Method and apparatus for detecting a semiconductor temperature field

Abstract

The application relates to the technical field of semiconductor detection, and provides a semiconductor temperature field detection method and device. The method comprises the following steps: point-by-point scanning of a wafer surface to obtain radiation information of each spatial position point on the wafer surface; radiation energy of different wavelengths is separated from each spatial position point in sequence according to the radiation information; radiation data of each spatial position point under multiple wavelengths is generated based on the radiation energy of each wavelength; the time sequence of the point-by-point scanning and the separation of the radiation energy of each wavelength is coordinated, so that each spatial position point matches corresponding multi-wavelength radiation data; the multi-wavelength radiation data is obtained, a radiation equation group is constructed based on the multi-wavelength radiation data of the same spatial position point, and the temperature value of each point on the wafer surface is calculated by synchronous inversion according to the radiation equation group. The application is used for the semiconductor manufacturing process, and can realize self-adaption and high-reliability non-contact detection of the surface temperature field without relying on preset parameters of emissivity.

Description

Technical Field

[0001] This invention relates to the field of semiconductor detection technology, and in particular to a method and apparatus for detecting semiconductor temperature fields. Background Technology

[0002] In semiconductor manufacturing processes, accurate temperature field detection is crucial for process quality control. Current non-contact infrared thermometry faces a key challenge: the emissivity of semiconductor materials is influenced by multiple factors, including intrinsic wafer characteristics, surface oxidation state, microstructure, and dynamic process dynamics, exhibiting significant unknowns, spatial non-uniformity, and time-varying characteristics. Traditional single-wavelength temperature measurement methods rely on preset emissivity parameters for temperature calculation. When the actual emissivity deviates from the preset value, it directly introduces systematic temperature measurement errors, leading to inaccurate process monitoring.

[0003] To reduce emissivity dependence, dual-wavelength thermometry eliminates the emissivity parameter by using the ratio of radiation intensity of specific wavelength combinations. However, this method relies on the theoretical assumption that "emissivity satisfies a simplified functional relationship at the selected wavelength." Semiconductor surfaces generally exhibit complex spectral selectivity in the infrared band, making this assumption difficult to apply in most practical manufacturing scenarios, and the inversion results are still affected by model mismatch. While multispectral thermometry has the potential to simultaneously solve for temperature and emissivity, existing implementation paths suffer from structural defects: Area array multispectral imaging schemes are not only costly but also struggle to achieve sub-pixel-level spatial registration between images of different bands, leading to spatial misalignment of multi-wavelength radiation data at the same physical location; schemes based on mechanical scanning and temporal filtering lack a precise temporal coordination mechanism for scanning position and wavelength switching, resulting in temporal shifts in the radiation acquisition times of the same spatial point at different wavelengths. When the measured surface has a temperature gradient or is undergoing dynamic changes, this shift will cause data mismatch and motion blur, undermining the spatiotemporal consistency foundation necessary for constructing the radiation equations and rendering temperature inversion physically unreliable.

[0004] Therefore, there is an urgent need for a method and apparatus for detecting semiconductor temperature fields to improve the above-mentioned problems. Summary of the Invention

[0005] This invention provides a method and apparatus for detecting semiconductor temperature fields. This invention is used in semiconductor manufacturing processes to perform adaptive, highly reliable, non-contact detection of surface temperature fields without relying on preset emissivity parameters.

[0006] According to a first aspect of the present invention, a method for detecting a semiconductor temperature field is provided, applied to a semiconductor temperature field detection device. The method includes: scanning a wafer surface point by point to acquire radiation information at each spatial location point on the wafer surface; sequentially separating radiation energy of different wavelengths based on the radiation information at each spatial location point; generating radiation data at each spatial location point at multiple wavelengths based on the radiation energy of each wavelength; coordinating the timing of point-by-point scanning and separation of radiation energy of each wavelength so that each spatial location point matches its corresponding multi-wavelength radiation data; acquiring multi-wavelength radiation data; constructing a radiation equation set based on the multi-wavelength radiation data of the same spatial location point; and synchronously inverting and calculating the temperature value at each point on the wafer surface based on the radiation equation set.

[0007] In one implementation, coordinating the timing of point-by-point scanning and separation of radiation energy at each wavelength includes: performing spatial scanning according to a preset scanning path; acquiring the actual feedback value of the scanning position in real time and converting the actual feedback value into spatial coordinates through a geometric calibration relationship; continuously detecting the filter wheel angle and converting the filter wheel angle into the corresponding wavelength parameter through a preset mapping relationship; generating a sampling trigger signal when it is determined that the scanning position has reached a stable state at the target point and the center of the filter exactly passes through the optical axis; acquiring radiation signals based on the sampling trigger signal and associating the radiation signals with corresponding spatial coordinates and wavelength parameters to establish a correspondence between spatial position and multi-wavelength radiation data.

[0008] In one implementation, before constructing the radiation equations, the data preprocessing steps include: acquiring raw data containing spatial coordinates, wavelength labels, and raw radiation signals; subtracting dark noise at the corresponding wavelength from the raw radiation signals to obtain net radiation signals; correcting the net radiation signals based on a pre-calibrated system spectral response function to obtain accurate radiation intensity values ​​at each wavelength; and generating a corrected multi-wavelength radiation intensity dataset for each spatial location point as input data for constructing the radiation equations.

[0009] In one implementation, the temperature values ​​at various points on the wafer surface are simultaneously calculated based on the radiation equations, including: establishing an emissivity wavelength model for each spatial location point, where the emissivity wavelength model is expressed as a linear function of wavelength; constructing a theoretical radiation intensity model based on Planck's radiation law and the emissivity wavelength model; defining an optimization criterion with the goal of minimizing the sum of squared deviations between the theoretical and measured radiation intensities; solving for the temperature parameters and emissivity model coefficients using an iterative optimization method, which includes parameter initialization, calculation of update amounts, parameter updates, and convergence judgment steps; terminating the iteration when the change in iterative parameters or the residual is less than a preset threshold, and outputting the temperature value and emissivity model parameters for that spatial location point; and calculating the emissivity value of that spatial location point at each wavelength based on the emissivity model parameters.

[0010] In one embodiment, after synchronously inverting and calculating the temperature values ​​of each point on the wafer surface according to the radiation equations, the method further includes: integrating the temperature values ​​of all spatial points on the wafer surface based on the multi-wavelength radiation data of the same spatial location point and the emissivity values ​​of each point on the wafer surface calculated by inversion, to generate full-field temperature distribution data; performing spatial interpolation processing on the full-field temperature distribution data to form a continuous temperature field distribution; converting the continuous temperature field distribution into a visual image for output; and generating an emissivity distribution map of the wafer surface based on the emissivity values ​​of each point on the wafer surface calculated by inversion, to characterize the state characteristics of the wafer surface.

[0011] In one implementation, dark noise at the corresponding wavelength is subtracted from the original radiation signal to obtain the net radiation signal. The following formula is used for calculation:

[0012] ;

[0013] in, It is the base voltage of the corresponding wavelength channel measured by blocking the optical path. The original radiation signal value is used; based on the pre-calibrated system spectral response function, the net radiation signal is corrected to obtain the radiation intensity value at each wavelength. The following formula is used for calculation:

[0014] ;

[0015] in, The system obtained through blackbody furnace calibration at wavelength The spectral response function under the given conditions.

[0016] In one implementation, the theoretical radiation intensity The model is constructed using the following expression:

[0017] ;

[0018] in, This indicates that an ideal blackbody is at a wavelength and the spectral radiance at temperature T, and These are the emissivity model coefficients to be inverted.

[0019] According to a second aspect of the present invention, a semiconductor temperature field detection device is provided. The device includes a spatial scanning mechanism, a wavelength timing separation mechanism, a single-point radiation detection unit, a synchronization control mechanism, and a data processing mechanism. The spatial scanning mechanism is used to scan the wafer surface point by point to acquire radiation information at each spatial location point on the wafer surface. The wavelength timing separation mechanism is disposed in the optical path of the spatial scanning mechanism and is used to sequentially separate radiation energy of different wavelengths according to the radiation information of each spatial location point. The single-point radiation detection unit is used to receive the radiation energy of each wavelength separated by the wavelength timing separation mechanism and generate radiation data of each spatial location point at multiple wavelengths. The synchronization control mechanism is used to coordinate the working timing of the spatial scanning mechanism and the wavelength timing separation mechanism so that each spatial location point accurately matches its corresponding multi-wavelength radiation data. The data processing mechanism is used to receive the multi-wavelength radiation data generated by the single-point radiation detection unit, construct a radiation equation set based on the multi-wavelength radiation data of the same spatial location point, and synchronously calculate the temperature value of each point on the wafer surface according to the radiation equation set.

[0020] In one embodiment, the wavelength timing separation mechanism includes a filter wheel, a rotation drive unit, and an angle detection unit. The filter wheel is equipped with multiple narrowband filters with different center wavelengths. When the filter wheel rotates, the narrowband filters with different wavelengths enter the optical path sequentially, performing wavelength selective separation of the radiation information at the same spatial location. The rotation drive unit is used to drive the filter wheel to rotate at a constant angular velocity. The angle detection unit is installed coaxially with the filter wheel to continuously detect the angle of the filter wheel.

[0021] In one embodiment, the single-point radiation detection unit has an adjustable-gain preamplifier circuit whose gain is dynamically adjusted according to the radiation intensity to adapt to the intensity differences of radiation energy at different wavelengths.

[0022] Compared with existing technologies, the advantages of this invention are as follows: This invention, through a collaborative design of point-by-point scanning and wavelength separation, ensures that each spatial location point obtains strictly corresponding multi-wavelength radiation data, laying a data foundation for constructing a physically complete set of radiation equations. Based on this, a set of radiation equations is constructed using the multi-wavelength radiation data from the same spatial point, and synchronous inversion calculations are performed. This decouples the temperature solution process from the emissivity characteristics, eliminating the need for preset or assumed emissivity values. Instead, it relies on the inherent physical correlation between multi-wavelength radiation intensities (i.e., following Planck's radiation law), implicitly handling the influence of emissivity variables on radiative transmission during the inversion process. This frees the temperature calculation results from dependence on prior knowledge of emissivity. Thus, temporal coordination ensures the spatiotemporal consistency of the input data, multi-wavelength data provides sufficient conditions for solving the equations, and the synchronous inversion mechanism enables direct solution of temperature parameters. This fundamentally eliminates the systematic errors introduced by emissivity uncertainty, making the detection of semiconductor surface temperature fields adaptive and reliable, suitable for monitoring semiconductor manufacturing processes with dynamically changing emissivity and complex surface conditions. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a method for detecting a semiconductor temperature field according to an exemplary embodiment.

[0024] Figure 2 This is a wireframe diagram illustrating a semiconductor temperature field detection device according to an exemplary embodiment.

[0025] Explanation of the reference numerals in the figure:

[0026] 1. Semiconductor temperature field detection device; 2. Spatial scanning mechanism; 3. Wavelength timing separation mechanism; 4. Single-point radiation detection unit; 5. Synchronization control mechanism; 6. Data processing mechanism. Detailed Implementation

[0027] Unless otherwise defined, the technical or scientific terms used in this specification should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. Specific embodiments of the invention will be described below with reference to the accompanying drawings. It should be noted that, in order to provide a concise description, this specification cannot provide a detailed description of all features of the actual embodiments. Without departing from the spirit and scope of the invention, those skilled in the art can make modifications and substitutions to the embodiments of the invention, and the resulting embodiments are also within the protection scope of the invention.

[0028] like Figure 1 As shown, the first embodiment of the present invention provides a method for detecting a semiconductor temperature field, applied to a semiconductor temperature field detection device, the method comprising the following steps S1-S5:

[0029] S1 scans the wafer surface point by point to obtain the radiation information of each spatial location on the wafer surface.

[0030] S2, based on the radiation information of each spatial location point, sequentially separates the radiation energy of different wavelengths.

[0031] S3 generates radiation data for each spatial location point at multiple wavelengths based on the radiation energy of each wavelength.

[0032] S4 coordinates the timing of point-by-point scanning and separation of radiation energy at each wavelength, so that each spatial location point matches its corresponding multi-wavelength radiation data.

[0033] S5. Acquire multi-wavelength radiation data, construct a set of radiation equations based on the multi-wavelength radiation data at the same spatial location, and simultaneously calculate the temperature value of each point on the wafer surface based on the set of radiation equations.

[0034] It is worth noting that, based on radiation data collected at each spatial location point at multiple wavelengths (e.g., no fewer than four), this invention simultaneously inverts and solves for the temperature value and emissivity model parameters of that point using a nonlinear optimization algorithm, achieving single-point temperature calibration based on a physical model and completely eliminating the dependence on a preset emissivity. On this basis, the inversion results from all points in the field are spatially integrated and interpolated to generate a continuous, high-resolution temperature field distribution image, and simultaneously outputs an emissivity distribution map characterizing the surface state. This technical approach enables each measurement point to obtain an independent and self-consistent physical solution, effectively overcoming the challenge of temperature measurement for targets with spatially non-uniform emissivity, such as unevenly oxidized metals, composite materials, or functional coating surfaces. It significantly improves the realism and engineering applicability of two-dimensional temperature field reconstruction, providing reliable state awareness data for precision processes such as semiconductor manufacturing.

[0035] In one implementation, coordinating the timing of point-by-point scanning and separation of radiation energy at each wavelength includes: performing spatial scanning according to a preset scanning path; acquiring the actual feedback value of the scanning position in real time and converting the actual feedback value into spatial coordinates through a geometric calibration relationship; continuously detecting the filter wheel angle and converting the filter wheel angle into the corresponding wavelength parameter through a preset mapping relationship; generating a sampling trigger signal when it is determined that the scanning position has reached a stable state at the target point and the center of the filter exactly passes through the optical axis; acquiring radiation signals based on the sampling trigger signal and associating the radiation signals with corresponding spatial coordinates and wavelength parameters to establish a correspondence between spatial position and multi-wavelength radiation data.

[0036] In some specific embodiments, the signal acquisition and synchronization control unit serves as the core synchronization module of the system, responsible for achieving precise timing coordination among spatial scanning, spectral separation, and radiation acquisition. This unit uses a Field-Programmable Gate Array (FPGA) as its core controller, and its specific workflow is as follows: The FPGA, based on a preset scanning path (such as a raster scanning mode), outputs a drive signal through a digital-to-analog converter to control the galvanometer to perform spatial scanning; it also synchronously reads the angle signal fed back by the galvanometer encoder. And based on geometric calibration relationships Real-time calculation of spatial coordinates; continuous monitoring of the filter wheel encoder angle. An angle is mapped to the center wavelength of the corresponding filter using a built-in lookup table. When the galvanometer position is determined to be stable at the target point (position deviation less than the threshold) and the filter center is precisely aligned with the optical axis, the FPGA instantly generates a high-precision sampling trigger pulse. This pulse triggers the analog-to-digital converter to sample the radiation signal and compares the sampled value with the currently calculated spatial coordinates (x, y) and wavelength parameters. Precise association and binding are performed to form data units with complete spatiotemporal-spectral identification. These data units are transmitted to the host computer in real time via a gigabit Ethernet interface using the UDP protocol, providing strictly corresponding multi-wavelength radiation input data for the synchronous inversion of the temperature field, thus fundamentally ensuring the spatiotemporal consistency required for the modeling of the radiation equations.

[0037] In one implementation, before constructing the radiation equations, the data preprocessing steps include: acquiring raw data containing spatial coordinates, wavelength labels, and raw radiation signals; subtracting dark noise at the corresponding wavelength from the raw radiation signals to obtain net radiation signals; correcting the net radiation signals based on a pre-calibrated system spectral response function to obtain accurate radiation intensity values ​​at each wavelength; and generating a corrected multi-wavelength radiation intensity dataset for each spatial location point as input data for constructing the radiation equations.

[0038] In some specific embodiments, before constructing the radiation equations, the system first performs fine preprocessing on the acquired raw data: receiving raw data frames from the FPGA, which contain spatial coordinates (x, y) and wavelength parameters. Original radiation signal value And timestamps; then dark noise subtraction is performed, that is, for each wavelength The original radiation signal value Subtract the pre-measured base voltage of the corresponding channel when the optical path is blocked. Obtain net radiation signal Subsequently, the system spectral response function obtained based on the blackbody furnace calibration was... Correcting the net radiation signal by means of The accurate radiation intensity values ​​at each wavelength were calculated; finally, a set of corrected multi-wavelength radiation intensity datasets was generated for each spatial location point P(x, y). This preprocessing procedure effectively eliminates measurement biases introduced by detector background noise and system response non-uniformity, ensuring high fidelity and physical consistency of the data input to the radiation equations, and providing a data foundation for subsequent synchronous inversion of temperature and emissivity.

[0039] In one implementation, the temperature values ​​at various points on the wafer surface are simultaneously calculated based on the radiation equations, including: establishing an emissivity wavelength model for each spatial location point, where the emissivity wavelength model is expressed as a linear function of wavelength; constructing a theoretical radiation intensity model based on Planck's radiation law and the emissivity wavelength model; defining an optimization criterion with the goal of minimizing the sum of squared deviations between the theoretical and measured radiation intensities; solving for the temperature parameters and emissivity model coefficients using an iterative optimization method, which includes parameter initialization, calculation of update amounts, parameter updates, and convergence judgment steps; terminating the iteration when the change in iterative parameters or the residual is less than a preset threshold, and outputting the temperature value and emissivity model parameters for that spatial location point; and calculating the emissivity value of that spatial location point at each wavelength based on the emissivity model parameters.

[0040] In some specific embodiments, the synchronous inversion of temperature and emissivity is performed independently for each spatial location point, including: firstly, establishing an emissivity wavelength model for that point, using a linear function of wavelength. (in and (The emissivity coefficient to be inverted; this model is applicable to the spectral characteristics of most engineering materials); then, based on Planck's radiation law, the ideal blackbody radiation intensity is... Combined with the emissivity model, a theoretical radiation intensity model is constructed. ,in Defined by Planck's formula:

[0041] ;

[0042] Where c1 and c2 are standard physical constants in the radiation law; the inversion process is transformed into a nonlinear least squares optimization problem, with the goal of minimizing the sum of squared deviations between theoretical and measured radiation intensities; the temperature parameter T and emissivity coefficient are solved through iterative optimization methods. , This method includes parameter initialization, calculation of update amounts, parameter updates, and convergence judgment. The calculation terminates when the parameter change or residual falls below a preset threshold during iteration, outputting the temperature value and emissivity model parameters for that spatial point. Finally, the emissivity value of that point at each measurement wavelength is calculated based on the obtained parameters. This process, through the synergy of physical model constraints and data-driven optimization, achieves coupled solution of temperature and emissivity, effectively avoiding dependence on preset emissivity and providing a high-confidence single-point solution for temperature field reconstruction.

[0043] In one embodiment, after synchronously inverting and calculating the temperature values ​​of each point on the wafer surface according to the radiation equations, the method further includes: integrating the temperature values ​​of all spatial points on the wafer surface based on the multi-wavelength radiation data of the same spatial location point and the emissivity values ​​of each point on the wafer surface calculated by inversion, to generate full-field temperature distribution data; performing spatial interpolation processing on the full-field temperature distribution data to form a continuous temperature field distribution; converting the continuous temperature field distribution into a visual image for output; and generating an emissivity distribution map of the wafer surface based on the emissivity values ​​of each point on the wafer surface calculated by inversion, to characterize the state characteristics of the wafer surface.

[0044] In some specific embodiments, step S3 involves the synthesis and output of the full-field temperature field. First, all scan points on the wafer surface are traversed, and the inversion calculation in step S2 is performed for each point to obtain the temperature matrix. and emissivity coefficient matrix , Key datasets, including those from [specific datasets]. Based on this data, the temperature distribution across the entire wafer surface was integrated and processed using a bilinear interpolation method. This generates a spatially continuous full-field temperature distribution map and converts it into an easily understandable pseudo-color image, where different colors represent different temperature ranges, making temperature field changes readily apparent. Furthermore, it utilizes the emissivity coefficient matrix... and It can calculate a specific wavelength (e.g.) Emissivity distribution at (λ = 1.2 μm) This process generates emissivity distribution maps, which are then used for in-depth analysis of the wafer surface's state characteristics. Ultimately, these visualizations not only provide spatial distribution information on wafer surface temperature and emissivity but also help identify potential hotspots or anomalous regions, providing crucial information for further research and practical applications. This process effectively transforms the measurement results of physical quantities into intuitive graphical representations, greatly enhancing the efficiency and accuracy of data analysis.

[0045] In one implementation, dark noise at the corresponding wavelength is subtracted from the original radiation signal to obtain the net radiation signal. The following formula is used for calculation:

[0046] ;

[0047] in, It is the base voltage of the corresponding wavelength channel measured by blocking the optical path. The original radiation signal value is used; based on the pre-calibrated system spectral response function, the net radiation signal is corrected to obtain the radiation intensity value at each wavelength. The following formula is used for calculation:

[0048] ;

[0049] in, The system obtained through blackbody furnace calibration at wavelength The spectral response function under the given conditions.

[0050] In one implementation, the theoretical radiation intensity The model is constructed using the following expression:

[0051] ;

[0052] in, This indicates that an ideal blackbody is at a wavelength and the spectral radiance at temperature T, and These are the emissivity model coefficients to be inverted.

[0053] like Figure 2 As shown, according to a second embodiment of the present invention, a semiconductor temperature field detection device 1 is provided, including a spatial scanning mechanism 2, a wavelength timing separation mechanism 3, a single-point radiation detection unit 4, a synchronization control mechanism 5, and a data processing mechanism 6. The spatial scanning mechanism 2 is used to scan the wafer surface point by point to obtain the radiation information of each spatial location point on the wafer surface. The wavelength timing separation mechanism 3 is disposed in the optical path of the spatial scanning mechanism 2 and is used to sequentially separate the radiation energy of different wavelengths according to the radiation information of each spatial location point. The single-point radiation detection unit 4 is used to receive the radiation energy of each wavelength separated by the wavelength timing separation mechanism 3 and generate radiation data of each spatial location point under multiple wavelengths. The synchronization control mechanism 5 is used to coordinate the working timing of the spatial scanning mechanism 2 and the wavelength timing separation mechanism 3 so that each spatial location point accurately matches its corresponding multi-wavelength radiation data. The data processing mechanism 6 is used to receive the multi-wavelength radiation data generated by the single-point radiation detection unit 4, construct a radiation equation set based on the multi-wavelength radiation data of the same spatial location point, and synchronously calculate the temperature value of each point on the wafer surface according to the radiation equation set.

[0054] In one embodiment, the wavelength timing separation mechanism 3 includes a filter wheel, a rotation drive unit, and an angle detection unit. The filter wheel is equipped with multiple narrowband filters with different center wavelengths. When the filter wheel rotates, the narrowband filters with different wavelengths enter the optical path in sequence, so as to perform wavelength selective separation of the radiation information at the same spatial location point. The rotation drive unit is used to drive the filter wheel to rotate at a constant angular velocity. The angle detection unit is installed in a coaxial position with the filter wheel to continuously detect the angle of the filter wheel.

[0055] In some specific embodiments, a time-modulated spectral acquisition architecture consisting of a single-point detector and a rotating filter wheel is employed. By rotating the filter wheel, different wavelength channels are sequentially switched in the time domain, achieving accurate acquisition of multispectral radiation information. This design eliminates the detector array and complex optical path required for traditional multi-channel parallel detection, significantly simplifying the system hardware structure and reducing hardware costs by an order of magnitude. The simplified structure also reduces the interference of environmental vibrations and temperature drift on the measurement link, improving the system's stability and environmental adaptability during long-term operation in industrial settings. Therefore, high-precision multispectral temperature field measurement technology overcomes the limitations of laboratory environments, demonstrating true engineering feasibility for reliable deployment and continuous application in harsh industrial scenarios such as semiconductor manufacturing and high-end equipment maintenance.

[0056] In other specific embodiments, efficient measurement is achieved through dual optimization of hardware timing and algorithm efficiency. The filter wheel rotates at a constant speed of 30–60 rpm, and the two-dimensional galvanometer performs high-speed scanning along a preset grating path. The FPGA synchronously analyzes the real-time feedback signals from the galvanometer position encoder and the filter wheel angle encoder. When the galvanometer is stable at the target point (deviation less than the threshold) and the center of the filter is precisely aligned with the optical axis, a single-point detector is instantly triggered to complete the wavelength radiation sampling. Within the single-point dwell window (approximately 5–15 milliseconds), data from 4–8 wavelength channels can be continuously acquired, achieving microsecond-level precise coordination of the three stages of "spatial positioning-wavelength switching-signal acquisition". The data processing end employs a lightweight nonlinear optimization algorithm, combined with monochromatic radiation initial value estimation and physical constraints, to compress the inversion calculation of single-point temperature and emissivity parameters to the millisecond level, and supports multi-point data parallel processing. This collaborative mechanism enables the system to achieve the engineering practicality of field-of-view temperature field reconstruction and dynamic thermal process monitoring while maintaining the low-cost advantage of the single-detector architecture, providing a reliable technical path for real-time perception of complex thermal processes in industrial sites. Although this embodiment employs a single-point scanning architecture, through the precise timing coordination of high-speed filter wheel rotation and galvanometer scanning, it can complete the acquisition of radiation data for no fewer than four wavelength channels within a millisecond-level window of single-point residence. The single-point spectral acquisition efficiency is several times higher than that of traditional single-wavelength scanning. Combined with the rapid positioning capability of the high-speed galvanometer, the system can complete the complete thermal field reconstruction of a typical field of view within seconds to tens of seconds, meeting the "near real-time" monitoring requirements in industrial scenarios. This solution effectively avoids the bottleneck of the time-consuming static multispectral scanning, while also avoiding the high cost and complex calibration problems of high-speed area array multispectral systems. It achieves an engineering-feasible optimized balance among measurement speed, system cost, and inversion accuracy, making high-precision temperature field monitoring of dynamic thermal processes such as combustion oscillation tracking, rapid semiconductor thermal processing, and transient thermal response of materials a realistic and feasible technical path.

[0057] In some specific embodiments, the synchronous inversion algorithm of this embodiment solves for the temperature field while simultaneously calculating the emissivity coefficient ( , As an endogenous variable coupled with temperature, emissivity distribution information is jointly optimized, making it a direct output of the inversion process rather than a post-processing derivation. Within the same iterative framework for single-point temperature calculation, the algorithm instantly generates an emissivity parameter distribution map or an emissivity distribution at a specified wavelength that is strictly spatially aligned with the temperature field. This mechanism ensures complete synchronization of temperature and emissivity data in terms of physical location, acquisition time, and inversion logic, completely avoiding the spatial misalignment or model mismatch risks that may be introduced by step-by-step processing. As a result, the temperature measurement system transcends the single temperature sensing function and is upgraded to a comprehensive diagnostic platform for "thermal state-surface characteristics": the emissivity distribution map can intuitively map the intrinsic properties of materials, oxidation degree, coating uniformity, and surface contamination, enabling operators to simultaneously and holistically assess the thermal uniformity and surface process quality of workpieces. This provides in-depth diagnostic evidence with both thermodynamic and materials science significance for precision processes such as semiconductor manufacturing and high-end equipment thermal management, significantly enhancing the scientific nature of process monitoring and the reliability of quality decisions.

[0058] In one embodiment, the single-point radiation detection unit 4 has an adjustable-gain preamplifier circuit whose gain is dynamically adjusted according to the radiation intensity to adapt to the intensity differences of radiation energy at different wavelengths.

[0059] In this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. The term "multiple" refers to two or more unless otherwise expressly defined.

[0060] The above description of the embodiments is intended to enable those skilled in the art to understand and apply the present invention. It will be apparent to those skilled in the art that various modifications can be made to these embodiments, and the general principles described herein can be applied to other embodiments without creative effort. Therefore, the present invention is not limited to the embodiments described herein, and any improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope and spirit of the invention are within the scope of the present invention.

Claims

1. A method for detecting a semiconductor temperature field, applied to a semiconductor temperature field detection device, characterized in that, The method includes: The wafer surface is scanned point by point to obtain the radiation information of each spatial location on the wafer surface; The radiation energy of different wavelengths is separated sequentially based on the radiation information of each spatial location point; Based on the radiation energy of each wavelength, radiation data for each spatial location point at multiple wavelengths is generated; The timing of the point-by-point scanning and separation of radiation energy at each wavelength is coordinated so that each spatial location point matches its corresponding multi-wavelength radiation data; The process involves acquiring multi-wavelength radiation data, constructing a set of radiation equations based on the multi-wavelength radiation data at the same spatial location, and simultaneously calculating the temperature values ​​at various points on the wafer surface using the radiation equations. This includes: establishing an emissivity wavelength model for each spatial location, where the emissivity wavelength model is expressed as a linear function of wavelength; constructing a theoretical radiation intensity model based on Planck's radiation law and the emissivity wavelength model; defining an optimization criterion with the goal of minimizing the sum of squared deviations between the theoretical and measured radiation intensities; solving for the temperature parameters and emissivity model coefficients using an iterative optimization method, which includes parameter initialization, calculation of update amounts, parameter updates, and convergence judgment steps; terminating the iteration when the change in iterative parameters or the residual is less than a preset threshold, and outputting the temperature value and emissivity model parameters for that spatial location; and calculating the emissivity value at each wavelength for that spatial location based on the emissivity model parameters.

2. The method according to claim 1, characterized in that, Coordinating the timing of the point-by-point scanning and separation of radiant energy at each wavelength includes: Spatial scanning is performed according to a preset scanning path; the actual feedback value of the scanning position is acquired in real time, and the actual feedback value is converted into spatial coordinates through geometric calibration relationship; the filter wheel angle is continuously detected, and the filter wheel angle is converted into the corresponding wavelength parameter through a preset mapping relationship; when it is determined that the scanning position has reached a stable state at the target point and the center of the filter just passes through the optical axis, a sampling trigger signal is generated; the radiation signal is collected based on the sampling trigger signal, and the corresponding spatial coordinates and wavelength parameters are associated with the radiation signal to establish the correspondence between spatial position and multi-wavelength radiation data.

3. The method according to claim 1, characterized in that, Before constructing the radiation equations, the following data preprocessing steps are also included: Collect raw data containing spatial coordinates, wavelength labels, and raw radiation signals; The net radiation signal is obtained by subtracting the dark noise at the corresponding wavelength from the original radiation signal; Based on the pre-calibrated system spectral response function, the net radiation signal is corrected to obtain accurate radiation intensity values ​​at each wavelength; A calibrated multi-wavelength radiation intensity dataset is generated for each spatial location point and used as input data for constructing the radiation equations.

4. The method according to claim 1, characterized in that, After simultaneously inverting and calculating the temperature values ​​at various points on the wafer surface based on the radiation equations, the process also includes: Based on the multi-wavelength radiation data of the same spatial location point and the emissivity values ​​of each point on the wafer surface calculated by inversion, the temperature values ​​of all spatial locations on the wafer surface are integrated to generate full-field temperature distribution data. Spatial interpolation is performed on the overall temperature distribution data to form a continuous temperature field distribution; The continuous temperature field distribution is converted into a visual image for output. Based on the emissivity values ​​of each point on the wafer surface obtained from the inversion calculation, an emissivity distribution map of the wafer surface is generated to characterize the state characteristics of the wafer surface.

5. The method according to claim 3, characterized in that, The net radiation signal is obtained by subtracting the dark noise at the corresponding wavelength from the original radiation signal. The following formula is used for calculation: ； in, It is the base voltage of the corresponding wavelength channel measured by blocking the optical path. This represents the original radiation signal value. The net radiation signal is corrected based on the pre-calibrated system spectral response function to obtain the radiation intensity values ​​at each wavelength. The following formula is used for calculation: ； in, The system obtained through blackbody furnace calibration at wavelength The spectral response function under the given conditions.

6. The method according to claim 1, characterized in that, The theoretical radiation intensity The model is constructed using the following expression: ； in, This indicates that an ideal blackbody is at a wavelength and the spectral radiance at temperature T, and These are the emissivity model coefficients to be inverted.

7. A semiconductor temperature field detection device, characterized in that, The device includes a spatial scanning mechanism, a wavelength timing separation mechanism, a single-point radiation detection unit, a synchronization control mechanism, and a data processing mechanism. The spatial scanning mechanism is used to scan the wafer surface point by point to obtain the radiation information of each spatial location on the wafer surface; The wavelength timing separation mechanism is set in the optical path of the spatial scanning mechanism and is used to sequentially separate the radiation energy of different wavelengths according to the radiation information of each spatial location point. The single-point radiation detection unit is used to receive the radiation energy of each wavelength separated by the wavelength timing separation mechanism, and generate radiation data of each spatial location point at multiple wavelengths. The synchronization control mechanism is used to coordinate the working timing of the spatial scanning mechanism and the wavelength timing separation mechanism, so that each spatial location point can accurately match its corresponding multi-wavelength radiation data. The data processing mechanism receives multi-wavelength radiation data generated by a single-point radiation detection unit, constructs a set of radiation equations based on the multi-wavelength radiation data at the same spatial location, and synchronously calculates the temperature values ​​at various points on the wafer surface according to the set of radiation equations. This includes: establishing an emissivity wavelength model for each spatial location, where the emissivity wavelength model is expressed as a linear function of wavelength; constructing a theoretical radiation intensity model based on Planck's radiation law and the emissivity wavelength model; defining an optimization criterion with the goal of minimizing the sum of squared deviations between the theoretical and measured radiation intensities; solving for temperature parameters and emissivity model coefficients using an iterative optimization method, which includes parameter initialization, calculation of update amounts, parameter updates, and convergence judgment steps; terminating the iteration when the change in iterative parameters or the residual is less than a preset threshold, and outputting the temperature value and emissivity model parameters for that spatial location; and calculating the emissivity value at each wavelength for that spatial location based on the emissivity model parameters.

8. The apparatus according to claim 7, characterized in that, The wavelength timing separation mechanism includes a filter wheel, a rotation drive unit, and an angle detection unit. The filter wheel is equipped with multiple narrowband filters with different center wavelengths. When the filter wheel rotates, the narrowband filters with different wavelengths enter the optical path sequentially, performing wavelength selective separation of the radiation information at the same spatial location point. The rotation drive unit is used to drive the filter wheel to rotate at a constant angular velocity. The angle detection unit is installed coaxially with the filter wheel to continuously detect the angle of the filter wheel.

9. The apparatus according to claim 7, characterized in that, The single-point radiation detection unit has an adjustable-gain preamplifier circuit, the gain of which is dynamically adjusted according to the radiation intensity to adapt to the intensity of radiation energy at different wavelengths.

Images