Homogeneous epitaxial wafer carrier concentration detection method and system based on raman spectrum

Abstract

This invention provides a method and system for detecting carrier concentration in SiC epitaxial wafers based on Raman spectroscopy, relating to the fields of integrated circuits and precision measurement. The method includes: exciting a pre-calibrated standard sample with pre-calibrated carrier concentration using laser, acquiring Raman spectra, extracting the LO / LOPC mode frequency shift Δω using Lorentz fitting, and obtaining a linear model using the least squares method; determining the carrier concentration of the homogeneous epitaxial wafer under test, placing the wafers in the same temperature-controlled environment, locating the detection point using a confocal system, acquiring the original spectrum, and sequentially performing wavelet transform, adaptive baseline correction, and LO / LOPC peak fitting extraction, then substituting the extraction results into the linear model to obtain the carrier concentration. This invention utilizes the above-mentioned method and system for detecting carrier concentration in homogeneous epitaxial wafers based on Raman spectroscopy to achieve in-situ, non-destructive, and high-precision detection of carrier concentration in SiC epitaxial wafers.

Description

Technical Field

[0001] This invention relates to the fields of integrated circuits and precision measurement, and in particular to a method and system for detecting carrier concentration in homogeneous epitaxial wafers based on Raman spectroscopy. Background Technology

[0002] In the field of 4H-SiC semiconductor manufacturing, carrier concentration detection mainly relies on destructive methods (such as mercury probe CV method which requires electrode deposition) or non-destructive techniques (such as PL spectroscopy which is affected by temperature drift with an error of ±15%, and THz time-domain spectroscopy which requires a vacuum environment). Raman spectroscopy, as a known non-destructive method, can identify the T0 mode (780 cm⁻¹) of SiC. -1 ) and LO model (797cm) -1 However, existing publicly available technologies have not established a quantitative correlation model between the carrier concentration of epitaxial wafers and the signal strength of thin layers, nor have they solved epitaxial wafer-specific problems such as weak signal strength and substrate interference.

[0003] like Figure 1 As shown, current Raman spectroscopy for detecting carrier concentration in 4H-SiC homoepitaxial wafers faces three core limitations: First, existing standards for LOPC coupled-mode detection on GaN substrates rely on complex plasma frequency formulas, which are not adapted to the phonon characteristics of 4H-SiC, such as the LO mode at 797 cm⁻¹. -1 First, the frequency shift response difference leads to insufficient accuracy when directly transferred to SiC epitaxial wafers. Second, due to the thin thickness of the homogeneous epitaxial layer and its identical substrate material, the Raman signal is severely interfered with by substrate scattering. Furthermore, the existing solution CN111880072A is only designed for bulk single crystals, and its optical system lacks sufficient spatial resolution to effectively extract thin-layer characteristic signals. Third, when the carrier concentration > 1×10⁻⁶... 17 cm -3 At that time, the surge in fluorescence background intensity caused the LOPC peak signal-to-noise ratio to plummet to below 2. Simultaneously, the high-temperature environment of the epitaxial process induced frequency shift and thermal drift, approximately 0.02 cm⁻¹. -1 At ℃, current technologies lack a synergistic solution to simultaneously suppress fluorescence interference and temperature drift. The resulting scientific problem is essentially how to establish a quantitative mapping model of LOPC peak-carrier concentration suitable for 4H-SiC epitaxial wafers, and simultaneously solve the physical dilemmas of thin-layer signal attenuation, high-doped fluorescence flooding, and high-temperature frequency shift inaccuracy, so as to achieve in-situ non-destructive high-precision detection.

[0004] While patent CN111880072A achieves Raman spectroscopy detection of SiC carrier concentration, its technical solution has three limitations: First, this solution relies on the TO phonon mode (780 cm⁻¹). -1 The peak shift was modeled, but experiments showed that the sensitivity of the TO mode shift to carrier concentration changes in 4H-SiC epitaxial wafers was only 0.15 cm⁻¹. -1 / decade, when concentration > 1 × 10 17 cm -3 Time-frequency shift Δω < 0.3 cm -1 The inherent resolution of the instrument (±0.5cm) -1 First, noise overwhelms the measurement, causing errors exceeding 15% in highly doped regions. Second, its optical system uses a conventional micro-Raman spectroscopy configuration with a spatial resolution of only 5 μm, failing to penetrate the epitaxial layer and suppress substrate scattering interference; actual measurements show a signal-to-noise ratio as low as 4 dB for a 10 μm thin-layer sample. Third, the lack of an integrated temperature control module means that ambient temperature fluctuations of ±2℃ cause a frequency shift of 0.04 cm. -1 (Equivalent concentration error 12%). Patent JP2020150705A, due to forced coupling of PL spectrum, requires freezing carrier movement in a liquid nitrogen environment at -196℃ to obtain PL characteristic peaks. At the same time, the dual-system optical path calibration deviation causes a positioning error of ±50μm, which significantly reduces the accuracy of local concentration detection in epitaxial wafers. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for detecting carrier concentration in homogeneous epitaxial wafers based on Raman spectroscopy. Addressing the insufficient sensitivity of the TO mode, the LO phonon mode (964.2 cm⁻¹) is employed. -1 As a carrier concentration indicator, its frequency shift sensitivity reaches 0.45 cm⁻¹. -1 / decade, making the highly doped region 1e18cm -3 Δω exceeds 1.35cm -1 It can be accurately captured by standard instruments; to address substrate noise, a confocal pinhole architecture was developed, improving the detection spatial resolution to 1.5μm and suppressing substrate noise to -65dB, ensuring a signal-to-noise ratio >10dB for the 10μm epitaxial layer; to address temperature drift issues, a thermoelectric cooling temperature control stage (±0.1℃) is integrated for real-time compensation of frequency shift. Compared to the PL dependence of JP2020150705A, this solution features a unique wavelet adaptive algorithm that directly strips away the fluorescence background, eliminating the need for liquid nitrogen cooling, improving detection speed, and the single-system optical path design eliminates positioning errors.

[0006] Based on the logarithmic linear relationship between the LO phonon mode frequency shift and carrier concentration in 4H-SiC epitaxial wafers, an online detection system based on 532nm laser confocal Raman spectroscopy was developed. This system utilizes a confocal optical path design to penetrate the epitaxial layer and suppress 60dB of substrate noise. A wavelet transform-adaptive baseline correction algorithm is combined to remove fluorescence background interference under high doping conditions, improving the signal-to-noise ratio of characteristic peaks. Simultaneously, a temperature control module is integrated to compensate for frequency shift thermal drift in real time. The Δω=k·log(N)+b model reduces single-point detection time and minimizes errors. Ultimately, this system achieves in-situ, non-destructive, and high-precision detection of carrier concentration in SiC epitaxial wafers, completely replacing destructive methods.

[0007] To achieve the above objectives, this invention provides a method for detecting carrier concentration in homogeneous epitaxial wafers based on Raman spectroscopy, comprising the following steps: Calibration stage: A standard sample with pre-calibrated carrier concentration is excited by a 532nm laser, Raman spectra are collected, the frequency shift Δω of LO mode / LOPC mode is extracted by Lorentz fitting, and a linear model is obtained by least squares fitting. Detection phase: Determine the carrier concentration of the homogeneous epitaxial wafer to be tested. Place the homogeneous epitaxial wafer to be tested in the same temperature-controlled environment. Locate the detection point through a confocal system. After acquiring the original spectrum, perform wavelet transform, adaptive baseline correction and LO peak / LOPC peak fitting in sequence to extract Δω_LO / Δω_LOPC. Substitute Δω_LO / Δω_LOPC into the linear model to obtain the carrier concentration.

[0008] Preferred detection mechanism for different charge carriers during the detection phase: When carrier concentration > 1×10 17 cm -3 At that time, extract the LOPC mode frequency shift Δω_LOPC; When the carrier concentration is between 5×10 16 ~1×10 17 cm -3 At that time, extract the LO mode frequency shift Δω_LO; When carrier concentration <5×10 16 cm -3 At that time, switch to the 355nm laser-induced LOPC mode and extract Δω_LOPC.

[0009] Preferably, after acquiring the original spectrum, wavelet transform, adaptive baseline correction, and LO peak / LOPC peak fitting are performed sequentially to extract Δω_LO / Δω_LOPC, specifically as follows: First, the original spectrum was decomposed using Daubechies wavelet basis to separate the high-frequency fluorescence background from the low-frequency Raman characteristic signal. Then, the adaptive baseline was fitted using the iterative least squares method to completely remove the fluorescence background caused by high doping. The purified spectrum is in the range of 500-1000 cm⁻¹ -1 Lorentz peak fitting is performed in the interval to accurately extract the LO / LOPC mode frequency shift Δω_LO / Δω_LOPC.

[0010] Preferably, the specific steps in the calibration stage are as follows: A standard sample of 4H-SiC homoepitaxial wafer with pre-calibrated carrier concentration using the Hall effect method was selected. This calibrated standard sample was placed on a temperature-controlled sample stage with an accuracy of ±0.1℃. The surface of the epitaxial layer was focused using a 532nm laser at a constant temperature of 25℃, and samples were collected from 500-1000 cm⁻¹. -1The Raman spectra of the interval were obtained by extracting the frequency shift Δω of the LO mode / LOPC mode through Lorentz fitting, and the coefficients k and b of the linear model Δω=k·log(N)+b were fitted using the least squares method.

[0011] A Raman spectroscopy-based carrier concentration detection system for homogeneous epitaxial wafers has an optical path that sequentially passes through a laser excitation module, a main optical path and co-convergence module, and a spectrophotometer and detector module. The laser excitation module is equipped with a parallel DPSS laser and an ultraviolet laser; the wavelength of the DPSS laser is 532nm and the wavelength of the ultraviolet laser is 355nm. A dichroic mirror DM2 is installed on the main optical path and the co-focusing module. The dichroic mirror DM2 is also connected to a third-order aberration correction objective and a converging lens. The third-order aberration correction objective is also connected to the sample / epitaxial layer under test and the thermoelectric cooling temperature control stage in sequence. The converging lens is also connected to a confocal pinhole. The spatial resolution of the confocal pinhole is ≤2μm, and the accuracy of the thermoelectric cooling temperature control stage is ±0.1℃.

[0012] Preferably, the laser excitation module further includes a first reflector, a second reflector, and a dichroic mirror DM1; wherein, the DPSS laser is connected to the first reflector, the ultraviolet laser is connected to the second reflector, and both the first and second reflectors are connected to the dichroic mirror DM1.

[0013] Preferably, the main optical path and co-focusing module also include a beam expander and a collimating lens; The beam expander is positioned between dichroic mirrors DM2 and DM1. The collimating lens is connected to the confocal pinhole.

[0014] Preferably, the spectral and detection module also includes a long-pass filter, an imaging spectrometer, a high-sensitivity CCD detector, and a data processing system connected in sequence.

[0015] Therefore, this invention employs the aforementioned Raman spectroscopy-based method and system for detecting carrier concentration in homogeneous epitaxial wafers, with the following technical advantages: This application proposes establishing a dual-mode collaborative detection system for epitaxial layers, designing differentiated detection mechanisms for different concentration ranges in 4H-SiC homogeneous epitaxial wafers. A linear model of LO / LOPC peak frequency shift and carrier concentration in 4H-SiC homogeneous epitaxial wafers is established: Δω = k·log(N) + b, with coefficients calibrated using ion implantation of standard samples. This model, based on the strong linear correlation between LO / LOPC frequency shift and log(N) found in the original model, improves computational efficiency, reduces errors, and solves the problem of direct transplantation errors caused by the existing model's incompatibility with SiC phonon characteristics.

[0016] This application proposes a confocal optical path design with a spatial resolution ≤2μm. By using a pinhole architecture, the substrate scattering noise is suppressed to -65dB, thereby improving the signal-to-noise ratio of the 10μm epitaxial layer from <4dB to >10dB. The LO mode signal originates from the epitaxial layer (depth 0-50μm), and the LOPC mode signal originates from the substrate (depth >50μm). Physical separation of the signals is achieved through spatial filtering.

[0017] This application proposes a joint algorithm of Daubechies wavelet transform and iterative least squares baseline correction to directly strip concentrations >1×10⁻⁶. 17 cm -3 This algorithm avoids the liquid nitrogen cooling step required for PL spectroscopy coupling in JP2020150705A by analyzing the high-doped fluorescence immersion mechanism, thus achieving high-doping detection in a single system.

[0018] This application proposes using 355nm ultraviolet laser to excite photogenerated carriers, raising the detection limit of LOPC from the conventional 2×10⁻⁶. 16 cm -3 Expanded to 5×10 15 cm -3 When the energy above the band gap can excite a low-concentration sample to generate sufficient charge carriers, the LOPC peak is significantly enhanced, solving the problem of traditional 532nm lasers in the <5×102 range. 15 cm -3 Industry blind spots where concentration-range failures occur.

[0019] This application proposes a ±0.1℃ thermoelectric cooling temperature control station to suppress frequency shift thermal drift caused by high temperatures (>1600℃) in epitaxial processes in real time. The frequency shift fluctuation is controlled to <0.002cm. -1 (Equivalent concentration error <0.5%). Attached Figure Description

[0020] Figure 1 A schematic diagram summarizing the shortcomings of existing technical solutions and testing methods; Figure 2 This is a flowchart of a method for detecting carrier concentration in homogeneous epitaxial wafers based on Raman spectroscopy. Figure 3 Processing procedures for different carrier concentrations; Figure 4 A complete signal processing flowchart; Figure 5 The standard sample of 4H-SiC homoepitaxial wafer with carrier concentration pre-calibrated by the Hall effect method; Figure 5 (a) is a 10mm*10mm 4H-SiC homoepitaxial wafer; Figure 5 (b) Fabrication of carrier electrodes for Hall effect testing; Figure 6This is a schematic diagram of a carrier concentration detection device for homogeneous epitaxial wafers based on Raman spectroscopy. Figure 7 This is a fitting diagram of the Raman spectrum peaks of a 4H-SiC homoepitaxial wafer; Figure 8 The linear fitting curve of LO / LOPC mode frequency shift versus carrier concentration; Figure 9 The results are for different carrier concentration ranges. Figure 9 (a) is n>1×10 17 cm -3 Detection results within the concentration range; Figure 9 (b) is 5 × 10 16 cm -3 <n<1×10 17 cm -3 Detection results within the concentration range; Figure 9 (c) is n < 5 × 10 16 cm -3 The detection results for the interval; Figure 10 Raman spectra of SiC epitaxial wafers with three different carrier concentrations.

[0021] Figure Labels 1. DPSS laser; 2. Ultraviolet laser; 3. First reflecting mirror; 4. Second reflecting mirror; 5. Dichroic mirror DM1; 6. Beam expander; 7. Dichroic mirror DM2; 8. Third-order aberration corrector objective lens; 9. Sample to be tested; 10. Thermoelectric cooling temperature control stage; 11. Converging lens; 12. Confocal pinhole; 13. Collimating lens; 14. Long-pass filter; 15. Imaging spectrometer; 16. High-sensitivity CCD detector; 17. Data processing system. Detailed Implementation

[0022] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0024] Example 1 like Figure 2 As shown, the method for detecting carrier concentration in homogeneous epitaxial wafers based on Raman spectroscopy includes the following steps: Calibration stage: The carrier concentration (range 1×10⁻⁶) was pre-calibrated using the Hall effect method. 16 ~1×10 19 cm -3Standard samples of 4H-SiC homoepitaxial wafers (thickness 5-50μm) were placed on a temperature-controlled sample stage with an accuracy of ±0.1℃. A 532nm laser (power 5mW, spot diameter ≤2μm) was used to focus on the epitaxial layer surface under a constant temperature of 25℃, collecting samples from 500-1000cm². -1 The Raman spectra of the interval were obtained by extracting the frequency shift Δω of the LO mode / LOPC mode through Lorentz fitting, and the coefficients k and b of the linear model Δω = k·log(N) + b were fitted using the least squares method. Detection Phase: The homogeneous epitaxial wafers to be tested are placed in the same temperature-controlled environment. The detection point is located using a confocal system. After acquiring the original spectrum, wavelet transform (Daubechies wavelet basis separation of fluorescence background), adaptive baseline correction (iterative least squares fitting method), and LO peak / LOPC peak fitting are performed sequentially to extract Δω_LO / Δω_LOPC. Δω_LO / Δω_LOPC is then substituted into a calibrated linear model to output the carrier concentration in real time. For concentrations <5×10⁻⁶, ... 15 cm -3 The sample was switched to a 355nm ultraviolet laser (power density 0.5mW / μm). 2 After stimulating the photogenerated carrier enhancement effect to increase the LOPC signal-to-noise ratio to >8, repeat the above process.

[0025] like Figure 3 As shown, the detection mechanism for different charge carriers during the detection phase: When carrier concentration > 1×10 17 cm -3 At that time, extract the LOPC mode frequency shift Δω_LOPC; When the carrier concentration is between 5×10 16 ~1×10 17 cm -3 At that time, extract the LO mode frequency shift Δω_LO; When carrier concentration <5×10 16 cm -3 At that time, switch to the 355nm laser-induced LOPC mode and extract Δω_LOPC.

[0026] like Figure 4 As shown, after acquiring the original spectrum, wavelet transform, adaptive baseline correction, and LO peak / LOPC peak fitting are performed sequentially, specifically as follows: First, the original spectrum was decomposed using Daubechies wavelet basis to separate the high-frequency fluorescence background from the low-frequency Raman characteristic signal. Then, the adaptive baseline was fitted using the iterative least squares method to completely remove the fluorescence background caused by high doping. The purified spectrum is in the range of 500-1000 cm⁻¹ -1Lorentz peak fitting is performed in the interval to accurately extract the LO / LOPC mode frequency shift Δω; The concentration calculation was performed using a linear model specific to 4H-SiC: Δω = k·log(N) + b, where the coefficients k and b were calibrated using ion implantation standard samples.

[0027] The method for calculating the carrier concentration of a 4H-SiC homoepitaxial wafer is as follows: Correction formula: ; In the formula, The corrected A(LO) peak position value, in centimeters (cm). -1 ); The values ​​represent the A(LO) peak positions of the sample obtained in Raman spectroscopy, expressed in centimeters (cm). -1 ); For the sample obtained in Raman spectroscopy Peak position value, in centimeters (cm) -1 ).

[0028] ; The frequency of charge carriers (plasma) is measured in Hertz (Hz). The wavenumber of the LOPC coupled mode is determined by the peak position (LPP+ or LPP-) of either the high-frequency or low-frequency branch. This value is obtained by fitting the Raman spectrum and is expressed in centimeters (cm). -1 ); The wavenumber of the A1(LO) phonon mode is 964.20 cm⁻¹. -1 ; The wavenumber of the A1(TO) phonon mode is 776.0 cm⁻¹. -1 .

[0029] ; In the formula, n is the carrier concentration, expressed in cubic centimeters (cm³). -3 ); The vacuum permittivity is 8.85 x 10⁻⁶. -12 The unit is farads per meter (F / m); It is the high-frequency dielectric constant, with a value of 5.35; The effective mass of the electron is taken as 0.48m0, where m0 is 9.1 x 10⁻⁶. -31 kg; e is the unit charge, with a value of 1.6 x 10⁻⁶. -19 The unit is coulomb (C).

[0030] like Figure 5As shown, the coefficients k and b of the 4H-SiC homoepitaxial wafer standard sample with carrier concentration pre-calibrated by the Hall effect method are used to fit the linear model Δω=k·log(N)+b.

[0031] like Figure 6 As shown, in the Raman spectroscopy-based homogeneous epitaxial wafer carrier concentration detection system, the optical path sequentially passes through the laser excitation module, the main optical path and co-focusing module, and the spectroscopic and detection module; among which, The laser excitation module is equipped with a parallel DPSS laser 1 and an ultraviolet laser 2; the wavelength of DPSS laser 1 is 532nm and the wavelength of ultraviolet laser 2 is 355nm; the laser excitation module also includes a first reflector 3, a second reflector 4, and a dichroic mirror DM15; wherein, DPSS laser 1 is connected to the first reflector 3, ultraviolet laser 2 is connected to the second reflector 4, and both the first reflector 3 and the second reflector 4 are connected to the dichroic mirror DM15.

[0032] A beam expander 6 is installed on the main optical path and the co-focusing module. A dichroic mirror DM27 is installed behind the beam expander 6. The beam expander 6 is located between the dichroic mirror DM15 and the dichroic mirror DM27.

[0033] The DM27 dichroic mirror is also connected to a third-order aberration correction objective lens 8 and a converging lens 11. The third-order aberration correction objective 8 is also connected in sequence to the sample to be tested 9 / epitaxial layer and the thermoelectric cooling temperature control stage 10. Following the converging lens 11 are the confocal pinhole 12 and the collimating lens 13; the spatial resolution of the confocal pinhole 12 is ≤2μm, and the accuracy of the thermoelectric cooling temperature control stage 10 is ±0.1℃.

[0034] The spectrometer and detector module also includes a long-pass filter 14, an imaging spectrometer 15, a high-sensitivity CCD detector 16, and a data processing system 17, which are connected in sequence.

[0035] like Figure 7 As shown, the complete Raman spectrum was plotted. The test results of different small pieces of the same wafer (S1, S2, S3 represent the numbers of each small piece) were the same, demonstrating the good repeatability of this method.

[0036] like Figure 8 As shown, the linear fitting curve of LO / LOPC mode frequency shift versus carrier concentration is presented, along with the least squares fitting result (Δω = k·log(N) + b), intuitively demonstrating a good linear relationship between Δω and log(N). (Note: The last part is incomplete and likely refers to a label or annotation.) Value, slope intercept The goodness of fit and parameters of the quantification model are crucial. The calibrated linear model is the core formula for subsequent calculation of the concentration of the sample. During the detection stage, only the Δω of the sample needs to be extracted and substituted into the model to quickly calculate the carrier concentration, thus achieving non-destructive and quantitative detection.

[0037] like Figure 9 As shown, this embodiment intuitively demonstrates the mechanism of adaptively selecting the detection mode based on carrier concentration. Figure 9 The paper divides the carrier concentration into three intervals, each corresponding to a different Raman phonon mode extraction strategy, and presents the spectral characteristics and data processing results for each interval. A detailed explanation follows: High concentration range (>1×10) 17 cm -3 At this point, the LOPC mode (phonon-plasma coupling mode) dominates, and its frequency shift is sensitive to concentration changes, while the LO mode is severely suppressed. Therefore, the LOPC mode is extracted for concentration calculation. In the figure, the red fitted curve corresponds to the LOPC mode, the black line is the original spectrum, the gray dashed line is the baseline, and the peak labels show the fitted peak positions.

[0038] Medium concentration range (5×10) 16 ~ 1×10 17 cm -3 The LO mode is dominant and has a high signal-to-noise ratio, while the LOPC mode is weaker. Therefore, the LO mode is extracted. The red fitted curve in the figure corresponds to the LO mode, and other elements are the same as above.

[0039] Low concentration range (<5×10) 16 cm -3 The LOPC mode excited by a conventional 532 nm laser is extremely weak. Therefore, a 355 nm ultraviolet laser was used to enhance the LOPC mode signal by utilizing the photogenerated carrier effect, thus ensuring reliable extraction of the LOPC mode. The figure shows the enhanced spectrum under 355 nm excitation and its fitting results.

[0040] like Figure 10 As shown, three different high concentrations (>1×10⁻⁶) are illustrated. 17 cm -3 Raman spectra of 4H-SiC epitaxial wafers. The black curve represents the original measurement data (including noise and fluorescence background), and the red curve represents the LOPC simulation results extracted after processing using this method. As the carrier concentration increases from 2.3 × 10⁻⁶... 17 Increased to 1.2 × 10 18 cm -3 The peak position of the LOPC model was 970.8 cm. -1 The blue color shifted to 977.2 cm. -1The frequency shift is significant and correlates well with concentration, proving that the LOPC mode is highly sensitive to concentration changes under high doping conditions. At the same time, the red fitting curve is in high agreement with the original data, verifying that the wavelet transform, adaptive baseline correction and Lorentz fitting algorithm proposed in this method can effectively remove background noise and accurately restore the characteristic peak shape. This provides a reliable data foundation for subsequent quantitative calculation of carrier concentration based on the Δω= k·log(N) + b model, and intuitively demonstrates the detection accuracy and discrimination ability of this method in the high concentration range.

[0041] Therefore, the present invention employs the above-mentioned Raman spectroscopy-based method and system for detecting carrier concentration in homogeneous epitaxial wafers, realizing in-situ, non-destructive, and high-precision detection of carrier concentration in SiC epitaxial wafers.

[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for detecting carrier concentration of a homoepitaxial wafer based on Raman spectroscopy, characterized by, Includes the following steps: Calibration stage: A standard sample with pre-calibrated carrier concentration is excited by a 532nm laser, Raman spectra are collected, the frequency shift Δω of LO mode / LOPC mode is extracted by Lorentz fitting, and a linear model is obtained by least squares fitting. Detection phase: Determine the carrier concentration of the homogeneous epitaxial wafer to be tested. Place the homogeneous epitaxial wafer to be tested in the same temperature-controlled environment. Locate the detection point through a confocal system. After acquiring the original spectrum, perform wavelet transform, adaptive baseline correction and LO peak / LOPC peak fitting in sequence to extract Δω_LO / Δω_LOPC. Substitute Δω_LO / Δω_LOPC into the linear model to obtain the carrier concentration.

2. The Raman spectroscopy-based homoepitaxial wafer carrier concentration detection method according to claim 1, characterized by, Detection mechanism for different charge carriers during the detection phase: When the carrier concentration > 1 x 10 17 cm -3 , the LOPC module frequency shift Δω_LOPC is extracted; When the carrier concentration is between 5×10 16 ~1×10 17 cm -3 At that time, extract the LO mode frequency shift Δω_LO; When carrier concentration <5×10 16 cm -3 At that time, switch to the 355nm laser-induced LOPC mode and extract Δω_LOPC.

3. The Raman spectroscopy-based homoepitaxial wafer carrier concentration detection method of claim 1, wherein, After acquiring the raw spectrum, wavelet transform, adaptive baseline correction, and LO / LOPC peak fitting were performed sequentially to extract Δω_LO / Δω_LOPC, specifically as follows: First, the original spectrum was decomposed using Daubechies wavelet basis to separate the high-frequency fluorescence background from the low-frequency Raman characteristic signal. Then, the adaptive baseline was fitted using the iterative least squares method to completely remove the fluorescence background caused by high doping. The purified spectrum is Lorenz peak fitted in the range of 500-1000 cm -1 to accurately extract the LO / LOPC mode shift Δω_LO / Δω_LOPC.

4. The Raman spectroscopy-based homoepitaxial wafer carrier concentration detection method of claim 1, wherein, The specific steps in the calibration phase are as follows: A standard sample of 4H-SiC homoepitaxial wafer with pre-calibrated carrier concentration using the Hall effect method was selected. This calibrated standard sample was placed on a temperature-controlled sample stage with an accuracy of ±0.1℃. The surface of the epitaxial layer was focused using a 532nm laser at a constant temperature of 25℃, and samples were collected from 500-1000 cm⁻¹. -1 The Raman spectra of the interval were obtained by extracting the frequency shift Δω of the LO mode / LOPC mode through Lorentz fitting, and the coefficients k and b of the linear model Δω=k·log(N)+b were fitted using the least squares method.

5. A system for detecting carrier concentration in a homoepitaxial wafer based on Raman spectroscopy, characterized by, For executing the method of claim 1, the optical path sequentially passes through the laser excitation module, the main optical path and co-focusing module, and the beam splitting and detection module; wherein... The laser excitation module is equipped with a parallel DPSS laser and an ultraviolet laser; the wavelength of the DPSS laser is 532nm and the wavelength of the ultraviolet laser is 355nm. A dichroic mirror DM2 is installed on the main optical path and the co-focusing module. The dichroic mirror DM2 is also connected to a third-order aberration correction objective and a converging lens. The third-order aberration correction objective is also connected to the sample / epitaxial layer under test and the thermoelectric cooling temperature control stage in sequence. The converging lens is also connected to a confocal pinhole. The spatial resolution of the confocal pinhole is ≤2μm, and the accuracy of the thermoelectric cooling temperature control stage is ±0.1℃.

6. The carrier concentration detection system for homogeneous epitaxial wafers based on Raman spectroscopy according to claim 5, characterized in that, The laser excitation module also includes a first reflector, a second reflector, and a dichroic mirror DM1; wherein, the DPSS laser is connected to the first reflector, the ultraviolet laser is connected to the second reflector, and both the first and second reflectors are connected to the dichroic mirror DM1.

7. The Raman spectroscopy based homoepitaxial wafer carrier concentration detection system of claim 5, wherein, The main optical path and co-focusing module also include a beam expander and a collimating lens; The beam expander is positioned between dichroic mirrors DM2 and DM1. The collimating lens is connected to the confocal pinhole.

8. The Raman spectroscopy based homoepitaxial wafer carrier concentration detection system of claim 5, wherein, The spectral and detection module also includes a long-pass filter, an imaging spectrometer, a high-sensitivity CCD detector, and a data processing system, which are connected in sequence.

Images