Single-pixel on-chip spectrometer based on local electric field modulation of multi-electrode structure
By employing a multi-electrode structure for local electric field modulation in a single-pixel photodetector and utilizing multiple control electrodes to change the carrier collection weight, the integration challenges of spectral coding and signal processing in existing technologies are solved, achieving efficient spectral reconstruction and detection.
Patent Information
- Application Number
- CN202611581894.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-13
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies struggle to achieve spectral encoding and signal processing within a limited chip area and rely on multiple independent detection channels or pixels, making it impossible to effectively utilize the differences in response configuration of the electrical state of the same detector for spectral reconstruction.
By employing a multi-electrode structure, local electric field modulation is used within the same photosensitive area. Multiple electrically isolated control electrodes are used to change the carrier collection weights, forming an adjustable spectral code. The incident spectrum is then reconstructed through multi-state measurements.
While maintaining the overall readout of the same photosensitive area, it provides additional spectral information, reduces reliance on independent optical coding structures and multi-pixel readout, and improves the resolution and integration of spectral detection.
Smart Images

Figure CN123111467A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a single-pixel on-chip spectrometer based on local electric field modulation of a multi-electrode structure. It modulates the spectral response of the same detector by local electric field modulation of multiple electrodes and obtains the incident spectrum by calculation reconstruction. It can be used for on-chip spectral detection and belongs to the field of photoelectric detection and spectral analysis technology. Background Technology
[0002] Spectroscopic detection analyzes the absorption, reflection, or emission characteristics of substances by acquiring the distribution of light intensity with wavelength, and can be used for substance identification, environmental monitoring, and industrial detection. Traditional spectrometers usually use gratings, prisms, or interference optical paths for spectral dispersion or encoding, and their size, assembly requirements, and applicable scenarios are affected by the optical structure and detector configuration. With the development of portable detection and integrated sensing, realizing spectral encoding, photoelectric conversion, and signal processing within a limited chip area has become an important direction for on-chip spectroscopic detection. This requires both reducing the size of the detection structure and readout system and retaining enough response information to distinguish the spectrum to be measured.
[0003] Computational spectrometers perform weighted measurements of incident light using a set of known response functions, and then estimate the spectrum based on the calibration response and reconstruction algorithm. The response differences can be provided by different filter structures, micro-nano optical structures, or multiple detector pixels, or they can be obtained by changing the operating state of the same detector. The former mainly utilizes the responses of different channels in space, while the latter utilizes the responses of different states in time. Using time-division measurement of a single photosensitive area can reduce the number of independent detector channels, but requires the spectrum to be measured to be basically stable during the scanning period. Therefore, on-chip design needs to consider not only the configuration of the encoding structure and readout channels, but also whether the obtained response contains enough information to support the target reconstruction task. The number of channels or states cannot be directly equated with the amount of recoverable information.
[0004] Chinese patent document CN109642822B, "Spectrometer and Spectral Measurement Method Thereof," involves first and second unit spectral filters and corresponding photodetectors, reconstructing the spectrum through signals from different filter channels. This approach utilizes the filter to provide response differences, requiring the system to be configured with corresponding optical coding structures and detection channels. If the method of adding independent filter channels to extend the measurement response is adopted, corresponding optical path coupling and signal acquisition must be arranged simultaneously, making it difficult to complete the response configuration solely through electrical control within the same detection area. Its limitations should be judged in conjunction with the channel implementation method, rather than simply categorizing all filter schemes as large in size or low in accuracy.
[0005] US Patent Application Publication US20220128408A1, "Multi-pixel spectral sensor," describes a multi-pixel sensor comprising a reflective structure, a phase adjustment element, and a detection element, which utilizes the wavelength response of different pixels for spectral analysis. This approach uses multiple spatial pixels to carry different responses, providing parallel measurement information, but still requires configuring corresponding photosensitive areas for these pixels and acquiring their outputs. For applications that wish to maintain a single continuous photosensitive area and overall readout, directly using this encoding method still relies on spatial pixel division and cannot simply rely on merging readouts to retain the original channel information, which differs from the requirement of using the same area to time-division change the response.
[0006] Chinese patent document CN111811651B, "Spectroscopic Chip, Spectrometer, and Method for Fabricating Spectroscopic Chip," involves setting a light modulation layer formed of metal and dielectric on the surface of an image sensor, and using a two-dimensional grating micro / nano structure unit array to encode spectral information to different pixels. The response difference depends on the coordination between the surface light modulation structure and pixel sampling, i.e., optical encoding is performed before light is detected. For the requirement to configure the response through the electrical operating state of the detector, a corresponding electrical modulation path needs to be established. The existing light modulation layer itself cannot replace the control of the carrier collection process. The above comparison focuses on the encoding position and implementation structure, and does not determine the performance superiority or inferiority of different schemes based on this.
[0007] Different detection responses can be obtained by changing the depletion region, carrier collection, or internal gain by applying an external bias. When only a single overall operating bias is adjusted, the potential of each region changes with the same control quantity, which limits the independently set spatial control boundary. If the response change within the target band is mainly manifested as approximately uniform amplitude scaling, it is difficult to increase the independent spectral information by increasing the bias state. This limitation depends on the device and operating conditions. Not all bias-adjustable detectors can only change the amplitude. Therefore, it is necessary to focus on whether it is possible to control the collection conditions at different locations separately while maintaining the overall readout of the same photosensitive area, and to confirm that the resulting wavelength-weighted differences are repeatable.
[0008] This invention employs multiple electrically isolated control electrodes within the same photoelectric conversion region. By altering the local potential boundary through bias combinations, the carrier collection weights at different generation locations become adjustable spectral coding factors. Compared to methods relying on multiple filter channels or spatial pixels, this structure utilizes the same photosensitive area and the same measurement output to acquire the response in a time-division manner, eliminating the need to add independent filter channels or detector pixels for response configuration. Compared to methods that only adjust the overall bias, multiple control electrodes provide separately configurable spatial control boundaries, independent of the signal collection end, allowing adjustment of local collection conditions while maintaining the detector end potential setting. When these adjustments are coupled with differences in generation distribution at different wavelengths and calibrated to obtain repeatable, non-redundant responses, additional effective measurement information can be provided for spectral reconstruction. This enables single-pixel spectral detection with an electrical configuration, reduces reliance on independent optical coding structures and multi-pixel readouts, increases the spectral information available for reconstruction, and facilitates the co-integration of the detector structure with the driving and readout circuits. Summary of the Invention
[0009] The purpose of this invention is to provide a single-pixel on-chip spectrometer based on local electric field modulation of a multi-electrode structure. Under the condition of maintaining the overall readout of the same photosensitive area, the carrier collection weight is adjusted by the locally settable potential boundary to form a calibrated distinguishable spectral response, and the incident spectrum is reconstructed through multi-state measurement.
[0010] The objective of this invention is achieved as follows:
[0011] The spectrometer consists of a single-pixel photodetector (1), a multi-electrode local electric field modulation unit (2), a programmable bias driving unit (3), a photoelectric signal acquisition unit (4), and a spectrum reconstruction processing unit (5). Incident light illuminates the detector (1), and the driving unit (3) applies voltage to each control electrode of the modulation unit (2) to make the detector (1) output photoelectric signals under different local electric field states. The acquisition unit (4) sends the signal to the processing unit (5), and the processing unit (5) reconstructs the spectrum according to the calibration response of the corresponding state and provides state selection information to the driving unit (3). Figure 1 The connection relationship in the middle distinguishes between measurement output and control information: the acquisition unit (4) transmits sampling data and status number to the processing unit (5), the processing unit (5) provides status selection information to the driving unit (3), and the driving unit (3) provides sampling synchronization information to the acquisition unit (4). The above control information is used to coordinate bias switching and sampling timing, and does not constitute an additional spectral measurement channel.
[0012] The single-pixel photodetector (1) has a single photosensitive area (11) that is continuous and used as a whole, including a semiconductor photoelectric conversion area (12), a signal collection end (13) and a return end (14), wherein the signal collection end (13) provides a spectral measurement output, and the return end (14) constitutes a return and reference end. Multiple control electrodes do not form independent spatial imaging pixels. When using a silicon photodiode, an n-type collection area can be set in the p-type photoelectric conversion area to form a photosensitive junction and lead out the signal collection end (13). The return end (14) is led out through the ohmic contact of the p-type area. This return contact is not used as a second spectral readout junction.
[0013] The control electrodes of the modulation unit (2) are located at different lateral positions on the same side of the photoelectric conversion region (12), separated from it by the dielectric layer (25), and are separate from the signal collection end (13) and the return end (14); the electrodes can be independent strip gates or staggered comb gates, formed of polysilicon or metal; each finger of the same comb gate shares a potential, different gates are independently connected to the driving unit (3), there is a gap between adjacent gates, and the electrodes and wiring leave a light path. The geometric dimensions, dielectric layer and bias voltage range are determined according to the device and process conditions; Figure 2 (a) and Figure 2 (b) represent two alternative layouts: strip grid and comb grid, respectively. Figure 3 Corresponding to the cross section of the strip gate, while multiple fingers of the comb gate are used to form the same potential control region, the number of fingers is not equal to the number of independent control terminals; both layouts maintain the same photosensitive area for continuous use and separate the control bias application path from the signal collection path, so as to change the local potential boundary while maintaining the working setting of the detection circuit; the gate potential is coupled to the semiconductor through the dielectric layer (25), forming an electric field that varies with position in the volume below the gate, below the gate gap and in its vicinity, the range of action depends on the gate layout, doping, dielectric layer and bias, and the lateral arrangement of the electrodes is not equal to adjusting only one lateral plane.
[0014] Let the number of independent control terminals be J, and the bias state of the k-th electrode be... = ( , …, ),in The reference potential of the j-th control electrode relative to the return terminal The voltage; K states are selected for one reconstruction; the signal collection terminal (13) maintains the set potential. , and The difference determines the operating bias voltage of the photojunction. Under this setting, changing the potential difference between adjacent control electrodes can adjust the potential boundary and carrier collection conditions in the region below and adjacent to the gap.
[0015] Under the approximation of linearity, no internal multiplication, and photogenerated carrier generation distribution that does not change with the control bias voltage, the spectral responsivity of the k-th state can be expressed as: (1)
[0016] In equation (1), λ is the wavelength, q is the elementary charge, Ω is the semiconductor volume involved in photoelectric conversion, x is the spatial position, g(x,λ) is the amount of photogenerated electron-hole pairs generated per unit volume per unit time under unit incident light power, and η(x; ) represents the carrier collection probability that contributes to the output in the corresponding state. The unit is ampere per watt; in the band where the absorption coefficient of the selected material varies with wavelength, different wavelengths can have different depths of generation distribution. These distributions overlap and do not correspond to mutually isolated wavelength regions; the local electric field changes the transport conditions of charge carriers to the collection region. When this change affects the competition between collection and recombination, the collection probability at the corresponding location can also change; if the spatial variation of the collection probability forms different weights for the generation distribution at different wavelengths, the overall readout response may show a non-proportional change, rather than directly switching a certain wavelength; this encoding requires effective coupling between the electric field region and the difference in generation distribution. If the charge carriers in the relevant region are almost completely collected in all states, or the change only forms a common scaling factor, then simply changing the path or adding electrodes is insufficient to provide new spectral information; whether the response has this difference should be confirmed by repeated calibration within the target band.
[0017] The driving unit (3) includes a controller, a digital-to-analog converter circuit, and an electrode driving circuit. It switches the bias voltage according to the status table and triggers sampling after stabilization. The acquisition unit (4) uses a transimpedance amplifier or integral sampling circuit and an analog-to-digital converter circuit to convert the signal into an equivalent photocurrent according to the gain and sampling time. The status number is saved with the data, and the calibration and measurement are kept the same. , Read out the configuration and lighting geometry.
[0018] The photocurrent at state k is measured by calibrating wavelength by wavelength using monochromatic light or sufficiently narrow-band light. (λ), dark signal and the non-zero optical power reaching the photosensitive area (11) (λ), and determine the responsivity according to formula (2); the calibration power here refers to the actual light power reaching the photosensitive area, and cannot be directly replaced by the nominal output power of the light source; the optical path coupling and illumination range should be consistent with the calibration record, and the dark signal should correspond to the bias state and readout setting, so as not to mistake the state-related dark current change for the spectral response change. The formula for calculating the responsivity is: (2)
[0019] The working band is divided into N non-overlapping intervals. Width is The total optical power within each interval For the unknown quantity, a piecewise constant approximation of the spectral power density within the channel is used, and the elements of the response matrix are the interval average responsivity: (3)
[0020] R is a K×N matrix, with rows corresponding to states and columns corresponding to wavelength channels. The unit of each element is amperes per watt. Within a sufficiently narrow channel, the response approximation (3) representing the wavelength can be used. The dark current of the light under test is obtained. = - This forms the measurement vector y and the spectral vector F, establishing: (4)
[0021] In equation (4), y and ε are K-dimensional column vectors in amperes; F is a vector composed of... The resulting N-dimensional column vector, with units of watts; ε includes noise, drift, calibration, and discretization errors; due to This represents the total power of the channel; it will not be multiplied repeatedly in RF. The processing unit (5) can obtain the estimated spectrum by solving the non-negative regularization problem. : (5)
[0022] In equation (5), α is a non-negative regularization parameter, selected based on the known verification spectrum or noise level; L is the identity matrix or difference operator. When constraining spectral density smoothing is required under non-equal width channels, L acts on... / Calibration should check linearity and response repeatability; if the bandwidth of a finite light source cannot be ignored, its spectral shape should be taken into account; measurements should be performed in the calibrated state, and when selecting a subset, the matrix rows should be extracted synchronously and arranged in the same order; the spectrum to be measured should be basically stable during a single scan, and the reconstruction result should be obtained through the predicted signal R. The residual test with y; the state table stores the correspondence between the control electrode bias combination and the state number, and the response matrix stores the calibration response of these states on each wavelength channel. When selecting a subset of states, the dark signal, sampling results and matrix rows should be selected simultaneously. The original row order should not be used while only the measurement order is changed; the residual reflects the degree of conformity of the model to the measurement signal. Its small value does not prove the uniqueness of the spectral solution on its own; when the response difference is insufficient or the noise influence is large, it is still necessary to combine the calibration repeatability and the verification spectrum to judge the reconstruction result.
[0023] This invention combines independent gate control, overall readout of the same photosensitive area, and state matching reconstruction, so that spectral coding is achieved by electrical adjustment of spatial collection weights; multiple control electrodes provide combinable potential boundaries, and the control path is separate from the detector end, so that local modulation can be performed when the detector end potential is kept set, and the same measurement output corresponds to different calibrated states, without the need to set a separate detector pixel for each state; the detector (1) and the modulation unit (2) are set on the same chip, and the units (3) to (5) can be integrated or externalized according to conditions; when the response difference is repeatable and higher than the measurement uncertainty, and the state set can support the target reconstruction task, the above structure can reduce the dependence on independent optical coding structure and multi-pixel readout. The actual resolution, power consumption and CMOS adaptation still need to be evaluated in combination with devices, circuits and processes. Attached Figure Description
[0024] Figure 1 This is a schematic diagram of the system structure of a single-pixel on-chip spectrometer. The diagram shows the connection of the detector (1), modulation unit (2), driving unit (3), acquisition unit (4) and processing unit (5), as well as the transmission relationship of incident light, control bias voltage, sampling synchronization and measurement data. The signal is output from the signal collection terminal (13), and the return terminal (14) provides return current and reference potential.
[0025] Figure 2 This is a top view schematic diagram of the control electrode layout, in which... Figure 2 (a) shows four independent strip grids (21) to (24) and their leads arranged sequentially at intervals within the same photosensitive area (11); Figure 2 (b) shows the staggered arrangement of the fingers of the two comb gates (21) and (22), the electrical connection within the group, and the isolation relationship between the groups; the signal collection end (13) and the return end (14) are located at the edge of the region, and A-A is Figure 2 (a) The corresponding sectioning position.
[0026] Figure 3 For along Figure 2 (a) is a schematic diagram of the device cross section in the A-A direction. The diagram shows the contact relationship between the control electrodes (21) to (24), the dielectric layer (25), the photoelectric conversion region (12), the signal collection end (13), and the return end (14). It shows the structure of the gate being isolated from the semiconductor and the collection end and the return end being connected to the detection circuit respectively. It also shows the direction of light entering the photoelectric conversion region through the gate gap.
[0027] Figure 4 For the same device in two candidate bias states , The diagram below illustrates the local electric field and carrier collection principle. The two sub-plots use the same structure and illumination to show the overlapping distribution of different wavelengths and the local modulation range. The plotted range and generation location are qualitative descriptions and do not represent simulation results. The differences in response between states are based on calibration.
[0028] Figure 5 This is a schematic diagram of the bias state switching and spectral response calibration process. The diagram shows the relationship between wavelength scanning and state traversal, stabilization waiting, dark signal acquisition, photosensitive surface light power measurement, responsivity calculation, channel discretization and state screening. The resulting matrix is arranged with states as rows and wavelength channels as columns, and is stored together with the state table and readout configuration.
[0029] Figure 6 This is a schematic diagram of the spectral measurement and reconstruction process. The diagram shows the sampling of the light under test in each state, the darkening and scale conversion, the matching of the measurement vector and matrix rows according to the state number, the non-negative regularization solution, and the prediction residual check. The output is the channel optical power and its corresponding spectral density. When using a subset of states, the measurement vector and matrix are adjusted simultaneously. Detailed Implementation
[0030] The present invention will be further illustrated below with reference to specific embodiments.
[0031] This embodiment adopts Figures 1 to 3 The four-gate silicon photodiode structure has an n-type collection area set in the p-type photoconversion area (12) of the continuous photosensitive area (11) and a signal collection terminal (13) led out. The return terminal (14) is led out through the ohmic contact of the p-type area. The strip gates (21) to (24) are arranged on the dielectric layer (25) at intervals and are respectively connected to the driving unit (3). The light path is left between the gates. The signal collection terminal (13) is connected to the acquisition unit (4) and the return terminal (14) is connected to the common reference terminal.
[0032] During operation, the return end (14) remains The acquisition unit (4) holds the signal collection end (13) at the position. The two are set with working bias voltage according to the polarity of the photosensitive junction; the driving unit (3) is based on... Using this as a reference, within the device's allowable range, the remaining gate voltages are first fixed, then the bias voltage is changed gate by gate, and finally the combination of two gates changing simultaneously is selected as the candidate state. To change the local collection conditions of the photoelectric conversion zone (12); after switching, continuous sampling is performed until the output change drops to within the stability criterion set according to the repeated measurement fluctuation, and then the signal and state number are recorded. States with saturation or continuous drift are not entered into subsequent calibration.
[0033] During calibration, according to Figure 5 First, block out light and traverse all states to obtain dark signals. Then, a single photosensitive area is illuminated with monochromatic light wavelength by wavelength (11), and the light power reaching the photosensitive surface is measured simultaneously. (λ); Candidate bias voltage combinations are applied sequentially at each wavelength, and the photocurrent is collected. (λ), by subtracting dark and normalizing according to equation (2), the spectral responsivity of each state is obtained.
[0034] The processing unit (5) evaluates the response fluctuation based on repeated calibration, compares the differences between candidate responses after deducting the optimal scaling ratio, and eliminates redundant states whose shape differences do not exceed the measurement uncertainty; it checks the overall information redundancy by combining the singular values and noise scales of the candidate matrix, and uses known spectra that have not participated in the state screening to test the selected set, retaining K states that meet the target task; it forms a K×N matrix R according to formula (3), and saves the state table, dark signal and readout configuration. If the verification does not meet the requirements, it adjusts the candidate bias or narrows the reconstruction task range.
[0035] During measurement, the light to be measured is coupled to a single photosensitive area (11), maintaining the illumination range during calibration. , and readout gain; press Figure 6 The selected states are called sequentially to obtain the dimming current. The matrix rows are matched according to the state number, and the processing unit (5) solves for the non-negative spectral vector according to equations (4) and (5). Output the total optical power of each channel; divide by the corresponding channel width to obtain the segmented spectral power density.
[0036] The spectrum under test is kept basically stable during a single scan, and the predicted signal R is used. The residual check results with the measured vector y; when an anomaly occurs, check for state mismatch, drift or saturation; after the working conditions affecting the response change, recalibrate or call up the calibration data under the corresponding conditions.
Claims
1. A single-pixel on-chip spectrometer based on local electric field modulation of a multi-electrode structure, characterized in that: It consists of a single-pixel photodetector (1), a multi-electrode local electric field modulation unit (2), a programmable bias driving unit (3), a photoelectric signal acquisition unit (4), and a spectral reconstruction processing unit (5); in the system, the detector (1) has a continuous and integrally read-out single photosensitive area (11), a semiconductor photoelectric conversion area (12), a signal collection end (13), and a return end (14); the modulation unit (2) includes at least two electrically isolated control electrodes, each control electrode being laterally spaced on the same side of the photoelectric conversion area (12), and connected via a dielectric... The electric layer (25) is separated from it and is independent of the signal collection terminal (13) and the return terminal (14); the driving unit (3) is connected to each control electrode respectively, and changes the local electric field and carrier collection weight by switching the bias combination to obtain a multi-state spectral response that is not completely proportional; the acquisition unit (4) collects the overall photoelectric signal of the same photosensitive area through the signal collection terminal (13) in a time-division manner, and the return terminal (14) provides the return and reference potential; the processing unit (5) corresponds the sampling data with the bias state and reconstructs the incident spectrum according to the matched calibration response matrix.
2. The single-pixel on-chip spectrometer based on multi-electrode structure local electric field modulation according to claim 1, wherein the multi-electrode local electric field modulation unit (2) used in the system is an insulated gate modulation device, characterized in that: The control electrode uses four independent strip grids, namely the first control electrode (21), the second control electrode (22), the third control electrode (23) and the fourth control electrode (24), or at least two sets of staggered comb grids; each strip grid is connected to the independent output of the driving unit (3), the fingers of each set of comb grids are electrically connected within the set and staggered between sets, and the whole set is an independent control terminal; light paths are left between adjacent electrodes and their wiring, and the dielectric layer (25) isolates the gate and the photoelectric conversion area (12); the signal collection terminal (13) and the return terminal (14) are located at the edge of the single photosensitive area (11) and are respectively connected to the detection circuit, and multiple control electrodes jointly modulate the same photoelectric conversion area (12) without forming separate detection pixels.
3. The single-pixel on-chip spectrometer based on local electric field modulation of a multi-electrode structure according to claim 1, characterized in that: The driving unit (3) includes a controller, a digital-to-analog converter circuit and an electrode driving circuit. The voltage combination of each control electrode relative to the reference potential of the return terminal (14) is set according to the state table. The calibration and measurement keep the potential setting and readout configuration of the signal collection terminal (13) and the return terminal (14) consistent. After the bias voltage is switched, wait for the response to stabilize and trigger the acquisition unit (4) to sample. The acquisition unit (4) includes a transimpedance amplifier circuit or an integral sampling circuit and an analog-to-digital converter circuit. The output is converted into an equivalent photocurrent according to the gain and sampling time, and the state number is saved along with the sampling data so that the measurement results of each state correspond one by one with the calibration response.
4. The single-pixel on-chip spectrometer based on local electric field modulation of a multi-electrode structure according to claim 1, characterized in that: The processing unit (5) uses the illumination current obtained by sampling wavelength by wavelength and bias voltage by narrowband light, subtracts the corresponding dark signal of the state and divides it by the calibrated light power reaching the photosensitive area (11) to obtain the state responsivity, and then calculates the average response of each wavelength channel to establish a response matrix with the state as the row and the wavelength channel as the column; the signal obtained by the light under test according to the same state table is darkened and the scale is unified, the measurement vector and matrix rows are arranged according to the state number, and the total light power of each channel is used as the non-negative unknown quantity to solve the optimization problem containing the measurement residual term and the regularization term, outputs the channel light power and its corresponding segmented spectral density, and checks the reconstruction result by the residual between the predicted signal and the measured signal.
Citation Information
Patent Citations
Spectrometer and spectrum measurement method using the same
CN109642822B
Spectral chip, spectrometer and spectral chip preparation method
CN111811651B
Multi-pixel spectral sensor
US20220128408A1
