An on-chip spectrometer device based on compact double-helix structure and a preparation method thereof

Abstract

This invention belongs to the field of spectral detection technology and discloses an on-chip spectrometer device and its fabrication method based on a compact double-helix structure. The device includes a silicon nitride substrate, a waveguide structure integrated on the substrate, a temperature control module, and a signal detection and reconstruction module. The waveguide structure is a double-helix rounded-corner rectangular waveguide structure. Its unequal-length straight waveguide segments and double-helix waveguide segments are configured to generate and amplify phase mismatch in optical signals of different wavelengths passing through it, thereby forming differentiated wavelength-correlated characteristic modes at the output. The temperature control module is used to regulate the temperature of the silicon nitride substrate. This invention solves the contradiction between miniaturization and high performance by achieving ultra-wideband, high-resolution spectral detection in a compact structure through controlled mode mismatch and a multi-channel collaborative mechanism.

Description

Technical Field

[0001] This invention relates to the field of spectral detection technology, and in particular to an on-chip spectrometer device based on a compact double-helix structure and its fabrication method. Background Technology

[0002] On-chip spectrometers hold significant promise for applications in portable sensing, consumer electronics, and wearable devices. The inherent contradiction between miniaturization and high performance is a core challenge hindering their development. Traditional miniature spectrometers primarily rely on wave decomposition and multiplexing, filtering, or Fourier transform architectures. Achieving high resolution and wide bandwidth typically requires extending the optical path, increasing the number of detector arrays, or introducing moving parts. In the pursuit of device miniaturization, shortening the optical path leads to a decrease in the distinguishability of characteristic modes, while increasing the number of detectors introduces a deterioration in the signal-to-noise ratio, making it difficult to simultaneously achieve ultra-wideband operating range and high spectral resolution on a single chip. Furthermore, some existing solutions attempt to reduce dependence on multiple detectors by cascading multiple interferometers and utilizing thermo-optical effects to increase the number of power channels, but their structural design is complex, sensitive to temperature disturbances, and lacks sufficient detection stability.

[0003] Therefore, there is an urgent need for a new technical approach that can achieve ultra-wideband, high-resolution spectral detection in a compact, miniaturized structure with a simple, stable, and easy-to-integrate solution. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and to propose an on-chip spectrometer device based on a compact double-helix structure, comprising: Silicon nitride substrate; A waveguide structure is integrated on the silicon nitride substrate. The waveguide structure is a double-helix rounded rectangular waveguide structure. The double-helix rounded rectangular waveguide structure includes an input end, an output end, a straight waveguide segment of unequal length, and a double-helix waveguide segment. The straight waveguide segment of unequal length and the double-helix waveguide segment are configured to generate and amplify phase mismatch of optical signals of different wavelengths passing through them, thereby forming a differentiated wavelength-related characteristic mode at the output end. A temperature control module is thermally connected to the silicon nitride substrate, and the temperature control module is used to control the temperature of the silicon nitride substrate; The signal detection and reconstruction module is connected to the output end of the waveguide structure. The signal detection and reconstruction module is used to collect the light signal intensity corresponding to the wavelength-related characteristic mode at multiple different temperatures, and output the reconstructed target spectrum based on the collected light signal intensity and the compressed sensing algorithm.

[0005] Preferably, the unequal-length straight waveguide segment and the double-helix waveguide segment together constitute a bidirectional evanescent wave coupling region. The unequal-length straight waveguide segment is connected to the input end to couple the evanescent waves and generate phase mismatch. The double-helix waveguide segment is connected to the unequal-length straight waveguide segment to enable bidirectional propagation of the optical signal to amplify the phase mismatch.

[0006] More preferably, the temperature control module is a thermoelectric cooling controller, which is physically connected to the silicon nitride substrate, and the temperature control stability accuracy of the thermoelectric cooling controller is ±0.01K.

[0007] More preferably, the temperature control range of the thermoelectric cooling controller is configured to be from 286.15K to 326.15K, and the thermoelectric cooling controller is configured to perform gradient control in temperature increments of 0.1K to generate 400 sets of optical signal datasets at different temperatures.

[0008] More preferably, the signal detection and reconstruction module includes a photodetector and a processor; the photodetector is connected to the output end of the waveguide structure and is used to detect the intensity of the optical signal; the processor is electrically connected to the photodetector and is used to construct a wavelength-temperature mapping matrix based on the collected optical signal intensity, and execute the compressed sensing algorithm to solve the underdetermined reconstruction problem to obtain the target spectrum.

[0009] More preferably, the processor solves for the target spectrum by executing the compressed sensing algorithm. The measured light intensity received from the photodetector With the constructed wavelength-temperature mapping matrix The following relationship must be satisfied: ; in, Temperature is a thermodynamic quantity that characterizes temperature. It is a length measure that characterizes the wavelength; To detect the lower limit wavelength of the bandwidth; To detect the upper limit wavelength of the bandwidth.

[0010] More preferably, the external dimensions of the double-helix rounded rectangular waveguide structure are 920μm×200μm, the working center wavelength of the on-chip spectrometer device is 1565nm, the detection bandwidth is 150nm, and the resolution is 0.08nm.

[0011] A method for fabricating an on-chip spectrometer device based on a compact double-helix structure, applicable to any one of the on-chip spectrometer devices based on a compact double-helix structure described above, the fabrication method comprising: S1: Using silicon nitride as the substrate material, a silicon nitride thin film is prepared, and the silicon nitride thin film is pretreated; S2: On the pretreated silicon nitride thin film, a double-helix rounded rectangular waveguide structure is prepared by photolithography and etching processes. The double-helix rounded rectangular waveguide structure forms an input end, an output end, a straight waveguide segment of unequal length, and a double-helix waveguide segment. The preset spacing between adjacent waveguides is controlled so that the straight waveguide segment of unequal length generates phase mismatch through evanescent wave coupling. The double-helix waveguide segment amplifies the phase mismatch through bidirectional propagation of light, so that light signals of different wavelengths form wavelength-dependent characteristic modes at the output end. S3: A thermoelectric cooling controller is physically connected to a silicon nitride substrate carrying the waveguide structure by thermal contact, and the temperature control circuit is welded and debugged to make the temperature control stability accuracy of the thermoelectric cooling controller reach ±0.01K. S4: Build a signal detection and reconstruction module. The building includes connecting a photodetector to the output end of the waveguide structure and electrically connecting a processor to the photodetector. Configure a spectral reconstruction program in the processor. The spectral reconstruction program is used to perform spectral reconstruction based on compressed sensing algorithm and using light signal intensity data corresponding to the wavelength-related characteristic mode collected at different temperatures. S5: The device prepared by the above steps is calibrated. The calibration includes inputting a known single-wavelength optical signal to the input end of the waveguide structure at multiple different temperatures, and the signal detection and reconstruction module collects and records the corresponding output optical signal intensity to construct a mapping matrix of wavelength, temperature and signal intensity.

[0012] More preferably, the double-helix rounded rectangular waveguide structure is fabricated on the silicon nitride thin film by electron beam lithography to draw the pattern, and the pattern transfer is completed by inductively coupled plasma etching, with the etching precision controlled at the sub-micron level.

[0013] More preferably, the step of calibration at multiple different temperatures includes using the thermoelectric cooling controller to perform gradient control in increments of 0.1K within a temperature range of 286.15K to 326.15K, collecting and recording the light signal intensity at each temperature point, and generating a dataset of 400 temperature channels for light signals.

[0014] Technical Effects: This invention utilizes unequal-length straight waveguide segments and double-helix waveguide segments in a waveguide structure to generate and amplify phase mismatches in optical signals of different wavelengths, thereby forming highly differentiated wavelength-correlated characteristic modes within a compact size. This controlled mode mismatch mechanism fundamentally solves the problem of insufficient mode discrimination caused by optical path limitations in traditional structures. Combined with a temperature control module that generates multi-channel optical signal sets at multiple temperature points and works in conjunction with compressed sensing algorithms, this invention achieves ultra-wideband, high-resolution spectral reconstruction on miniaturized devices without increasing the number of detectors, overcoming the inherent difficulty in balancing miniaturization and high performance in previous technologies. Attached Figure Description

[0015] Figure 1 This is a block diagram of the on-chip spectrometer device based on a compact double-helix structure.

[0016] Figure 2 The diagram shows the overall structure and experimental characterization of the spectrometer of this invention; (a) is a schematic diagram of the spectrometer architecture, (b) is a graph of the spectral autocorrelation function of 400 temperature channels, (c) is a comparison graph of the reconstructed continuous broadband spectrum (solid line) and the reference signal (dashed line), and (d) is a graph of the evolution of reconstruction error with the number of channels (100~400). Figure 3 These are experimental diagrams of narrowband spectral reconstruction characterization of the spectrometer of the present invention; wherein, (a) is a single-peak narrowband spectral reconstruction diagram, and (b) is a double-peak narrowband spectral reconstruction diagram; Figure 4 These are complex broadband spectral characterization diagrams of the spectrometer of the present invention; wherein, (a), (b), and (c) are reconstruction diagrams of the characteristic peaks of 1530 nm, 1590 nm, and 1610 nm superimposed on the broadband continuous spectrum, and (d), (e), and (f) are reconstruction diagrams of the absorption valleys of 1545 nm, 1550 nm, and 1555 nm in the broadband spectrum.

[0017] Figure 5 This invention relates to a method for fabricating an on-chip spectrometer device based on a compact double-helix structure. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Traditional technical solutions have the following technical problems: During the miniaturization process, on-chip spectrometers are limited by optical path length and cannot generate and amplify wavelength-sensitive phase differences in a compact device size. This results in insufficient differentiation of wavelength-related feature modes formed at the output end, making it difficult to simultaneously achieve ultra-wideband operating range and high spectral resolution on a single chip.

[0020] Based on this, please refer to Figures 1-5 This embodiment provides an on-chip spectrometer device based on a compact double-helix structure, including a silicon nitride substrate, a waveguide structure, a temperature control module, and a signal detection and reconstruction module. The waveguide structure is integrated on the silicon nitride substrate and is a double-helix rounded-corner rectangular waveguide structure. The double-helix rounded-corner rectangular waveguide structure includes an input end, an output end, unequal-length straight waveguide segments, and a double-helix waveguide segment. The unequal-length straight waveguide segment and the double-helix waveguide segment are configured to generate and amplify phase mismatch in optical signals of different wavelengths passing through them, thereby forming a differentiated wavelength-correlated characteristic mode at the output end. The temperature control module is thermally connected to the silicon nitride substrate and is used to control the temperature of the silicon nitride substrate. The signal detection and reconstruction module is connected to the output end of the waveguide structure and is used to acquire the optical signal intensity corresponding to the wavelength-correlated characteristic mode at multiple different temperatures, and output the reconstructed target spectrum based on the acquired optical signal intensity and a compressed sensing algorithm.

[0021] The silicon nitride substrate serves as the photonic integration platform for the entire device, providing a low-loss optical transmission medium for the waveguide structure. The waveguide structure is the physical core for achieving controlled mode mismatch. A broadband optical signal or the optical signal under test is coupled into the input end of the waveguide structure via optical fiber, first reaching the unequal-length straight waveguide segment. This unequal-length straight waveguide segment comprises two parallel rectangular waveguides with a very small spacing. This spacing is designed as a preset spacing for evanescent wave coupling, causing periodic energy exchange of the optical signals transmitted in the two waveguides, i.e., generating evanescent wave coupling. Since the unequal-length straight waveguide segment consists of two waveguides with a preset physical length difference, the optical path length of the optical signal transmitted in the two waveguides differs. This difference accumulates a phase difference highly sensitive to wavelength during transmission, resulting in a fundamental phase mismatch. This causes initial mode differences to form when optical signals of different wavelengths are output from the unequal-length straight waveguide segment. The optical signal then enters a double-helix waveguide section directly connected to the unequal-length straight waveguide section. This waveguide section employs a rounded rectangular multi-turn helical geometry, and its zigzag optical path forces the optical signal to propagate bidirectionally in both the forward and reverse directions, multiplying the effective optical path length without increasing the physical size of the device. The minute phase mismatch generated by the unequal-length straight waveguide section is repeatedly accumulated and significantly amplified during the bidirectional helical transmission in the double-helix waveguide section, ultimately forming differentiated wavelength-dependent characteristic patterns at the output end. Through the combined effect of the phase mismatch generated by the unequal-length straight waveguide section and the amplification of the phase mismatch by the double-helix waveguide section, incident light of different wavelengths forms highly differentiated and stably distinguishable intensity distribution patterns at the output end of the waveguide structure, thus forming unique wavelength-dependent characteristic patterns, much like giving each wavelength a unique fingerprint.

[0022] The temperature control module, thermally connected to the silicon nitride substrate, supports system operation from two dimensions by precisely controlling the substrate temperature. Firstly, the high-stability temperature control provided by the module effectively suppresses interference from environmental thermal fluctuations and device self-heating effects on the refractive index of the silicon nitride waveguide, thus ensuring the stability and uniqueness of the wavelength-related characteristic modes. Secondly, the module can actively adjust the temperature to allow the waveguide structure to operate at multiple different preset temperature points. For the same input wavelength, the refractive index of the waveguide material changes slightly with temperature, resulting in a controllable change in the output wavelength-related characteristic mode pattern. Therefore, at different stable temperature points, the entire system obtains multiple sets of different, non-redundant light intensity responses. The signal detection and reconstruction module acquires data once at each preset temperature, essentially opening a new measurement channel. Through the signal detection and reconstruction module connected to the output terminal, all light signal intensity information corresponding to each temperature point is converted into electrical signals and recorded, forming a large multi-channel light signal intensity dataset modulated by both wavelength and temperature parameters. Subsequently, the signal detection and reconstruction module utilizes compressed sensing algorithms to solve the underdetermined problem constructed from this dataset, reconstructing the complete spectrum of the original input light with high probability from the seemingly aliased measurements. This scheme, through a compact, controlled mode mismatch waveguide design, cleverly utilizes temperature control to construct a virtual measurement channel, overcoming the dependence of traditional miniature spectrometers on long optical paths and multiple detectors, thus achieving both ultra-wideband and high-resolution performance on a miniaturized device.

[0023] It is worth mentioning that the silicon nitride substrate is a high-quality silicon nitride thin film prepared by chemical vapor deposition, with a thickness of 300 nanometers, exhibiting extremely low transmission loss in the near-infrared communication band. The waveguide structure pattern transfer is completed using electron beam lithography and inductively coupled plasma etching processes, with a waveguide width of 1 micrometer and a waveguide sidewall steepness controlled above 85 degrees, ensuring single-mode transmission and evanescent wave coupling efficiency. In the unequal-length straight waveguide segments, the length difference between the two parallel waveguide segments is designed to be 50 micrometers, which determines the spectral period of the basic phase mismatch. The total arm length of the double-helix waveguide segment is 2 millimeters, with semicircular rings of 30 micrometer radius connected end to end to form a rounded rectangular helix. The number of turns can be designed according to the required magnification; in this embodiment, it is 5 turns. The temperature control module uses a TEC1-12706 semiconductor cooler, whose hot side is tightly attached to the back of the silicon nitride substrate via thermal grease. The temperature switching and stabilization time during signal acquisition are optimized using a closed-loop PID control algorithm, enabling the signal to move from one temperature point to the next and stabilize within a preset accuracy range within 10 seconds. The photodetector in the signal detection and reconstruction module is an indium gallium arsenide photodiode with a spectral response range covering 1200 nm to 1700 nm. The processor is based on an ARM-based embedded platform, and the acquired mapping matrix data and reconstruction algorithm program are stored in onboard DDR memory. The compressed sensing algorithm specifically employs a fully variational regularized optimization solution, with a preset iteration count of 200 and a convergence threshold set to 10^-6. The final reconstructed spectrum is output to a host computer or display via a digital interface.

[0024] Traditional technical solutions suffer from the following technical problems: the physical mechanisms and structures underlying the generation and amplification of phase mismatch are unclear, making it difficult for those skilled in the art to implement designs that achieve efficient mode differentiation within a compact space. Therefore, the unequal-length straight waveguide segment and the double-helix waveguide segment together constitute a bidirectional evanescent wave coupling region. The unequal-length straight waveguide segment is connected to the input end to couple the evanescent waves and generate phase mismatch; the double-helix waveguide segment is connected to the unequal-length straight waveguide segment to enable bidirectional propagation of the optical signal, thereby amplifying the phase mismatch.

[0025] The bidirectional evanescent wave coupling region is a continuous functional area on the waveguide structure, seamlessly integrating two levels of functionality. The first level of functionality is provided by the unequal-length straight waveguide segment directly connected to the input end. Its working principle is that when light enters from the input end, it first distributes energy between the two parallel waveguide segments. The submicron-level spacing between the two waveguide segments ensures evanescent wave coupling, while the intentionally designed unequal lengths ensure that when the two beams reconverge at the waveguide exit, the existence of the optical path difference generates wavelength-sensitive interference and phase mismatch. This mismatch is directly manifested in the different energy distribution ratios of light of different wavelengths when output from this segment. The second level of functionality is provided by the double-helix waveguide segment directly connected to the unequal-length straight waveguide segment. After leaving the unequal-length straight waveguide segment, the optical signal immediately enters the curved and folded double-helix waveguide segment. Here, the optical signal no longer propagates in a single direction but travels along a spiral path. During this process, partial back reflection and scattering occur at the bends, allowing light propagating in both the forward and reverse directions to coexist within the waveguide cavity, forming bidirectional propagation. This bidirectional propagation creates a weak optical resonance and interference superposition environment, causing the energy distribution differences previously caused by phase mismatch to accumulate, amplify, and filter through repeated cycles. Ultimately, this results in extremely sharp and high-contrast characteristic peaks and valleys on the transmission spectrum of different wavelengths at the output, achieving spectral resolution far exceeding that of a single action. This combination cleverly utilizes spatial asymmetry and the folding structure to achieve controlled enhancement of mode differentiation.

[0026] It is worth mentioning that the center-to-center spacing between the two parallel waveguide segments of the unequal-length straight waveguide section is 1.2 micrometers, and the coupling length is 200 micrometers. This combination of parameters ensures the most efficient energy coupling at the center wavelength of 1565 nanometers. The rounded rectangular turning radius used in the double-helix waveguide section is designed to be 25 micrometers. Simulation verification shows that this radius effectively reduces bending loss and maintains single-mode transmission conditions. The total physical footprint of the waveguide structure is only 920 micrometers by 200 micrometers. The physical realization of the entire bidirectional evanescent wave coupling region requires no suspended structures or heterogeneous material integration. It is formed entirely on a silicon nitride planar thin film layer in a single step through a single-step etching process, greatly simplifying the fabrication complexity and improving structural robustness.

[0027] Traditional technical solutions suffer from the following problems: fluctuations in ambient temperature interfere with the refractive index of the waveguide, leading to instability in phase mismatch modes and severely degrading the accuracy and repeatability of spectral reconstruction. Therefore, the temperature control module is a thermoelectric cooling controller, which is physically connected to the silicon nitride substrate, and the temperature control stability accuracy of the thermoelectric cooling controller is ±0.01 Kelvin.

[0028] The core function of the temperature control module is to provide a stable and reproducible physical environment for the mapping relationship between wavelength, temperature, and light intensity. A thermoelectric cooling controller is chosen as the specific implementation component because it can both cool and heat, and by changing the direction and magnitude of the driving current, the temperature can be flexibly adjusted bidirectionally over a wide range. The cold side of the thermoelectric cooling controller is physically connected to the back side of the silicon nitride substrate carrying the waveguide structure through a thermal interface material with high thermal conductivity, forming a thermal contact. Its temperature control stability accuracy reaches ±0.01 Kelvin, meaning that the controller can suppress the actual temperature fluctuation of the substrate within this extremely narrow window of 0.01 Kelvin for a long time. This ultra-high precision temperature control is crucial because the refractive index of the silicon nitride waveguide is very sensitive to temperature. Even small temperature fluctuations can cause the wavelength-related characteristic modes to drift, introducing non-negligible noise in single or multiple measurements, disrupting the preset correlation between the elements of the mapping matrix, and ultimately leading to spectral reconstruction failure or a significant decrease in accuracy. By locking the temperature, the feature patterns generated at each preset temperature point are ensured to have a high degree of uniqueness and repeatability. This is the physical reliability basis for the subsequent compressed sensing algorithm to correctly solve underdetermined problems.

[0029] It is worth mentioning that the thermoelectric cooling controller uses a multi-stage cascaded thermocouple stack structure, and its control circuit employs a bridge-driven proportional-integral-derivative feedback control chip, model MAX1978. The temperature feedback element is a negative temperature coefficient thermistor with an accuracy of 0.001 Kelvin, which is directly cured onto the surface of the silicon nitride substrate with thermally conductive adhesive, adjacent to the waveguide structure, to achieve accurate sampling of the waveguide region temperature. The soldering and debugging of the control circuit must ensure that the ripple noise of the drive current is less than 1% to avoid the fluctuation of the drive current itself being converted into temperature disturbance through the Joule heating effect.

[0030] Traditional technical solutions suffer from the following problems: relying solely on high-precision temperature control cannot obtain rich multi-channel spectral data, and there is a lack of systematic temperature modulation strategies to provide a sufficient number of incoherent measurements for compressed sensing algorithms. Therefore, the temperature control range of the thermoelectric cooling controller is configured from 286.15 Kelvin to 326.15 Kelvin, and the thermoelectric cooling controller is configured to perform gradient control in temperature increments of 0.1 Kelvin to generate 400 sets of optical signal datasets at different temperatures.

[0031] The core of this technical solution lies in the creative use of temperature as a scanning dimension to generate virtual measurement channels, thereby overcoming the limitation imposed by the number of physical detectors on the dimension of the measurement matrix. The temperature control range of the thermoelectric cooling controller is specifically configured from 286.15 Kelvin to 326.15 Kelvin. This wide temperature window of 40 Kelvin is sufficient to cause a significant and regular change in the refractive index of the waveguide structure. Step-gradient control in increments of 0.1 Kelvin means setting 400 discrete, precisely controllable temperature operating points across the entire temperature range. At each temperature point, after the device stabilizes, the signal detection and reconstruction module acquires the light signal intensity across the entire wavelength band. Thus, for the same input spectrum, 400 different light intensity measurements can be obtained. These 400 sets of data constitute 400 independent spectral detection channels, far exceeding the information provided by a single physical photodetector. These measurement channels exhibit good incoherence because they originate from the distinct spectral fingerprints output by the waveguide structure at different temperatures. These 400 sets of data together constitute a huge optical signal dataset, providing sufficient equation entry points for solving the compressed sensing algorithm, making it possible to solve for 1875 spectral components from a single physical measurement, and serving as a key data foundation for solving the underdetermined reconstruction problem.

[0032] It is worth noting that the lower limit of the temperature control range, 286.15 Kelvin, is chosen slightly below room temperature to avoid introducing external heating interference in a conventional testing environment; the upper limit, 326.15 Kelvin, falls within the stable operating range of the waveguide and thermal interface materials. The 0.1 Kelvin temperature increment is a parameter optimized theoretically and experimentally, achieving a balance between spectral reconstruction accuracy and total measurement time. Scanning all 400 temperature points takes approximately 40 minutes, and the entire process is automatically controlled by the processor without manual intervention.

[0033] Traditional technical solutions suffer from the following technical problems: the hardware architecture for signal detection and algorithm reconstruction is unclear, making it impossible for those skilled in the art to implement the complete process from optical signal acquisition to spectral output. Therefore, the signal detection and reconstruction module includes a photodetector and a processor; the photodetector, connected to the output end of the waveguide structure, is used to detect the intensity of the optical signal; the processor, electrically connected to the photodetector, is used to construct a wavelength-temperature mapping matrix based on the acquired optical signal intensity and execute the compressed sensing algorithm to solve the underdetermined reconstruction problem to obtain the target spectrum.

[0034] The signal detection and reconstruction module serves as a bridge connecting the optical front-end and the digital back-end, completing the closed loop from physical quantities to digital information and finally to the spectrum. Its internal functions are clearly divided into two core components: the photodetector and the processor. The photodetector is directly aligned and connected to the output end of the waveguide structure via optical fiber or end-face coupling. Its sole function is to convert the optical signal emitted from the output end, containing the wavelength-related characteristic mode information, into an analog electrical signal. The magnitude of this analog electrical signal is linearly related to the incident light power. This step is fundamental to all quantitative analysis. The processor, acting as the computational and control hub, receives the digitized optical signal intensity data from the photodetector via an analog-to-digital converter. Its first key task is mapping matrix construction. During the calibration phase, the processor records the light intensity values ​​obtained by the photodetector at each known wavelength and at each preset temperature, forming a large two-dimensional data table. Each element in this data table represents a deterministic response relationship between a specific wavelength-temperature pair and light intensity; this is the wavelength-temperature mapping matrix. During the reconstruction phase, when faced with unknown input light, the processor collects a set of mixed light intensity values ​​at the current temperature, retrieves the pre-stored wavelength-temperature mapping matrix, executes the compressed sensing algorithm to solve it, and finally calculates and outputs the target spectrum.

[0035] It is worth mentioning that the photodetector is specifically an indium gallium arsenide photodetector assembly integrating a transimpedance amplifier, with a -3 dB bandwidth of 10 MHz, meeting the requirements for quasi-static measurement. The processor adopts a heterogeneous computing architecture based on FPGA and ARM, where the FPGA is used for high-speed parallel acquisition of the data stream output from the analog-to-digital converter, and the ARM is responsible for running the operating system and executing the compressed sensing algorithm. The wavelength-temperature mapping matrix is ​​stored in non-volatile memory in a sparse matrix format after calibration to save storage space and speed up matrix-vector multiplication operations during algorithm calls.

[0036] Traditional technical solutions suffer from the following problems: the mathematical model upon which compressed sensing reconstruction relies is not specifically defined in the patent documents, resulting in an unclear algorithm implementation process for the processor and making it difficult to ascertain how to obtain the target spectrum from the measured values. Therefore, the processor solves for the target spectrum by executing the compressed sensing algorithm. The measured light intensity received from the photodetector With the constructed wavelength-temperature mapping matrix The following relationship must be satisfied:

[0037] in, Temperature is a thermodynamic quantity that characterizes temperature. It is a length measure that characterizes the wavelength; To detect the lower limit wavelength of the bandwidth; To detect the upper limit wavelength of the bandwidth.

[0038] This formula is the core mathematical model connecting physical measurement and spectral reconstruction; its physical essence is a first-kind Friedholm integral equation. The left side of the equation... This represents a specific moment or a specific preset temperature. Under these conditions, the photodetector measures and outputs the total light intensity value to the processor. It is important to emphasize that this measured value... It is not a contribution from a single wavelength, but rather at that temperature. The result of the combined effect of all wavelength components is a mixed signal. The right side of the equation precisely describes this mixing process: the target spectrum. It is the unknown function we ultimately want, which describes different wavelengths. The optical power spectral density distribution at a given location, with wavelength as the independent variable. Mapping matrix It is the known transfer function of the system, which is also a function of wavelength. and temperature A bivariate function. Each element in the equation can be understood physically as: at a specific temperature Below, a unit intensity with a wavelength of When a monochromatic light input system is used, the output light intensity value that can be obtained at the photodetector is described. Integration process. Indicates the wavelength at the lower limit of the detection bandwidth. Up to the upper limit wavelength Within the range, all wavelength components Its corresponding response weight After multiplication, the equations are continuously accumulated. Since we cannot directly obtain continuous functions in physics, in the actual algorithm implementation, this integral equation will be discretized into a system of linear equations. ,in It is a 400-dimensional measurement vector. It is the N-dimensional vector representing the spectrum that is to be solved. It is a 400-row, N-column matrix. Since the number of temperature channels (400) is much smaller than the number of discrete spectral points (N), this system of equations is underdetermined, with infinitely many solutions. Compressed sensing algorithms are introduced here, and their core principle is to utilize… By solving a convex optimization problem with L1 norm regularization in a certain transform domain, the sparsest solution, which is also the most likely to be the true spectrum, is found from infinitely many solutions, thus enabling high-fidelity reconstruction of the original spectrum from a small number of mixed measurements.

[0039] It is worth mentioning that when the formula is discretized, the wavelength... The discrete interval was set to 0.04 nm, generating 3750 spectral channels within a 150 nm detection bandwidth, far exceeding the 400 temperature measurement channels, thus achieving a compression ratio of approximately 9:1. The specific algorithm used for solving the L1 norm regularization was a two-step iterative shrinkage threshold algorithm, which achieved a good balance between computational efficiency and reconstruction accuracy. In terms of dimensional analysis... The dimensionless quantity Kelvin, which has the properties of thermodynamic temperature. , and A function with the dimension of length, nanometer. The dimension of is power divided by length. The dimension of is power. The dimension of the formula is 1 divided by the length. This clear correspondence of physical dimensions ensures the correctness of the formula in physical calculations and eliminates any suspicion of pure mathematical operations.

[0040] Traditional technical solutions suffer from the following problems: the specific operating wavelength and key performance indicators of the device are not defined, and the key application window of the technical solution is unclear. Based on this, the external dimensions of the double-helix rounded rectangular waveguide structure are 920 micrometers by 200 micrometers, the operating center wavelength of the on-chip spectrometer device is 1565 nanometers, the detection bandwidth is 150 nanometers, and the resolution is 0.08 nanometers.

[0041] This scheme clearly defines the typical operating window and key performance indicators of the entire device in the near-infrared communication and sensing bands. The external dimensions of 920 μm x 200 μm define the physical area occupied by the waveguide structure, fully demonstrating its fundamental advantage in miniaturization, allowing the spectrometer to be integrated as an on-chip component into various portable devices. The operating center wavelength of 1565 nm is clearly located in the near-infrared band, covering a detection bandwidth of 150 nm from 1490 nm to 1640 nm. This band corresponds to the C-band and L-band of fiber optic communication and is also the region containing characteristic absorption lines for many molecular gases and substances, possessing significant application value. The high resolution of 0.08 nm means that the device can resolve approximately 1875 independent spectral channels within this ultra-wide 150 nm spectral range. This combination of performance parameters clearly outlines the application window and capability boundaries of this device: achieving ultra-wideband spectral detection with sub-nanometer resolution, covering mainstream communication and sensing bands, within a compact physical size sufficient for on-chip integration.

[0042] It is worth mentioning that the choice of the center wavelength of 1565 nm is matched with the low-loss window of silicon nitride material in the near-infrared band and common optical fiber communication bands. The resolution of 0.08 nm was obtained by calculating the autocorrelation function of the calibrated mapping matrix and taking the full width at half maximum (FWHM) value; this resolution has been experimentally verified. This device achieves this performance without any external mechanical scanning components or bulk optical elements.

[0043] In traditional methods, the fabrication of on-chip micro-spectrometers often involves complex three-dimensional integration or the release of suspended structures, resulting in numerous process steps and low yields. Therefore, this fabrication method includes using silicon nitride as the substrate material to prepare a silicon nitride thin film, followed by pretreatment of the silicon nitride thin film. On the pretreated silicon nitride thin film, a double-helix rounded-corner rectangular waveguide structure is fabricated using photolithography and etching processes. This structure forms an input end, an output end, unequal-length straight waveguide segments, and a double-helix waveguide segment. A preset spacing between adjacent waveguides is controlled so that the unequal-length straight waveguide segments generate phase mismatch through evanescent wave coupling. The double-helix waveguide segment amplifies this phase mismatch through bidirectional light propagation, resulting in different wavelengths of light signals forming differentiated wavelength-correlated characteristic modes at the output end. A thermoelectric cooling controller is physically connected to the silicon nitride thin film carrying the waveguide structure, and the temperature control circuit is soldered and adjusted. The test aims to achieve a temperature control stability accuracy of ±0.01 Kelvin for the thermoelectric cooling controller. A signal detection and reconstruction module is then constructed, comprising connecting a photodetector to the output terminal of the waveguide structure and electrically connecting a processor to the photodetector. A spectral reconstruction program is configured in the processor, which uses compressed sensing algorithms to reconstruct the spectrum based on optical signal intensity data corresponding to the wavelength-related characteristic modes collected at different temperatures. The device prepared in the above steps is calibrated by inputting a known single-wavelength optical signal to the input terminal of the waveguide structure at multiple different temperatures, and the signal detection and reconstruction module collects and records the corresponding output optical signal intensity to construct a mapping matrix of wavelength, temperature, and signal intensity.

[0044] This fabrication method is entirely based on standard CMOS-compatible planar processes, without any floating or three-dimensional stacking steps, greatly ensuring the repeatability and cost control of mass production. The process begins with growing a high-quality silicon nitride thin film on a silicon substrate using low-pressure chemical vapor deposition (LPCVD), followed by a standard cleaning process to remove surface particles and organic contaminants. The core patterning step then begins. After spin-coating electron beam resist, a double-helix rounded rectangular pattern is directly written using electron beam lithography. After development and fixing, the pattern is precisely transferred to the silicon nitride thin film using inductively coupled plasma etching (ICP-C) to form a waveguide structure. Here, precise control of lithography and etching parameters ensures the waveguide sidewall morphology and waveguide gap, thereby locking in the physical structure that generates evanescent wave coupling and phase mismatch. Afterward, the discrete TEC cold surfaces are uniformly coated with thermally conductive silver paste and bonded and cured to the back of the chip containing the waveguide structure, completing the physical connection. Finally, a temperature control circuit board, including a driver chip, H-bridge circuit, and thermistor feedback loop, is soldered on, and a stable accuracy of ±0.01 Kelvin is achieved by adjusting the PID parameters. The signal link setup involves high-precision alignment of the output terminal of the indium gallium arsenide photodetector with a pigtail to the output waveguide end face of the chip, followed by UV curing. The detector's electrical signal output is connected to the data acquisition and processing platform via a coaxial cable. The final calibration step is crucial for activating the spectrometer's functionality. A tunable laser scans in 0.1-nanometer steps within the 1490-1640 nm range, while simultaneously controlling the TEC to scan in 0.1-Kelvin steps within the 286.15-326.15 Kelvin range. Light intensity is acquired and recorded at each temperature-wavelength combination point, generating a multidimensional mapping matrix stored in the processor. This fabrication process clearly defines the input, output, and acceptance criteria for each step, allowing those skilled in the art to fabricate and calibrate the device without inventive effort.

[0045] It is worth mentioning that the thickness of the silicon nitride thin film was measured using an ellipsometry, and its uniformity was controlled within ±2%. The inductively coupled plasma etching gas used during etching was a mixture of sulfur hexafluoride and octafluorocyclobutane. By adjusting the gas flow ratio and RF bias power, a near 90-degree vertical sidewall and linewidth control accuracy better than ten nanometers were achieved. In the temperature control circuit debugging, the specific settings of the PID parameters were optimized by applying a pseudo-random binary sequence perturbation signal and observing the system's closed-loop step response, ensuring no overshoot and a minimum settling time. During the calibration process, the optical power data was repeatedly collected one hundred times at each point and averaged to eliminate the influence of high-frequency noise.

[0046] Traditional technical solutions suffer from the following problems: the sub-micron-level feature dimensions of the waveguide structure place extremely high demands on etching precision, and the implementation details of the process method are unclear. Therefore, a double-helix rounded-corner rectangular waveguide structure is fabricated on the silicon nitride thin film using electron beam lithography to create the pattern, and then transferred using inductively coupled plasma etching (ICP-C) to achieve sub-micron-level etching precision.

[0047] This scheme clarifies the specific process path for achieving high-quality submicron waveguide structures. Electron beam lithography (EBL) leverages its high resolution to precisely expose complex double-helix rounded rectangle patterns designed in computer-aided design software onto an electron beam-sensitive resist with nanometer-level precision. EBL avoids the overlay errors associated with physical masks, making it particularly suitable for structures like unequal-length straight waveguides and helical waveguides, which are extremely sensitive to absolute dimensions and spacing. Subsequently, inductively coupled plasma (ICP) etching transfers the pattern on the resist to the underlying silicon nitride film with high fidelity. ICP, through independent radio frequency source control of plasma density and ion bombardment energy, enables highly anisotropic etching, which is crucial for forming waveguides with steep sidewalls and precise linewidths. Controlling the etching precision to the sub-micron level is the key process guarantee for the success of this solution, because the core function of the waveguide structure—generating and amplifying phase mismatch—strongly depends on the accuracy of the waveguide's geometry and spacing. Any deviation exceeding the tolerance range will cause distortion of the designed phase mismatch mode, thereby degrading the spectral reconstruction performance.

[0048] It is worth mentioning that electron beam lithography, using an accelerating voltage of 100 kV and an electron beam current of 1 nanoamp, combined with a positive chemical amplification resist, can achieve a pattern resolution better than 10 nanometers on a 400-nanometer-thick resist layer. During inductively coupled plasma etching, the source power is set to 500 W, the bias power to 50 W, and the working pressure maintained at 10 mTorr. These conditions provide a stable silicon nitride etching rate of approximately 200 nanometers per minute. Submicron-level etching precision is specifically ensured through scanning electron microscopy linewidth measurements, with the linewidth controlled to ±10 nanometers of the target size.

[0049] Traditional technical solutions suffer from the following problems: the specific parameters and operating procedures of the calibration process are unclear, and no detailed instructions are provided for generating 400 sets of mapping datasets. Therefore, the step of calibration at multiple different temperatures includes using the thermoelectric cooling controller to perform gradient control in increments of 0.1 Kelvin within a temperature range of 286.15 Kelvin to 326.15 Kelvin, acquiring and recording the light signal intensity at each temperature point to generate 400 sets of light signal datasets for different temperature channels.

[0050] This scheme precisely parameterizes the crucial calibration step, making it a clear, repeatable, and standardized operational procedure. The core objective of calibration is to establish an accurate wavelength-temperature mapping matrix. The scheme defines the temperature operating range and stepping method for calibration: using the thermoelectric cooling controller as the actuator, a complete, step-by-step heating or cooling scan is performed within a temperature range of 286.15 Kelvin to 326.15 Kelvin, spanning 40 Kelvin, in fixed increments of 0.1 Kelvin. At each stabilized temperature point, the system initiates a wavelength scan measurement, recording the corresponding optical signal intensity. Ultimately, the entire process generates 400 sets of optical signal datasets, each precisely corresponding to a specific temperature point. These 400 temperature channels form the temperature dimension of the mapping matrix. By specifically defining the temperature range, increments, and the final number of channels, this scheme provides a followable technical procedure for quality control and process reproducibility. Any operator can automatically write control scripts according to this procedure to achieve automated calibration of the precision temperature-controlled spectrometer, ensuring that devices prepared in different batches can obtain a mapping matrix with consistency and integrity.

[0051] It is worth noting that before data acquisition at each temperature point, the control program actively checks the thermistor reading. Only after the temperature has stabilized within an error range of ±0.01 Kelvin for more than 3 seconds will the optical signal acquisition program be triggered. The optical signal data for each temperature point consists of all light intensity values ​​obtained by scanning the tunable laser across the entire wavelength range of 1490 nm to 1640 nm in 0.1 nm increments. Since full-wavelength scanning at a single temperature point is time-consuming, to prevent long-term drift, the system operates under high-precision constant temperature. After each temperature point measurement, the light intensity at a standard reference wavelength point is quickly remeasured to verify the system's stability. If the requirements are not met, all data at that temperature point is automatically remeasured. The generated 400 datasets are packaged and stored in binary file format, along with temperature and time metadata for verification and retrieval.

[0052] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An on-chip spectrometer device based on a compact double-helix structure, characterized in that, include: Silicon nitride substrate; A waveguide structure is integrated on the silicon nitride substrate. The waveguide structure is a double-helix rounded rectangular waveguide structure. The double-helix rounded rectangular waveguide structure includes an input end, an output end, a straight waveguide segment of unequal length, and a double-helix waveguide segment. The straight waveguide segment of unequal length and the double-helix waveguide segment are configured to generate and amplify phase mismatch of optical signals of different wavelengths passing through them, thereby forming a differentiated wavelength-related characteristic mode at the output end. A temperature control module is thermally connected to the silicon nitride substrate, and the temperature control module is used to control the temperature of the silicon nitride substrate; The signal detection and reconstruction module is connected to the output end of the waveguide structure. The signal detection and reconstruction module is used to collect the light signal intensity corresponding to the wavelength-related characteristic mode at multiple different temperatures, and output the reconstructed target spectrum based on the collected light signal intensity and the compressed sensing algorithm.

2. The on-chip spectrometer device based on a compact double-helix structure according to claim 1, characterized in that, The unequal-length straight waveguide segment and the double-helix waveguide segment together form a bidirectional evanescent wave coupling region. The unequal-length straight waveguide segment is connected to the input end to couple the evanescent waves and generate phase mismatch. The double-helix waveguide segment is connected to the unequal-length straight waveguide segment to enable bidirectional propagation of the optical signal to amplify the phase mismatch.

3. The on-chip spectrometer device based on a compact double-helix structure according to claim 2, characterized in that, The temperature control module is a thermoelectric cooling controller, which is physically connected to the silicon nitride substrate, and the temperature control stability accuracy of the thermoelectric cooling controller is ±0.01K.

4. The on-chip spectrometer device based on a compact double-helix structure according to claim 3, characterized in that, The temperature control range of the thermoelectric cooling controller is configured to be from 286.15K to 326.15K, and the thermoelectric cooling controller is configured to perform gradient control in temperature increments of 0.1K to generate 400 sets of optical signal datasets at different temperatures.

5. The on-chip spectrometer device based on a compact double-helix structure according to claim 2, characterized in that, The signal detection and reconstruction module includes a photodetector and a processor; the photodetector is connected to the output end of the waveguide structure and is used to detect the intensity of the optical signal; the processor is electrically connected to the photodetector and is used to construct a wavelength-temperature mapping matrix based on the collected optical signal intensity, and execute the compressed sensing algorithm to solve the underdetermined reconstruction problem to obtain the target spectrum.

6. The on-chip spectrometer device based on a compact double-helix structure according to claim 5, characterized in that, The processor obtains the target spectrum by executing the compressed sensing algorithm. The measured light intensity received from the photodetector With the constructed wavelength-temperature mapping matrix The following relationship must be satisfied: ； in, Temperature is a thermodynamic quantity that characterizes temperature. It is a length measure that characterizes the wavelength; To detect the lower limit wavelength of the bandwidth; To detect the upper limit wavelength of the bandwidth.

7. The on-chip spectrometer device based on a compact double-helix structure according to claim 2, characterized in that, The external dimensions of the double-helix rounded rectangular waveguide structure are 920μm×200μm, and the working center wavelength of the on-chip spectrometer device is 1565nm, the detection bandwidth is 150nm, and the resolution is 0.08nm.

8. A method for fabricating an on-chip spectrometer device based on a compact double-helix structure, applicable to the on-chip spectrometer device based on a compact double-helix structure as described in any one of claims 1 to 7, characterized in that, The preparation method includes: S1: Using silicon nitride as the substrate material, a silicon nitride thin film is prepared, and the silicon nitride thin film is pretreated; S2: On the pretreated silicon nitride thin film, a double-helix rounded rectangular waveguide structure is prepared by photolithography and etching processes. The double-helix rounded rectangular waveguide structure forms an input end, an output end, a straight waveguide segment of unequal length, and a double-helix waveguide segment. The preset spacing between adjacent waveguides is controlled so that the straight waveguide segment of unequal length generates phase mismatch through evanescent wave coupling. The double-helix waveguide segment amplifies the phase mismatch through bidirectional propagation of light, so that light signals of different wavelengths form wavelength-dependent characteristic modes at the output end. S3: A thermoelectric cooling controller is physically connected to a silicon nitride substrate carrying the waveguide structure by thermal contact, and the temperature control circuit is welded and debugged to make the temperature control stability accuracy of the thermoelectric cooling controller reach ±0.01K. S4: Build a signal detection and reconstruction module. The building includes connecting a photodetector to the output end of the waveguide structure and electrically connecting a processor to the photodetector. Configure a spectral reconstruction program in the processor. The spectral reconstruction program is used to perform spectral reconstruction based on compressed sensing algorithm and using light signal intensity data corresponding to the wavelength-related characteristic mode collected at different temperatures. S5: The device prepared by the above steps is calibrated. The calibration includes inputting a known single-wavelength optical signal to the input end of the waveguide structure at multiple different temperatures, and the signal detection and reconstruction module collects and records the corresponding output optical signal intensity to construct a mapping matrix of wavelength, temperature and signal intensity.

9. The method for fabricating an on-chip spectrometer device based on a compact double-helix structure according to claim 8, characterized in that, The double-helix rounded rectangular waveguide structure was fabricated on the silicon nitride thin film by electron beam lithography to draw the pattern, and the pattern was transferred by inductively coupled plasma etching, with the etching precision controlled at the submicron level.

10. The method for fabricating an on-chip spectrometer device based on a compact double-helix structure according to claim 8, characterized in that, The step of calibration at multiple different temperatures includes using the thermoelectric cooling controller to perform gradient control in increments of 0.1K within a temperature range of 286.15K to 326.15K, collecting and recording the light signal intensity at each temperature point, and generating a dataset of 400 temperature channels for light signals.

Images