Optoelectronic Devices and Integration XIII

This presentation introduces a novel polishing methodology centered around robotic systems. Specifically, deterministic polishing technologies like computer controlled optical surfacing (CCOS) and magnetorheological finishing (MRF) are synergistically integrated with robotic platforms to realize a versatile and economically viable multi-axis polishing apparatus. The presentation delves into the advantages and disadvantages inherent in Robot CCOS&MRF systems, elucidating various compensation techniques designed to enhance the precision of the robotic polishing process. Empirical evidence underscores the capability of robot-based CCOS & MRF systems to attain nanometer-level accuracy in the fabrication of mid-sized aspherical or freeform optics, all while operating within a cost-effective framework.

The 14-meter off-axis deployable space telescope, Single Aperture Large Telescope for Universe Studies (SALTUS), is designed to serve as an exceptionally large far-infrared observatory in space. SALTUS aims to observe thousands of faint astrophysical targets, including the earliest galaxies, protoplanetary disks in different stages of evolution, and various solar system objects. Its architecture incorporates a radiatively cooled, unobscured 14-meter aperture and cryogenic instruments, enabling both high spectral and spatial resolution with unprecedented sensitivity across a wavelength range that is largely inaccessible to current ground-based or space observatories. The innovative SALTUS optical system, featuring a large inflatable off-axis primary mirror, offers exceptional sensitivity, angular resolution, and imaging performance at far-infrared wavelengths over a wide ±0.02° × 0.02° field of view. SALTUS' compact design allows it to fit within existing launch fairings and be easily deployed in orbit.

Non-destructive techniques are critical for understanding the body’s structure and function, for both scientific discovery and development of diagnostic procedures. Optical coherence tomography (OCT) is outstanding for cross-sectional mapping of the microstructure of tissue up to 2 mm in depth. Variants of OCT can show blood flow, quantify tissue stiffness, and visualize movement of cellular structures. I will discuss two examples of how OCT has positively impacted medicine. First, monitoring tumor growth in mouse models of colon cancer is critical for understanding the value of chemopreventive and therapeutic compounds. Miniature endoscopes allow alterations in crypt structure and tumor size to be visualized. Second, the understudied fallopian tube is a significant imaging challenge, but careful ex vivo studies and development of flexible sub-mm diameter endoscopes allow elucidation of the complex microstructure of the FT including the coordinated movement of cilia. These insights can be used for understanding infertility and the development of ovarian cancer.

Harnessing integrated quantum photonic circuits for the generation, manipulation, and detection of quantum states of light offers a pathway towards implementing advanced quantum technologies for applications spanning quantum computation, quantum simulation, and quantum communication. In this presentation, we will show recent advancements in the large-scale integrated photonic devices and circuits for quantum computation, networking, and simulations. Our discussion will delve into the on-chip processes involved in creating, manipulating, and measuring complex entanglement structures, such as multidimensional and multiphoton entanglement. We will discuss quantum computing and networking with integrated photonics, highlighting several aspects of chip-based entanglement distribution, teleportation, Gaussian Boson sampling, qudit-based and cluster-state quantum computations. We will also discuss photonic simulations of topological physics on a lattice of coupled microresonators, by which we have observed topologically-protected entanglement and fast topological non-Hermitian phase transitions, and we implemented a fully programmable photonic insulator within this framework.

Optical nanoprobes, designed to emit or collect light in the close proximity of a sample, have been extensively used to sense and image at nanometer resolution. However, the available nanoprobes, constructed from artificial materials, are incompatible and invasive when interfacing with biological systems. In this talk, I will present a fully biocompatible nanoprobe for subwavelength probing of localized fluorescence from single-cells in human blood. The nanoprobe is built on a tapered fiber tip apex by optical trapping of a living sphere-shaped cell and a chain of rod-shaped living cells, which act as a high-aspect-ratio nanoprobe. Light propagating along the nanoprobe can be focused into a spot with a full width at half maximum of around 200 nm on the surface of single cells. Fluorescence signals are detected in real time at subwavelength spatial resolution.

An all-polymer spring optomechanical microresonator based ultrasonic sensor is presented. The microresonator is an all-polymer structure with a micro-disk and a cylindrical waveguide based microcavity. Experiment results show that its sensitivity is 987.8 times that of the commercial acoustic sensor at 200 KHz, and its acoustic frequency response range is within 50 KHz-400 KHz. It shows the advantages of smart structure, high sensitivity, and wideband response, and it is expected to have wide applications in smart grid, biomedicine and medical ultrasound imaging.

We introduce an on-chip integrated silicon nitride grating waveguide structure, leveraging Bragg scattered dispersion to achieve local phase-matching of dispersive wave and soliton pulse. This advanced grating structure facilitates precise tuning, thereby enabling localized supercontinuum power enhancement across a designated spectral range.

The operation of the device is affected by the increase in temperature caused by the rough electrodes surface of the thin-film lithium niobate modulator. We have developed simulation models to analyze the relationship between roughness and maximum surface temperature of gold electrodes and to examine the influence of snowball radius and surface area ratio on the maximum surface temperature. The results indicate that both surface area ratio and snowball radius affect maximum surface temperature change, with the influence of surface area ratio being greater than that of snowball radius within the range of 2-6nm and a surface area ratio between 1.6-2.5, respectively.

Silicon-based platform is highly competitive in the development of next generation on-chip sensing systems for chemical, biological, and medical sensing. We demonstrate an on-chip refractive index sensor simultaneously possessing high sensitivity, wide detectable range, and low limit of detection through monitoring the envelope spectrum of a subwavelength grating microring resonator. In addition, the device shows low temperature sensitivity. This work provides a strong candidate for the key component in lab-on-chip sensing systems.

Fiber-based distributed acoustic sensing (DAS) technique has recently captured great attention of academia and industry, showing wide applications in wide application. However, there still exists a key challenge to achieve DAS system equipped with long distance, high sensitivity, full distribution and wide bandwidth simultaneously. The traditional DAS system is based on standard optical fiber, but the Rayleigh scattering of the fiber is intrinsically weak that tends to make acoustic measurements inaccurate and unreliable with deteriorate signal-to-noise ratio (SNR) and sensitivity. In this talk, we will present our recent work about acoustic sensitivity enhancement in DAS system.

presentation title: Ultra-high resolution optical vector analysis based on optical double-sideband modulation author: WANG Wenyu, Zhou Jing, Yu Changyuan abstract: Optical devices, which are the cornerstone of a new generation of optical information systems, require multi-dimensional measurements. This article reviews the development and application of microwave photonics, and details the current main methods, advanced structures, and future development directions for building optical vector analysis technology. The most promising ultra-high-resolution optical vector analysis technology based on optical double-sideband modulation is analysed from the aspects of principle, classification, system simulation, etc. It also adopts measurement range broadening technology based on optical frequency comb, measurement error elimination technology based on carrier suppression and balanced photoelectric detection, and ultrafast measurement technology based on LFM to improve its performance. This solution can quickly obtain the spectral response of optical devices during measurement and discover specific problems that arise in design and manufacturing.

We measure electro-optic coefficient with highly accurate transmission Teng and Man method to eliminate measurement error due to multi-reflection inside an active material.

Three-axis measurements are very popular because they allow you to measure the position of the monitored object with a single instrument. Autocollimators as such devices allow to achieve high accuracy of measurements, which is in demand when controlling deformations or deflections of various roofing structures, bridge beams, overpasses, as well as for precise positioning of cutting or welding tools. Cube-corner reflectors with two cylindrical surfaces allow autocollimation systems to make independent measurements of three angles of rotation of the object at an increased distance. The focus of this work is to describe an algorithm that allows the angles of rotation of a controlled object to be determined from the angles of rotation of lines in the image of the described reflector.

Given the wide coverage of telecom fiber networks, the goal of turning optical fiber network as a sensor is proposed by fully integrating optical fiber sensing and communication in one system. In this talk, we will review our recent work on IOSAC. We will discuss the key parameters that influence the performance of IOSAC, demonstrate different joint designs of integrated optical communication and distributed optical fiber vibration sensing including forward sensing and Rayleigh backscattering based sensing and introduce the prospective applications of IOSAC.

Optical phased arrays can offer non-mechanical beam steering capabilities for light detection and ranging (LiDAR), wireless optical communications, and imaging. Chip-scale silicon optical phased arrays have made significant advances in their key performance parameters, such as field of view, sidelobe suppression ratio, main beam energy ratio, and beam steering speed. We will present key scientific insights and technical innovations that lead to these advances: how crosstalk-suppression in half-wavelength pitch waveguide arrays enables near 180 degrees field of view and high main beam energy ratio; how coordinating all phase shifters helps achieve fast beam steering; and how neural networks can help suppress sidelobes. These advances open up new opportunities such as independent multi-beam forming, and mobile wireless optical communications. We will analyze some simple trade-offs that can be made between some parameters and their practicality in real applications. Holistic advances through novel approaches are needed to look beyond such simple trade-offs.

We demonstrate waveguide Modified Uni-Traveling-Carrier (MUTC) Photodetectors (PDs) to complete optical-to-microwave conversion. Ultra-high 3-dB bandwidths of 137 GHz and 163 GHz, high responsivities of 0.32 A/W and 0.24 A/W, and low dark currents of 3 nA and 5 nA have been achieved for 5×12 µm2 and 5×8 µm2 devices, respectively. These PDs demonstrate promising performance for photodetection of high-repetition-rate optical pulses, laying the foundation for the photonic chip-based ultra-stable and low-phase noise microwave generation.

In order to address the challenges brought by high bandwidth requirements to optical communication networks and improve the capacity of ultra wideband wavelength division multiplexing systems, we conducted a study entitled "Power control of ultra wideband optical communication system based on optimization algorithm". An optical power optimization algorithm based on the marine predator algorithm is used to optimize the launched power to achieve the goal of maximizing channel capacity. At the same time, the maximum capacity strategy, high and flat capacity strategy, and flattest capacity strategy are applied to meet different transmission requirements, achieving a balance between capacity and ripple. The simulation results show that compared with the brute force search method, the marine predator algorithm based optimization algorithm significantly reduces the optimization time required and achieves advantages in both total capacity and ripple flatness. This method can be used to quickly optimize optical network design before actual deployment. Author list: Zhu Xingyu, Zhou Jing, Yu Changyuan

The cascaded Mach–Zehnder interferometers based on planar lightwave circuit are used to design and fabricate a low-crosstalk and flat-top 2-channel coarse wavelength-division multiplexing (de)multiplexer, which provides flat pass-bands with wide bandwidth of 40 nm and low crosstalk of -14 dB. It also has low insertion loss and low polarization dependent loss.

This paper presents a novel neural network (NN)-based equalization method for high-speed PAM8 silicon photonics intensity modulation and direct detection (IM/DD) systems. The proposed NN equalizer utilizes an activation function with eight saturation regions, significantly enhancing the performance of short-range optical communication systems. Experiments demonstrate successful transmission of 60 GBaud PAM8 signals with a bit error rate (BER) below the high-definition forward error correction (HD-FEC) threshold and 70 GBaud PAM8 signals with BER below the soft-decision forward error correction (SD-FEC) threshold. The results indicate that the innovative NN-based equalization offers up to three orders of magnitude improvement in BER compared to traditional activation functions, showcasing its potential for next-generation optical communication systems.

This work proposes an integrated, low-cost, low-loss spot size converter (SSC) for thin film lithium niobate (TFLN) . The proposed SSC uses a Si3N4 tapered waveguide and an amorphous silicon layer to achieve 88% coupling efficiency with standard lithography, reducing end-face coupling losses and improving integration with semiconductor lasers.

Perovskite photodetectors with regulated band alignment engineering by bridging molecules Xiangyu Zhou,Qiao He,Yixuan Huang,Jiang Wu With the rapid development of Internet of Things (IoT) and artificial intelligence (AI) technologies, the field of information technology is gradually moving towards the era of optical information processing. Photodetectors convert optical signals into electrical signals, and due to their superior optoelectronic properties, perovskite has become a promising candidate for the next generation of high-speed photodetectors. However, achieving high device efficiency requires the manufacture of perovskite thin films with fewer defects and high carrier extraction capability. Here, we propose that by finely controlling the transport layer and perovskite layer, favorable band bending has been achieved, series resistance has been reduced, and carrier separation has been enhanced. Therefore, the integrated 2Br-POOH-TFHV co processing method achieved a dark current of less than 1X10-9 a and a response speed of less than 30 ns. In summary, this study proposes a new method for manufacturing high-speed photodiodes.

A hybrid Bloch surface polarized waveguide comprising a dielectric nanowire incorporated at the top of a photonic crystal slab with a sandwiched dielectric arc groove is proposed, and the guiding properties of which are theoretically investigated at the telecommunication wavelength. The introduction of the arc groove offers additional freedom for tuning the supported hybrid mode. By properly choosing the geometrical parameters of the structure, the modified hybrid waveguide could feature significantly enhanced field confinement ability combined with long-range propagation length. The proposed waveguide is shown to be able to enable obviously improved optical guiding performance over the conventional hybrid Bloch surface waveguide due to the tight hybridization of the dielectric mode and the Bloch surface mode in the groove gap. The structure being studied is expected to open up a new route for high-performance photonic integrated devices.

Metasurface has the characteristics of small volume and high integration, which can realize the simultaneous regulation of various degrees of freedom such as phase, polarization and amplitude. The dimensional characteristics of its subwavelength structure have become a hot spot in the optical field. Therefore, the combination of the metasurface and the traditional optical system imaging and polarization imaging can obtain the intensity and polarization information of the target, but also ensure the miniaturization and high performance of the new imaging optical system.

In this work, a novel solid-core anti-resonant fiber with core diameter of 50 μm was proposed to propagate single LP11a mode with low confinement loss and high mode purity. By optimizing the structure parameters of the single-layer nonuniformly distributed anti-resonant elements, LP01 mode and other high-order modes suffer seriously core mode leakage while the LP11a mode is well confined in the fiber core. The simulated results show that in 1500-1600 nm band, the confinement loss of LP11a mode is about 0.04 dB/m and that of the other modes is greater than 10 dB/m. The corresponding mode suppression ratio, which is defined as the loss ratio between the mode with lowest confinement loss and the LP11a mode, is larger than 370 in the whole band.

The significant difference of mode field size between thin film lithium niobate waveguides and standard single-mode fibers (SMF) leads to high coupling losses. Reducing coupling losses is an urgent problem that needs to be addressed in the field of thin film lithium niobate devices.

Electrochromic devices (ECDs) are emerging as a new technology for a variety of applications such as commercially available smart skylights, eyeglasses and self-adjusting mirrors. They can dynamically control the transmission and reflection of solar radiation from buildings and vehicles. We used Poly(3,4-ethylenedioxythiophene) (PEDOT) prepared by roll-to-roll in-situ polymerization as an ion storage layer and electrode to prepare 3-layer structured electrochromic devices, which can achieve fast switching with a coloring response time and recovery time of 3.6 s and 4 s, respectively, and a transmittance modulation (ΔT) of 43.5% (@ 8V and -0.6 V). Due to the high conductivity PEDOT (~2500 S/cm) surface charge capacity (insertion/exhaustion) balance, the ECDs show excellent optical properties and cycling stability (100,000 cycles ΔT <1%). The solution and low-temperature processing capabilities based on all-organic ECDs will have a significant impact on the industrialization of large-area smart windows.

Optical upconverters that convert near-infrared (NIR) light into visible light by directly connecting organic photovoltaic (OPV) devices and organic light-emitting diodes (OLEDs) in series are attracting attention for their potential applications in medicine, night vision, security, and detection. The use of organic system materials eliminates the need for NIR Si integrated circuits and removes limitations on substrate types. The high performance of organic compounds, combined with the manufacturing flexibility of OLEDs, simplifies the production process and reduces costs. The color of the upconversion electroluminescence spectrum can vary across the visible region, from blue to red. Our device achieved a maximum luminance greater than 100 cd/m² of green light at 8.5 V when illuminated by an 840 nm NIR light (1 mW/cm²), and the device reaches an infrared-to-visible upconversion efficiency of 7.23%.

We have designed a spot size converter (SSC) utilizing a hybrid SiN-LN structure, optimized for operation at a wavelength of 780 nm. Simulation results demonstrate the coupling losses of our SSC are 0.41 dB/facet and 0.55 dB/facet, for TE and TM mode, respectively. The entire structure can be fabricated with two steps of photolithography and etching. This structure is expected to enhance the edge coupling performance of thin-film lithium niobate photonic integrated circuits at short wavelengths.

A packaging scheme for optical transmission modules based on fourth-order pulse amplitude modulation(PAM4)technology with a data transmission rate of up to 200Gbit/s is proposed to meet the design requirements of 200Gbit/s PAM4 optical transceiver modules. Package integrates four PAM4 electrical/optical conversion channels internally,with a single channel data transmission rate of 50Gbit/s. This article introduces the composition and technical difficulties of the 200Gbit/s PAM4 optical emission module, and then models, simulates, and optimizes the 50Gbit/s data transmission channel. Finally, the sample testing is completed. The test results show that the single channel PAM4 data rate of the sample can reach 50Gbit/s, and the overall PAM4 data rate can reach 200Gbit/s, meeting the design requirements of the optical transceiver module.

Flexible polymer optical waveguides demonstrate outstanding advantages over those on rigid substrates, including flexibility, bendability, lightweight, versatility, and ease of integration. However, flexible polymer optical waveguides face significant challenges in achieving efficient connection and integration with optical fibers. In this study, we propose a simple and reliable method for fabricating flexible polymer optical waveguides on polyimide (PI) using an ultraviolet (UV) lithography process. Additionally, MT connectors are fabricated at each end of the flexible waveguide with passive alignment features. This integration facilitates pluggable and efficient connections between waveguides and MT fiber arrays, significantly improving the ease of assembly and the reliability of optical connections. This integrated connector not only enhances the flexibility and reliability of applications in the field of optical interconnects but also lays the foundation for the development of future high-performance optoelectronic systems.

We proposed the simultaneous wavelength extensions of single-cavity dual-wavelength pulses by using a single erbium-doped fiber amplifier and one section of high nonlinear fiber. By additionally introducing a polarization dependent isolator, polarization dependent loss based gain profile tuning is adopted to obtain dual-wavelength pulses centered at 1533.7 and 1554.8 nm. When both dual-wavelength pulses are simultaneously launched into the same erbium-doped fiber amplifier with the bidirectional pump power, the spectral range of the output pulses could be expanded to be from 1500 to 1600 nm. Subsequently, the amplified dual-wavelength pulses further pass through a section of high nonlinear fiber, extending the spectrum from 1.2 um to more than 1.75 um. Dual-wavelength pulses are amplified simultaneously by using only one amplifier, showing the feasibility of the simplification of single-cavity dual-comb pulse amplification. These results show the high potential in the applications such as multi-color laser generation and spectroscopy.

In order to solve a series of application problems caused by the difficulty of adjusting prism collimation and the short working distance of small period gratings in current collimation solutions, this paper proposes a six-degree-of-freedom measurement scheme that uses a combination of large period gratings and long focus plano-convex lenses. This solution can achieve six degrees of freedom measurement at a working distance of more than 80mm. The main principle is that the displacement is calculated using the phase information formed by the interference of the four diffracted lights of the two-dimensional grating. The angle is based on the principle of autocollimation. According to the propagation direction of the diffracted light caused by the change of the grating diffracted light with the grating angle changes, thereby changing the angle obtained based on the location changes in the detected diffraction spot, ultimately form a measurement of six degrees of freedom.

Utilizing laser interference lithography, this study presents an efficient method for fabricating wire grid polarizers with high precision. The technique involves pattern generation using a helium-cadmium laser and subsequent transfer to a silicon substrate, resulting in gratings with an extinction ratio of 40 dB. This approach offers a promising solution for cost-effective and high-quality polarizer production.

In this paper, we propose an alignment method for sub-area multiple exposures that utilizes the substrate profile to mitigate the effects of alignment errors between the mask and the substrate.

In this manuscript, a novel scheme for generating a frequency-octupled triangular waveform based on an external modulator and fiber grating is proposed. The octave frequency modulation is realized by dual-parallel Mach-Zehnder modulator (DP-MZM), and two main optical sidebands (±4th) are generated in the spectrum. Then, the intensity modulator (IM) is used to modulate the optical double sideband (ODSB), and four sideband signals (±4th and ±12th) are obtained. The four spectral lines are separated inside and outside by an optical circulator and fiber grating. After polarization control, the power ratio of ±4 and ±12 sidebands can be tuned to 9.5 dB, and the expression of light intensity is approximately equal to the Fourier expansion of the ideal triangular wave. By tuning the driving frequency of local oscillator, continuous tunability of the repetition rate (2.5, 5, 7.5 and 10 GHz) is obtained.

The beam steering of phased array radars is achieved through the complex weighting of signals radiated by multiple antennas, with the phase of each antenna element channel controlled to direct the beam towards a specific target direction. Traditional phased array antennas utilize phase shifters for beam scanning; however, for wideband signals, this can result in beam squint phenomena. To compensate for this, true time delay (TTD) technology can be employed for phase correction. Optical fibers offer inherent advantages such as lightweight, broad bandwidth, low loss, and immunity to electromagnetic interference, making the implementation of optical TTD through microwave photonic technology an extremely promising solution. Therefore, we propose a tunable Mach-Zehnder delay interferometer (MZDI) structure that compensates for spatial phase differences within a certain bandwidth and scanning angle, enabling squint-free beam imaging.

We propose an optical transformer model based on the MZI cascaded mesh architecture. By constructing the self-attention layer using MZI cascaded mesh, the model achieved 96.12% accuracy on the MNIST dataset, demonstrating the effectiveness and potential of implementing large transformer models using optical architecture.

Requirements for continuous improvement in performance, increasing miniaturization and density of systems are the main driver in the semiconductor industry to improve the structure of chips and develop new packaging principles. Installations that install semiconductor crystals are necessarily equipped with technical vision systems. These systems include two or more industrial video cameras. The cameras are connected to a powerful computer. Computer software recognizes the position of the crystals in the storage system, monitors and corrects the error in the position of the crystal during transfer and flip operations, recognizes the mounting point of the crystal on the substrate and determines the correctness of the operation in order to identify defects. The article discusses a technical vision system for an automatic installation of semiconductor crystals. The requirements for cameras are justified and calculated based on the required chip positioning accuracy. A description of the applied technical vision algorithms and the final software product is given.

This paper addresses the challenges of signal loss and attenuation in high-frequency LVDS (Low-Voltage Differential Signaling) transmission for CMOS image sensors. It establishes a model for LVDS drivers and transmission lines, detailing attenuation mechanisms and their impact on differential signal transmission. Focusing on matching LVDS driver design with transmission line characteristics, the study analyzes interactions between MOS transistor operations and key variables of the RLGC transmission line model. Simulation experiments using 1 Gbps LVDS data over a 20 cm flexible ribbon cable validate the effectiveness of the matching method, providing technical support for reliable high-speed LVDS driver design and valuable references for selecting LVDS high-frequency transmission lines.

Low light detectors refer to photodetectors that perform imaging in lunar and lower illumination environments. The main types of detectors include electrical gain type detectors represented by EMCCD, photoelectric gain type detectors represented by Microchannel Panel (MCP) ICCD, and the latest scientific level CMOS image sensor (sCMOS). Enhanced low light detectors are more suitable for target detection and edge bright spot detection applications. Extremely dim targets can exceed the noise threshold with extremely high gain, so that the target will not be submerged in noise. Scientific grade CMOS image sensors are more suitable for array low light imaging applications, with better imaging quality than enhanced low light detectors. This article calculates the MTF based on the optical and electrical parameters of the low light detector. The comprehensive MTF at the Nyquist frequency of the three low light detectors are 0.48, 0.13, and 0.63, respectively.

Multi-degree-of-freedom grating interferometers play a crucial role in high-precision fields such as lithography machines and astronomical telescope mirror alignment. The phase calculation of interference signals is an important component. Current algorithms hard to balance real-time performance and accuracy effectively. Therefore, to enhance both the precision and real-time performance of grating interferometers, this paper proposes using an extended Kalman filter (EKF) method to solve the phase difference. In the EKF model, the state variables are set as the phase, frequency, and amplitude of the sinusoidal signal, and the calculation is performed through prediction and correction steps. Since the Kalman filter algorithm only uses the current sampling point data for model calculation, it has lower latency. The algorithm was deployed on an FPGA to test the signals generated by a signal generator, achieving a measurement accuracy of 0.03° and a resolution of 0.01°. This research contributes to improving the real-time performance and accuracy of grating interferometers.

In this study, we explore the application of spectral analysis techniques in software development, focusing on overcoming the technical limitations of existing platforms. We present the development of an integrated spectral analysis platform with enhanced data processing capabilities, algorithmic intelligence, and user interaction. Key innovations include a modern peak-seeking algorithm optimized for spectral data analysis and a robust system architecture utilizing C++/Qt programming. Our advanced multiscale based peak detection algorithm significantly improves the accuracy of spectral data interpretation. This work not only increases the efficiency and accuracy of spectral analysis, but also provides a more user-friendly operating environment, setting the stage for future research and wider applications across disciplines.

Due to increasingly large computational resources, modern neural networks are severely constrained due to their processing speed and energy consumption. Optical neural networks (ONNs), which use photonic structures to process signals at the physical level as an alternative to the computation in the electronic domain provided by traditional neural networks, are an attractive approach to implementing ultra-high-speed, low-energy parallel computation. Nevertheless, the gradient-free nature of ONNs makes it an obstacle to train them using the back-propagation. In this work, a stochastic function-based gradient-free training method, i.e., stochastic function direct feedback alignment (SF-DFA), is used to train a spectral slicing reservoir computing (SS-RC) architecture. Our results demonstrate that the training of ONN using SF-DFA can converge efficiently, with higher processing speed and lower energy consumption compared to the back-propagation.