Integrated photonic convolution acceleration core for wearable devices

With the advancement of deep learning and neural networks,the computational demands for applications in wearable devices have grown exponentially.However,wearable devices also have strict requirements for long battery life,low power consumption,and compact size.In this work,we propose a scalable optoelectronic computing system based on an integrated optical convolution acceleration core.This system enables high-precision computation at the speed of light,achieving 7-bit accuracy while maintaining extremely low power consumption.It also demonstrates peak throughput of 3.2 TOPS(tera operations per second)in parallel processing.We have successfully demonstrated image convolution and the typical application of an interactive first-person perspective gesture recognition application based on depth information.The system achieves a comparable recognition accuracy to traditional electronic computation in all blind tests.

A review:Photonics devices,architectures,and algorithms for optical neural computing

The explosive growth of data and information has motivated various emerging non-von Neumann computational approaches in the More-than-Moore era.Photonics neuromorphic computing has attracted lots of attention due to the fascinating advantages such as high speed,wide bandwidth,and massive parallelism.Here,we offer a review on the optical neural computing in our research groups at the device and system levels.The photonics neuron and photonics synapse plasticity are presented.In addition,we introduce several optical neural computing architectures and algorithms including photonic spiking neural network,photonic convolutional neural network,photonic matrix computation,photonic reservoir computing,and photonic reinforcement learning.Finally,we summarize the major challenges faced by photonic neuromorphic computing,and propose promising solutions and perspectives.

Time-space multiplexed photonic-electronic digital multiplier

Optical computing has shown immense application prospects in the post-Moore era.However,as a crucial component of logic computing,the digital multiplier can only be realized on a small scale in optics,restrained by the limited functionalities and inevitable loss of optical nonlinearity.In this paper,we propose a time-space multiplexed architecture to realize large-scale photonic-electronic digital multiplication.

Integrated spatial-temporal random speckle spectrometer with high resolution in the C-band

The increasing demand for diverse portable high-precision spectral analysis applications has driven the rapid development of spectrometer miniaturization. However, the resolutions of existing miniaturized spectrometers mostly remain at the nanometer level, posing a challenge for further enhancement towards achieving picometerlevel precision. Here, we propose an integrated reconstructive spectrometer that utilizes Mach–Zehnder interferometers and a tunable diffraction network. Through random tuning in the time domain and disordered diffraction in the space domain, the random speckle patterns closely related to wavelength information are obtained to construct the transmission matrix. Experimentally, we achieve a high resolution of 100 pm and precisely reconstruct multiple narrowband and broadband spectra. Moreover, the proposed spectrometer features a simple structure, strong portability, and fast sampling speed, which has great potential in the practical application of high-precision portable spectral analysis.

Integrated contra‑directionally coupled chirped Bragg grating waveguide with a linear group delay spectrum

Due to the advantages of low propagation loss,wide operation bandwidth,continuous delay tuning,fast tuning speed,and compact footprints,chirped Bragg grating waveguide has great application potential in wideband phased array beamforming systems.However,the disadvantage of large group delay error hinders their practical applications.The nonlinear group delay spectrum is one of the main factors causing large group delay errors.To solve this problem,waveguides with nonlinear gradient widths are adopted in this study to compensate for the nonlinear efect of the grating apodization on the mode efective index.As a result,a linear group delay spectrum is obtained in the experiment,and the group delay error is halved.

Self-calibrating microring synapse with dual-wavelength synchronization

As a resonator-based optical hardware in analog optical computing, a microring synapse can be straightforwardly configured to simulate the connection weights between neurons, but it faces challenges in precision and stability due to cross talk and environmental perturbations. Here, we propose and demonstrate a self-calibration scheme with dual-wavelength synchronization to monitor and calibrate the synaptic weights without interrupting the computation tasks. We design and fabricate an integrated 4 × 4 microring synapse and deploy our self-calibration scheme to validate its effectiveness. The precision and robustness are evaluated in the experiments with favorable performance, achieving 2-bit precision improvement and excellent robustness to environmental temperature fluctuations(the weights can be corrected within 1 s after temperature changes 0.5°C). Moreover, we demonstrate matrix inversion tasks based on Newton iterations beyond 7-bit precision using this microring synapse. Our scheme provides an accurate and real-time weight calibration independently parallel from computations and opens up new perspectives for precision boost solutions to large-scale analog optical computing.

On-chip photonic spatial-temporal descrambler

As an indispensable part to compensate for the signal crosstalk in fiber communication systems,conventional digital multi-input multi-output(MIMO)signal processor is facing the challenges of high com-putational complexity,high power consumption and relatively low processing speed.The optical MIMOenables the best use of light and has been proposed to remedy this limitation.However,the currently existing optical MIMO methods are all restricted to the spatial di-mension,while the temporal dimension is neglected.Here,an on-chip spatial-temporal descrambler with four channels were devised and its MIMO functions were experimentally verified simultaneously in both spatial and temporal dimensions.The spatial crosstalk of single-channel descrambler and four-channel descrambler is respectively less than-21 dB and-18 dB,and the time delay is simultaneously com-pensated successfully.Moreover,a more universal model extended to mode-dependent loss and gain(MDL)compensation was further de-veloped,which is capable of being cascaded for the real optical trans-mission system.The first attempt at photonic spatial-temporal de-scrambler enriched the varieties of optical MIMO,and the proposed scheme provided a new opportunity for all-optical MIMO signal pro-cessing.

On-chip programmable pulse processor employing cascaded MZI-MRR structure

Optical pulse processor meets the urgent demand for high-speed,ultra wideband devices,which can avoid electrical confinements in various fields,e.g.,alloptical communication,optical computing technology,cohcrcnt control and microwave fields.To date,great efforts have been madc particularly in on-chip programmable pulse proccessing.Here,we experimentally demonstrate a programmable pulse proceesor employing 16cascaded Mach-Zehnder interferometer coupled microring resonator(MZI-MRR)structure based on silicon-oninsulator wafer.With micro-heaters loaded to the device,both amplitude and frequency tunings can be realized in each MZI-MRR unit.Thanks to its reconfigurability and integration ability,First,it can serve as a fractional differentiator whose tuning range is 0.51-2.23 with deviation no more than 7%.Second,the device can be tuned into a programmable optical filter whose bandwidth varies from 0.15to 0.97nm.The optical filter is also shape tunable.Especially,15-channel wavelength selective switches are generated.

Photonic matrix multiplication lights up photonicaccelerator and beyond

Matrix computation,as a fundamental building block of information processing in science and technology,contributes most of the computational overheads in modern signal processing and artificial intelligence algorithms.Photonic accelerators are designed to accelerate specific categories of computing in the optical domain,especially matrix multiplication,to address the growing demand for computing resources and capacity.Photonic matrix multiplication has much potential to expand the domain of telecommunication,and artificial intelligence benefiting from its superior performance.Recent research in photonic matrix multiplication has flourished and may provide opportunities to develop applications that are unachievable at present by conventional electronic processors.In this review,we first introduce the methods of photonic matrix multiplication,mainly including the plane light conversion method,Mach–Zehnder interferometer method and wavelength division multiplexing method.We also summarize the developmental milestones of photonic matrix multiplication and the related applications.Then,we review their detailed advances in applications to optical signal processing and artificial neural networks in recent years.Finally,we comment on the challenges and perspectives of photonic matrix multiplication and photonic acceleration.

Performance of integrated optical switches based on 2D materials and beyond

Applications of optical switches,such as signal routing and data-intensive computing,are critical in optical interconnects and optical computing.Integrated optical switches enabled by two-dimensional(2D)materials and beyond,such as graphene and black phosphorus,have demonstrated many advantages in terms of speed and energy consumption compared to their conventional silicon-based counterparts.Here we review the state-of-the-art of optical switches enabled by 2D materials and beyond and organize them into several tables.The performance tables and future projections show the frontiers of optical switches fabricated from 2D materials and beyond,providing researchers with an overview of this field and enabling them to identify existing challenges and predict promising research directions.

Tunable optical delay line based on integrated grating-assisted contradirectional couplers

Tunable optical delay lines are one of the key building blocks in optical communication and microwave systems.In this work, tunable optical delay lines based on integrated grating-assisted contradirectional couplers are proposed and experimentally demonstrated. The device performance is comprehensively improved in terms of parameter optimization, apodization analysis, and electrode design. Tunable group delay lines of 50 ps at different wavelengths within the bandwidth of 12 nm are realized with a grating length of 1.8 mm. Under thermal tuning mode, the actual delay tuning range is around 20 ps at 7.2 V voltage. At last, a new scheme adopting an ultra-compact reflector for doubling group delay is proposed and verified, achieving a large group delay line of 400 ps and a large dispersion value up to 5.5 × 10~6 ps∕(nm · km) within bandwidth of 12 nm. Under thermal tuning mode, the actual delay tuning range is around 100 ps at 8 V voltage.

Compact and broadband multimode waveguide bend by shape-optimizing with transformation optics

Multimode waveguide bend is one of the key components for realizing high-density mode-division multiplexing systems on chip.However,the reported multimode waveguide bends are either large,bandwidth-limited or fabrication-complicated,which hinders their applications in future high-density multimode photonic circuits.Here we propose a compact multimode waveguide bend supporting four TE modes simply by shape-optimizing with transformation optics.The shape of the waveguide is optimized in the virtual space with gradient distribution of the refractive index,so that the scattering loss and intermode cross talk are well suppressed.After conformal mapping back into the physical space,a compact(effective radius of 17μm)multimode bending waveguide is obtained.Simulations show that the proposed multimode waveguide bend has little loss(<0.1 dB)and low cross talk(<−20 dB)throughout an ultrabroad wavelength range of 1.16–1.66μm.We also fabricated the shape-optimized multimode bending waveguide on a silicon-on-insulator wafer.At 1550 nm wavelength,the measured excess losses for the four lowest-order TE modes are less than 0.6 dB,and the intermode cross talks are all below−17 dB.Our study paves the way for realizing high-density and large-scale multimode integrated optical circuits for optical interconnect.

Broadband on-chip integrator based on silicon photonic phase-shifted Bragg grating

All-optical integrators are key devices for the realization of ultra-fast passive photonic networks, and, despite their broad applicability range(e.g., photonic bit counting, optical memory units, analogue computing, etc.), their realization in an integrated form is still a challenge. In this work, an all-optical integrator based on a silicon photonic phase-shifted Bragg grating is proposed and experimentally demonstrated, which shows a wide operation bandwidth of 750 GHz and integration time window of 9 ps. The integral operation for single pulse, inphase pulses, and π-shifted pulses with different delays has been successfully achieved.

A small microring array that performs large complex-valued matrix-vector multiplication

As an important computing operation,photonic matrix-vector multiplication is widely used in photonic neutral networks and signal processing.However,conventional incoherent matrix-vector multiplication focuses on real-valued operations,which cannot work well in complex-valued neural networks and discrete Fourier transform.In this paper,we propose a systematic solution to extend the matrix computation of microring arrays from the real-valued field to the complex-valued field,and from small-scale(i.e.,4×4)to large-scale matrix computation(i.e.,16×16).Combining matrix decomposition and matrix partition,our photonic complex matrix-vector multiplier chip can support arbitrary large-scale and complex-valued matrix computation.We further demonstrate Walsh-Hardmard transform,discrete cosine transform,discrete Fourier transform,and image convolutional processing.Our scheme provides a path towards breaking the limits of complex-valued computing accelerator in conventional incoherent optical architecture.More importantly,our results reveal that an integrated photonic platform is of huge potential for large-scale,complex-valued,artificial intelligence computing and signal processing.

Wideband adaptive microwave frequency identification using an integrated silicon photonic scanning filter

Photonic-assisted microwave frequency identification with distinct features, including wide frequency coverage and fast tunability, has been conceived as a key technique for applications such as cognitive radio and dynamic spectrum access. The implementations based on compact integrated photonic chips have exhibited distinct advantages in footprint miniaturization, light weight, and low power consumption, in stark contrast with discrete optical-fiber-based realization. However, reported chip-based instantaneous frequency measurements can only operate at a single-tone input, which stringently limits their practical applications that require wideband identification capability in modern RF and microwave applications. In this article, we demonstrate, for the first time, a wideband, adaptive microwave frequency identification solution based on a silicon photonic integrated chip,enabling the identification of different types of microwave signals from 1 to 30 GHz, including single-frequency,multiple-frequency, chirped-frequency, and frequency-hopping microwave signals, and even their combinations.The key component is a high Q-factor scanning filter based on a silicon microring resonator, which is used to implement frequency-to-time mapping. This demonstration opens the door to a monolithic silicon platform that makes possible a wideband, adaptive, and high-speed signal identification subsystem with a high resolution and a low size, weight, and power(SWaP) for mobile and avionic applications.

Anti-parity-time symmetry enabled on-chip chiral polarizer

Encircling an exceptional point(EP) in a parity-time(PT) symmetric system has shown great potential for chiral optical devices,such as chiral mode switching for symmetric and antisymmetric modes.However,to our best knowledge,chiral switching for polarization states has never been reported,although chiral polarization manipulation has significant applications in imaging,sensing,communication,etc.Here,inspired by the anti-PT symmetry,we demonstrate,for the first time to our best knowledge,an on-chip chiral polarizer by constructing a polarization-coupled anti-PT symmetric system.The transmission axes of the chiral polarizer are different for forward and backward propagation.A polarization extinction ratio of over 10 dB is achieved for both propagating directions.Moreover,a telecommunication experiment is performed to demonstrate the potential applications in polarization encoding signals.It provides a novel functionality for encircling-an-EP parametric evolution and offers a new approach for on-chip chiral polarization manipulation.

Large-range tunable fractional-order differentiator based on cascaded microring resonators

In this paper, we experimentally demonstrate an all-optical continuously tunable fractional-order differentiator using on-chip cascaded electrically tuned microring resonators （MRRs）. By changing the voltage applied on a MRR, the phase shift at the resonance frequency of the MRR varies, which can be used to implement tunable fractional-order differentiator. Hence fractional-order differentiator with a larger ttmable range can be obtained by cascading more MRR units on a single chip. In the experiment, we applied two direct current voltage sources on two cascaded MRRs respectively, and a tunable order range of 0.57 to 2 have been demonstrated with Gaussian pulse injection, which is the largest tuning range to our knowledge.

Metasurface for highly-efficient on-chip classical and quantum all-optical modulation

Metasurface made of artificially two-dimensional structured subwavelength-scaled nanostructures gives rise to unprecedented efficient way to realize on-chip all-optical modulation,in both classical regime and quantum regime.

All-optical PtSe2 silicon photonic modulator with ultra-high stability

All-optical modulation based on the photothermal effect of two-dimensional(2 D)materials shows great promise for all-optical signal processing and communication.In this work,an all-optical modulator with a 2 D Pt Se2-onsilicon structure based on a microring resonator is proposed and demonstrated utilizing the photothermal effect of Pt Se2.A tuning efficiency of 0.0040 nm·m W-1 is achieved,and the 10%–90%rise and decay times are 304μs and 284μs,respectively.The fabricated device exhibits a long-term air stability of more than 3 months.The experimental results prove that 2 D Pt Se2 has great potential for optical modulation on a silicon photonic platform.

The smallest nanowire spectrometers

Optical spectroscopy is a versatile characterization technique for a wide range of applications.Developing miniaturized spectrometers is the trend for applications in which small footprint takes precedence over high resolution.However,development of micro-spectrometers based on miniaturized or integrated optics is approaching a bottleneck toward submillimeter scales because of the inherent scale limitation of their optical components or path lengths.Although these constraints can be circumvented with computational spectral reconstruction by addressing a full range of spectral components simultaneously at multiple detectors,complex millimeter-scale arrays of individually prepared filters arranged over charge-coupled device or complementary metal-oxide semiconductor detectors are difficult to be miniaturized.