimal for neuroprosthetic systems designed to reanimate a patient's own paralyzed hand. Here, we developed an implantable tactile sensing system intended for subdermal placement. The system is composed of a microfabricated capacitive pressure sensor, a custom integrated circuit supporting wireless powering and data transmission, and a laser-fused hermetic silica package. The miniature device was validated through simulations, benchtop assessment, and testing in a primate hand. The sensor implanted in the fingertip accurately measured applied skin forces with a resolution of 4.3 mN. The output from this novel sensor could be encoded in the brain with microstimulation to provide tactile feedback. More broadly, the materials, system design, and fabrication approach establish new foundational capabilities for various applications of implantable sensing systems.

Wideband tunable laser phase noise reduction using single sideband modulation in an electro-optical feed-forward scheme.

A wideband laser phase noise reduction scheme is introduced where the optical field of a laser is single sideband modulated with an electrical signal containing the discriminated phase noise of the laser. The proof-of-concept experiments on a commercially available 1549 nm distributed feedback laser show linewidth reduction from 7.5 MHz to 1.8 kHz without using large optical cavity resonators. This feed-forward scheme performs wideband phase noise cancellation independent of the light source and, as such, it is compatible with the original laser source tunability without requiring tunable optical components. By placing the proposed phase noise reduction system after a commercial tunable laser, a tunable coherent light source with kilohertz linewidth over a tuning range of 1530-1570 nm is demonstrated.

Multi-dimensional data transmission using inverse-designed silicon photonics and microcombs.

The use of optical interconnects has burgeoned as a promising technology that can address the limits of data transfer for future high-performance silicon chips. Recent pushes to enhance optical communication have focused on developing wavelength-division multiplexing technology, and new dimensions of data transfer will be paramount to fulfill the ever-growing need for speed. Here we demonstrate an integrated multi-dimensional communication scheme that combines wavelength- and mode- multiplexing on a silicon photonic circuit. Using foundry-compatible photonic inverse design and spectrally flattened microcombs, we demonstrate a 1.12-Tb/s natively error-free data transmission throughout a silicon nanophotonic waveguide. Furthermore, we implement inverse-designed surface-normal couplers to enable multimode optical transmission between separate silicon chips throughout a multimode-matched fibre. All the inverse-designed devices comply with the process design rules for standard silicon photonic foundries. Our approach is inherently scalable to a multiplicative enhancement over the state of the art silicon photonic transmitters.

A 10.8 µW Neural Signal Recorder and Processor With Unsupervised Analog Classifier for Spike Sorting.

Implantable brain machine interfaces for treatment of neurological disorders require on-chip, real-time signal processing of action potentials (spikes). In this work, we present the first spike sorting SoC with integrated neural recording front-end and analog unsupervised classifier. The event-driven, low power spike sorter features a novel hardware-optimized, K-means based algorithm that effectively eliminates duplicate clusters and is implemented using a novel clockless and ADC-less analog architecture. The 1.4 mm2 chip is fabricated in a 180-nm CMOS SOI process. The analog front-end achieves a 3.3 µVrms noise floor over the spike bandwidth (400 - 5000 Hz) and consumes 6.42 µW from a 1.5 V supply. The analog spike sorter consumes 4.35 µW and achieves 93.2% classification accuracy on a widely used synthetic test dataset. In addition, higher than 93% agreement between the chip classification result and that of a standard spike sorting software is observed using pre-recorded real neural signals. Simulations of the implemented spike sorter show robust performance under process-voltage-temperature variations.