A non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers

By acquiring reflection interferograms and transmission diffraction patterns using dual-wavelength laser beams and fusing the interference phase gradient and diffraction pattern broadening coefficient, the problem of non-destructively obtaining dislocation strain field information of gallium nitride epitaxial wafers is solved, enabling accurate detection of dislocation type and density distribution, and supporting process optimization and quality evaluation.

CN122793757BActive Publication Date: 2026-11-10YU TENG (TONGCHUAN) SEMICONDUCTORTECHNOLOGY CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202611264547.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-08-20
Publication Date
2026-11-10
Estimated Expiration
2046-08-20

AI Technical Summary

Technical Problem

Existing technologies struggle to simultaneously acquire dislocation strain field information on the surface and bulk of gallium nitride epitaxial wafers under non-destructive conditions, and also struggle to accurately distinguish dislocation types and densities at different depths.

Method used

The epitaxial wafer was excited simultaneously using a dual-wavelength laser beam. Reflection interferograms and transmission diffraction patterns were collected. By fusing the interference phase gradient distribution and the diffraction pattern broadening coefficient, spatial distribution data of the strain field was constructed, dislocation types were identified, and the dislocation density distribution of each depth layer was determined.

Benefits of technology

It enables comprehensive and quantitative detection of dislocation defects in gallium nitride epitaxial wafers through nondestructive testing, accurately identifying dislocation types and density distributions, and providing a basis for process optimization and interface quality evaluation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122793757B_ABST
    Figure CN122793757B_ABST
Patent Text Reader

Abstract

This invention discloses a non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers, belonging to the field of semiconductor material testing technology. The method includes: incidenting a dual-wavelength laser beam onto the surface of the gallium nitride epitaxial wafer to be tested, and simultaneously acquiring a reflection interferogram and a transmission diffraction pattern; extracting the interference phase gradient distribution from the reflection interferogram, and extracting the diffraction pattern broadening coefficient from the transmission diffraction pattern; aligning the interference phase gradient distribution and the diffraction pattern broadening coefficient according to spatial coordinates and then fusing them to construct spatial distribution data of the dislocation strain field; reconstructing a spatial map of the dislocation defects based on the spatial distribution data of the dislocation strain field; identifying the dislocation type based on the clustering regions of singularity feature points in the strain field in the spatial map; and determining the dislocation density distribution at each depth layer based on the dislocation type and the attenuation curve of the strain field gradient amplitude in the spatial map.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of semiconductor material testing technology, specifically a non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers. Background Technology

[0002] Dislocation defects in gallium nitride (GaN) epitaxial wafers significantly impact device performance and reliability. Current methods for detecting dislocation defects in GaN epitaxial wafers largely rely on chemical etching followed by microscopic observation, or on transmission electron microscopy (TEM) and cathodoluminescence spectroscopy (CAT). Chemical etching requires destructive treatment of the epitaxial wafer, making non-destructive online detection impossible. Precision instruments like TEM are slow and can only analyze small local areas, failing to provide a comprehensive picture of the overall dislocation distribution. Other methods utilize single-wavelength laser interferometry or diffraction. However, single-wavelength interferometry is easily affected by surface morphology and thickness variations, making it difficult to separate phase changes purely caused by dislocation strain. Single-wavelength diffraction only reflects the average contribution of dislocations to the broadening of diffraction spots, failing to obtain details of the strain field direction and spatial distribution of individual dislocations. Therefore, how to simultaneously acquire dislocation strain field information on the surface and within the epitaxial wafer while maintaining non-destructive testing, and accurately distinguish dislocation types and densities at different depths, has become a pressing problem. Summary of the Invention

[0003] This invention provides a non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers. The purpose is to obtain reflection interference patterns and transmission diffraction patterns simultaneously using dual-wavelength laser beams, and to construct strain field spatial data by fusing the interference phase gradient distribution and diffraction pattern broadening coefficient, thereby identifying dislocation types and determining the dislocation density distribution at each depth layer.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers. This method involves simultaneously exciting the epitaxial wafer with dual-wavelength lasers and acquiring reflection interferograms and transmission diffraction patterns. The interference phase gradient reflecting the in-plane strain distribution is extracted from the reflection interferogram, and the diffraction pattern broadening coefficient reflecting the accumulation of bulk strain is extracted from the transmission diffraction pattern. The two are then aligned in spatial coordinates and fused to construct three-dimensional spatial distribution data of the dislocation strain field, capable of simultaneously characterizing both strain amplitude and direction. Based on this spatial distribution data of the dislocation strain field, a spatial map of dislocation defects is reconstructed, thereby identifying the dislocation type and determining the dislocation density distribution at each depth layer. This enables comprehensive, quantitative, and non-destructive detection of dislocation defects in gallium nitride epitaxial wafers.

[0005] As a technical solution of the present invention, a first-wavelength laser beam illuminates the upper surface of the gallium nitride epitaxial wafer under test at a first incident angle, and a reflection interferogram formed by reflection from the upper surface is acquired. A second-wavelength laser beam illuminates the lower surface of the gallium nitride epitaxial wafer under test at a second incident angle, and diffraction patterns transmitted from the lower surface are acquired. The difference between the first and second wavelengths is preferably maintained between 20 nanometers and 50 nanometers, and the first and second incident angles are respectively set as complementary angles on both sides of the Brewster angle. This configuration of dual-sided, dual-wavelength, and complementary incident angles can effectively suppress the crosstalk between surface scattering and volume scattering, so that the reflection interferogram mainly carries the dislocation strain information of the surface and shallow layer, while the transmission diffraction patterns contain the volume defect modulation information that runs through the entire thickness of the epitaxial wafer. The complementary fusion of the two lays a dual foundation of signal isolation and information completeness for the subsequent construction of high-precision spatial distribution data of dislocation strain field.

[0006] In constructing the spatial distribution data of dislocation strain fields, the interference intensity values ​​of each pixel in the reflection interferogram are unwrapped to obtain the interference phase distribution matrix. Then, the phase change rate is calculated along both the horizontal and vertical directions to obtain the interference phase gradient distribution. The full width at half maximum (FWHM) of each order of diffraction spots in the transmission diffraction pattern is measured, and the difference between each FWHM and the corresponding order of diffraction spots in a defect-free standard epitaxial wafer is determined as the diffraction pattern broadening coefficient. The gradient amplitude at each spatial coordinate point in the interference phase gradient distribution is multiplied by the broadening value at the corresponding spatial coordinate point in the diffraction pattern broadening coefficient to obtain the strain field amplitude at each coordinate point. Simultaneously, the interference gradient direction angle is used as the strain field direction angle, thereby generating the strain field vector at each coordinate point. Finally, the data are arranged according to spatial coordinates to form the spatial distribution data of the dislocation strain field. This fusion method enables the two-dimensional synthesis of amplitude and direction of surface interference information and volume diffraction information in the same coordinate system, truly reflecting the continuous distribution of the dislocation strain field in three-dimensional space.

[0007] In a preferred embodiment of the present invention, when reconstructing the spatial map of dislocation defects, coordinate points whose strain field amplitude exceeds a preset threshold are first extracted from the spatial distribution data of dislocation strain field as candidate defect points; the rate of change of strain field direction angle in the neighborhood of each candidate defect point is calculated, and candidate defect points whose rate of change of direction angle is greater than the angle threshold are identified as dislocation core points; using each dislocation core point as a seed point, region growth is performed along the path of continuous change of strain field direction angle to obtain the spatial contour region of each dislocation defect; the spatial contour regions of each dislocation defect and their corresponding strain field amplitude distributions are superimposed to generate a spatial map. This core point identification and region growth method based on strain field singularity can filter out noise, spurious strain, and weak scattering interference, accurately locate the core position of each dislocation and restore its influence range, and the generated spatial map intuitively shows the spatial morphology, distribution, and interaction relationship of dislocations in the epitaxial wafer.

[0008] More preferably, when determining the dislocation core point, a square neighborhood window of predetermined side length is selected centered on each candidate defect point. The difference in strain field orientation angle between the center point of the window and each peripheral point within the window is calculated. The absolute values ​​of all orientation angle differences are then averaged to obtain the orientation angle change rate of the candidate defect point. The orientation angle change rate is compared with a preset angle threshold. When the orientation angle change rate is greater than the preset angle threshold, the candidate defect point is marked as a dislocation core point. This processing method utilizes the physical characteristic of the rapid change in strain direction in the dislocation core, quantitatively measuring the singularity of the strain field through the local orientation angle change rate. This effectively distinguishes between the true dislocation core and false positive regions caused by growth stress or surface undulations, improving the accuracy of dislocation identification.

[0009] In dislocation type identification, this invention extracts edge contour points where the strain field amplitude exhibits a step abrupt change from the spatial spectrum, and defines the region enclosed by these edge contour points as a clustering region. The cumulative rotation of the strain field direction angle within the clustering region is calculated, and the dislocation spin direction is determined based on the sign of the cumulative rotation. According to the aspect ratio of the clustering region and the dislocation spin direction, the dislocation type is determined to be an edge dislocation, screw dislocation, or mixed dislocation. More specifically, the strain field direction angle values ​​are read point by point along the boundary of the clustering region in a counterclockwise direction, and the differences in strain field direction angles between adjacent boundary points are summed to obtain the cumulative rotation. When the cumulative rotation is positive, the dislocation spin direction is determined to be right-handed; when it is negative, it is determined to be left-handed. Therefore, without relying on diffraction contrast imaging or other electron microscopy techniques, non-destructive determination of dislocation types can be achieved solely through the geometric and topological characteristics of the spatial distribution of the strain field, providing a direct basis for analyzing the correlation between dislocation behavior and growth processes.

[0010] In obtaining the longitudinal distribution of dislocation density, multiple depth layers are divided at equal intervals along the thickness direction of the gallium nitride epitaxial wafer under test. For each depth layer, the strain field gradient amplitude at the corresponding spatial location is extracted, and a decay curve of the strain field gradient amplitude as a function of depth is plotted. Piecewise linear fitting is performed on the decay curve, and the number of dislocation nuclei per unit volume within that depth layer is calculated based on the slope of each segment, thus obtaining the dislocation density distribution. Since the decay law of the dislocation strain field in the propagation direction is directly related to the dislocation type and the crystal structure of the epitaxial layer, the piecewise fitting method of the strain field gradient amplitude decay curve can eliminate the interference of low-frequency fluctuations of the background strain field on dislocation counting, obtaining independent dislocation densities for each depth layer. The resulting longitudinal evolution map of dislocations can clearly reflect the annihilation, multiplication, and reorientation behavior of dislocations during epitaxial growth.

[0011] Based on this, the present invention can also generate a cumulative distribution curve of dislocation density as a function of depth according to the dislocation density distribution of each depth layer; extract the abrupt depth location where the rate of change of dislocation density exceeds the rate of change threshold on the cumulative distribution curve, and mark this abrupt depth location as the interface defect enrichment layer; calculate the defect enrichment factor of the interface defect enrichment layer based on the ratio of the peak dislocation density of the interface defect enrichment layer to the average dislocation density of the adjacent depth layer. The introduction of the defect enrichment factor enables the quantitative characterization of the degree of defect aggregation at the interface between epitaxial layers or between the epitaxial layer and the substrate, providing a sensitive and comparable physical criterion for optimizing the process window of gallium nitride epitaxial wafer fabrication and evaluating interface quality, which helps guide the adjustment of process parameters to reduce interface defect density.

[0012] The technical effects and advantages provided by the present invention in the above technical solution are as follows: By incident a dual-wavelength laser beam onto the surface of a gallium nitride epitaxial wafer under test, and simultaneously acquiring the reflection interferogram and transmission diffraction patterns, with the difference between the two wavelengths and the incident angle set as complementary angles on either side of the Brewster angle, it is possible to simultaneously obtain interference information formed by surface reflection and diffraction information formed by transmission through the epitaxial wafer without damaging the wafer. The reflection interferogram is sensitive to lattice distortion in the near-surface region of the epitaxial wafer, while the transmission diffraction patterns are sensitive to local crystal plane tilting and strain broadening caused by through-dislocations in the epitaxial wafer. The two signals are complementary, allowing for the simultaneous acquisition of surface and bulk strain responses.

[0013] The interference phase gradient distribution and diffraction pattern broadening coefficient are aligned in spatial coordinates and then fused. The strain field amplitude at each coordinate point is obtained through multiplication, and the strain field vector is generated using the interference gradient direction angle as the strain field direction angle. This fusion method allows the strain effect of internal dislocations, which cannot be reflected by a single reflected interference signal, to be corrected and enhanced by the diffraction broadening coefficient. Furthermore, local directional information that cannot be captured by the diffraction pattern is supplemented by the interference phase gradient. The resulting spatial distribution data of the dislocation strain field can simultaneously present the directional characteristics of the dislocation strain field in the plane and its broadening degree along the longitudinal direction.

[0014] After reconstructing the spatial map, clustered regions enclosed by edge contour points exhibiting abrupt changes in strain field amplitude are extracted. The rotational cumulative amount of the strain field direction angle within these regions is calculated, and the dislocation type (edge ​​dislocation, screw dislocation, or mixed dislocation) is determined based on the sign of the rotational cumulative amount and the aspect ratio of the region shape. Depth layers are divided along the thickness direction, and piecewise linear fitting is performed based on the attenuation curve of the strain field gradient amplitude with depth to calculate the number of dislocation nuclei per unit volume in each depth layer. This process directly utilizes the longitudinal strain attenuation difference contained in the dual-wavelength fused data to solve for the dislocation density in different depth layers, obtaining the dislocation density distribution along the thickness direction without the need for layer-by-layer physical stripping of the epitaxial wafer. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a flowchart of a non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers; Figure 2 This is a flowchart of the process for constructing spatial distribution data of dislocation strain fields; Figure 3 This is a flowchart of dislocation type identification and dislocation density distribution determination for gallium nitride epitaxial wafers; Figure 4 These are the distribution curves of strain field amplitude and diffraction pattern broadening coefficient at spatial coordinate points of a gallium nitride epitaxial wafer; Figure 5 These are the distribution curves of different types of dislocation density with depth in gallium nitride epitaxial wafers. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] See Figure 1 This invention provides a non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers, comprising: incidenting a dual-wavelength laser beam onto the surface of the gallium nitride epitaxial wafer to be tested, and simultaneously acquiring a reflection interferogram and a transmission diffraction pattern on the surface of the epitaxial wafer; extracting the interference phase gradient distribution based on the reflection interferogram, and extracting the diffraction pattern broadening coefficient based on the transmission diffraction pattern; aligning the interference phase gradient distribution and the diffraction pattern broadening coefficient according to spatial coordinates and then fusing them to construct spatial distribution data of the dislocation strain field of the gallium nitride epitaxial wafer to be tested; reconstructing a spatial map of dislocation defects in the gallium nitride epitaxial wafer to be tested based on the spatial distribution data of the dislocation strain field; identifying the dislocation type of the gallium nitride epitaxial wafer to be tested based on the clustering region of strain field singularity feature points in the spatial map; and determining the dislocation density distribution of each depth layer in the gallium nitride epitaxial wafer to be tested based on the dislocation type and the attenuation curve of the strain field gradient amplitude in the spatial map.

[0019] Example 1: In practice, the method of incidenting a dual-wavelength laser beam onto the surface of a gallium nitride epitaxial wafer to be tested and simultaneously acquiring the reflection interference pattern and transmission diffraction pattern on the surface of the epitaxial wafer includes configuring a dual-wavelength laser source, setting the incident angle, and simultaneously acquiring the optical signal.

[0020] When configuring a dual-wavelength laser source, a first laser and a second laser are provided. The first laser outputs a first-wavelength laser beam, and the second laser outputs a second-wavelength laser beam. The difference between the first and second wavelengths is controlled between 20 nm and 50 nm. The selection of the wavelength difference is based on the following: when the wavelength difference is less than 20 nm, the difference in penetration depth between the two laser beams in the gallium nitride epitaxial wafer under test is too small to effectively distinguish the strain response caused by surface and subsurface dislocations; when the wavelength difference is greater than 50 nm, the difference in absorption coefficients of the gallium nitride epitaxial wafer under test for the two wavelengths is too large, resulting in a significant difference in the intensity of the transmission diffraction pattern, affecting the accuracy of diffraction pattern broadening coefficient extraction. One selectable wavelength combination is a first wavelength of 405 nm and a second wavelength of 435 nm, with a difference of 30 nm.

[0021] A first-wavelength laser beam, after collimation and beam expansion, illuminates the upper surface of the gallium nitride epitaxial wafer under test at a first incident angle. A second-wavelength laser beam, after collimation and beam expansion, illuminates the lower surface of the gallium nitride epitaxial wafer under test at a second incident angle. The first and second incident angles are set as complementary angles on either side of the Brewster angle, i.e., the sum of the first and second incident angles is 90 degrees. The setting method includes: first obtaining the refractive index of the gallium nitride material at the first and second wavelengths, and then calculating the corresponding Brewster angle. Brewster angle Calculate using the following formula: ; in, The refractive index of gallium nitride at the incident laser wavelength is given by [insert value here]. The refractive index of air is set to 1. Since the first and second wavelengths are close, the difference in refractive index of gallium nitride at these two wavelengths is minimal. Therefore, the average of the two refractive indices can be used to calculate the shared Brewster angle, or the average of the two refractive indices can be calculated separately and used as the reference Brewster angle. Introducing angular offset , The value ranges from 0.5 degrees to 5 degrees. Set the first incident angle. Second angle of incidence Therefore, the first angle of incidence is less than the Brewster angle, and the second angle of incidence is greater than the Brewster angle; the two are complementary angles. Angular offset. The value of is determined by: when When the angle of incidence is less than 0.5 degrees, the first incident angle is too close to the Brewster angle, the p-polarization component in the reflected light from the upper surface weakens sharply, and the contrast of the reflection interferogram decreases; when When the angle is greater than 5 degrees, the second incident angle deviates too far from the Brewster angle, and the diffraction patterns in the transmitted light from the lower surface may be constrained by the total internal reflection condition at the interface, affecting the complete acquisition of the transmitted diffraction patterns. (Selectable...) Degrees, under typical gallium nitride refractive index of 2.4, reference Brewster angle The angle of incidence is approximately 67.4 degrees. degrees, second angle of incidence Spend.

[0022] When a first-wavelength laser beam illuminates the upper surface of the gallium nitride epitaxial wafer under test at a first incident angle, the reflected light from the upper surface carries information about the surface morphology and refractive index changes caused by dislocation strain. This information is collected by the first imaging objective and forms a reflection interferogram on the photosensitive surface of the first image sensor. When a second-wavelength laser beam illuminates the lower surface of the gallium nitride epitaxial wafer under test at a second incident angle, the transmitted light passes through the epitaxial wafer and exits from the lower surface. This transmitted light is collected by the second imaging objective and forms a transmission diffraction pattern on the photosensitive surface of the second image sensor. The exposure of the first and second image sensors is controlled by a synchronous trigger signal, ensuring that the reflection interferogram and the transmission diffraction pattern are acquired within the same time window, with a time synchronization accuracy on the order of microseconds. The synchronous trigger signal is issued by the control unit, simultaneously initiating frame exposure of both image sensors, thereby ensuring that the reflection interferogram and the transmission diffraction pattern correspond to the instantaneous state of the same region of the gallium nitride epitaxial wafer under test.

[0023] Example 2: In practice, the process of extracting the interference phase gradient distribution from the reflection interferogram includes unwrapping the phase of the reflection interferogram and calculating the phase gradient based on the phase distribution matrix.

[0024] When performing phase unwrapping processing on the interference intensity values ​​of each pixel in the reflection interferogram, a phase unwrapping algorithm based on a quality-guided path is employed. The interference intensity value of each pixel is obtained from the reflection interferogram, and the wrapped phase is extracted from the interference intensity value using either a phase-shifting method or a Fourier transform method. The range of the wrapped phase value is... The wrapper contains phase data where the phase value is not zero and the absolute value of the phase difference between adjacent pixels is greater than 1. The pixels are identified as phase transition points. For the identified phase transition points, the pixels are determined by adding or subtracting... The phase is expanded in multiples of each other until the absolute value of the phase difference between adjacent pixels is less than or equal to the value of the phase difference between adjacent pixels. This eliminates phase jumps. After performing the phase unrolling operation on all pixels, an interference phase distribution matrix is ​​obtained, where each element represents the continuous phase value of the corresponding spatial coordinate point.

[0025] After obtaining the interference phase distribution matrix, the phase change rate is calculated along the horizontal and vertical directions of the interference phase distribution matrix. Using any coordinate point in the interference phase distribution matrix... Centered on the coordinate point, take the coordinate point and its horizontally adjacent coordinate points. The phase difference is divided by the spatial distance between the two coordinate points to obtain the phase change rate of the coordinate point along the horizontal direction; the coordinate point is then taken as its perpendicular neighbor. The phase difference is divided by the spatial distance between the two coordinate points to obtain the phase change rate of the coordinate point in the vertical direction. The phase change rates in the horizontal and vertical directions together constitute the interference phase gradient vector of the coordinate point. By traversing all coordinate points in the interference phase distribution matrix and calculating the interference phase gradient vector of each coordinate point, the set of interference phase gradient vectors of all coordinate points is the interference phase gradient distribution.

[0026] The process of extracting the broadening factor of the diffraction pattern based on the transmission diffraction pattern includes measuring the full width at half maximum (FWHM) of each order of diffraction spots in the transmission diffraction pattern and comparing it with the full WHM of the corresponding order of diffraction spots in a defect-free standard epitaxial wafer.

[0027] When measuring the full width at half maximum (FWHM) of diffraction spots of each order in a transmission diffraction pattern, the positions of each order of diffraction spots are identified on the transmission diffraction pattern image, and the intensity distribution curves of each order of diffraction spots in the azimuth direction are extracted. After background subtraction of the intensity distribution curves, the two azimuth positions corresponding to half the peak intensity of the curve are searched, and the width between these two azimuth positions is calculated as the FWHM of the diffraction spot, denoted by . ,in The order numbering of the diffraction spots. Take the integer value. Represents the zero-order diffraction spot. This represents a first-order diffraction spot, and so on.

[0028] The full width at half maximum (FWHM) values ​​of each order of diffraction spots are compared with the corresponding FWHM values ​​of diffraction spots on a defect-free standard epitaxial wafer. The FWHM values ​​of the corresponding orders of diffraction spots on the defect-free standard epitaxial wafer are obtained through pre-calibration and are denoted by the symbol . The calibration method is as follows: A gallium nitride epitaxial wafer verified by high-resolution transmission electron microscopy to be free of dislocation defects is selected as a defect-free standard epitaxial wafer. Standard transmission diffraction patterns are acquired under the same second-wavelength laser beam incident and acquisition conditions as the epitaxial wafer under test. The full width at half maximum (FWHM) of each order of diffraction spots in the standard transmission diffraction pattern is measured to obtain... And store. The difference between the full width at half maximum (FWHM) of each order of diffraction spots and the corresponding order of diffraction spots on a defect-free standard epitaxial wafer is determined as the diffraction pattern broadening coefficient, denoted as . Its defining formula is: ; in, For the first The diffraction pattern broadening factor of a diffraction spot, expressed in angular units or pixel units. The first [item] in the transmission diffraction pattern of the gallium nitride epitaxial wafer to be tested Full width at half maximum (FWHM) of the diffraction spot. For the defect-free standard epitaxial wafer corresponding to the first Full width at half maximum (FWHM) of the diffraction spot. Number the diffraction orders. Calculate the diffraction patterns for different orders separately. The broadening factors of all orders of diffraction patterns together constitute the transmission diffraction pattern broadening factor, which is used to characterize the degree of diffraction pattern broadening caused by dislocation strain.

[0029] Example 3: In specific implementation, please refer to Figure 2 To construct spatial distribution data of dislocation strain field after aligning the interference phase gradient distribution with the diffraction pattern broadening coefficient according to spatial coordinates, it is necessary to first convert the diffraction pattern broadening coefficient from a value indexed by the diffraction order into a distribution map of broadening values ​​corresponding to each spatial coordinate point in the interference phase gradient distribution.

[0030] Diffraction pattern broadening factor extracted from transmission diffraction patterns The overall broadening of diffraction spots at all levels has not yet been correlated with spatial coordinates. The process of establishing this spatial correspondence involves windowing the transmission diffraction pattern image, with the window size matching the size of the epitaxial wafer surface region corresponding to a single pixel in the reflection interferogram. The spatial coordinates of the window center are then used as the reference point. This represents the detection position corresponding to the window. Within each scanning window, the transmission diffraction pattern within that local region is extracted individually, and the full width at half maximum (FWHM) of each order of diffraction spots within that local region is calculated using the same method as for the entire transmission diffraction pattern. This yields the FWHM values ​​for each local order. The full width at half maximum (FWHM) values ​​of local diffraction patterns at various levels are compared with the corresponding FWHM values ​​of diffraction spots on a defect-free standard epitaxial wafer. By subtracting, the local diffraction pattern broadening coefficient is obtained. To obtain a single broadening value corresponding to the same spatial coordinate point as the interference phase gradient distribution, the broadening coefficients of the local diffraction patterns of each order are weighted and summed to obtain the broadening value at the corresponding spatial coordinate point. The weighting coefficients are set according to the rule that lower diffraction orders have a higher weight than higher diffraction orders, and positive and negative orders have the same weight. Alternatively, the average of the broadening coefficients of the first three orders of diffraction patterns can be used as the broadening value at that spatial coordinate point. This generates a diffraction pattern broadening coefficient distribution map that corresponds one-to-one with the spatial coordinates of the reflection interferogram.

[0031] Each spatial coordinate point contained in the interferometric phase gradient distribution is mapped to the broadening value of the same spatial coordinate point in the diffraction pattern broadening coefficient distribution map. During spatial coordinate alignment, image registration methods are used to eliminate potential geometric distortions and offsets between the reflected and transmitted optical paths. Alignment mark patterns are set at the same location on the gallium nitride epitaxial wafer under test. Reflected and transmitted images of the alignment mark patterns are acquired separately, and the coordinate difference of the mark center in the two images is extracted. The affine transformation parameters between the spatial coordinates of the reflected interferogram and the spatial coordinates of the transmitted diffraction pattern are calculated. The affine transformation parameters are used to map the diffraction pattern broadening coefficient distribution map to the spatial coordinate system of the reflected interferogram, ensuring that the same physical location of the epitaxial wafer uses the same coordinate values ​​in both datasets. express.

[0032] After spatial coordinate alignment, a fusion process is performed. This involves merging each spatial coordinate point in the interference phase gradient distribution. The magnitude of the interference gradient at that point corresponds to the same spatial coordinate point in the distribution of the diffraction pattern broadening coefficient. The amplitude values ​​at each point are multiplied, and the product is used as the strain field amplitude at that spatial coordinate point. (Coordinate point in the interferometric phase gradient distribution) The interference gradient vector at point is denoted as Its modulus is denoted as Its direction angle is denoted as Coordinates of the diffraction pattern broadening coefficient distribution. The width value at the location is denoted as Strain field amplitude The formula for calculation is: ; in, Represents spatial coordinates The strain field amplitude at that location, Represents the spatial coordinates of the interferometric phase gradient distribution. The magnitude of the interference gradient vector at that point. Represents the spatial coordinates of the diffraction pattern broadening coefficient distribution. The width value at that location, and These are the x-coordinate index and y-coordinate index of the spatial coordinate point, respectively.

[0033] The interference gradient direction angle at each spatial coordinate point in the interference phase gradient distribution. The strain field direction angle is directly used as the corresponding spatial coordinate point. The strain field amplitude is used as the... The scalar length of the strain field vector is given by the strain field direction angle. Given the pointing angle of the strain field vector, the strain field vectors at each coordinate point are synthesized. The synthesis method is as follows: the horizontal component of the strain field vector... The vertical component of the strain field vector The strain field vectors at all spatial coordinate points constitute a two-dimensional vector field.

[0034] Index the strain field vectors of all spatial coordinate points according to their respective x-coordinates and y-axis index Arranged into a matrix structure with the same number of rows and columns as the original reflection interferogram, each matrix element is a vector data containing amplitude and orientation angle. This matrix structure represents the spatial distribution data of the dislocation strain field. The position of each element in the spatial distribution data of the dislocation strain field corresponds to a detection point on the surface of the gallium nitride epitaxial wafer to be tested, and the element value is the strain field vector information at that detection point.

[0035] See Figure 4 In the figure, the horizontal axis represents the spatial position on the surface of the gallium nitride epitaxial wafer, in micrometers (μm); the vertical axis represents the corresponding numerical amplitude. The figure shows three curves, representing the magnitude of the interference phase gradient. (Blue solid line) Diffraction pattern broadening value (Orange dashed line) and strain field amplitude (Green dotted line) corresponds to the spatial distribution data of dislocation strain field obtained by fusing the interference phase gradient distribution extracted from the reflection interferogram with the spatial distribution of the transmission diffraction pattern broadening coefficient in Example 3.

[0036] The curves exhibit three main peaks, located at approximately 20 μm, 50 μm, and 75 μm. The blue interference phase gradient magnitude curve shows obvious peak-valley fluctuations, with peak values ​​of approximately 1.0, 1.4, and 0.8, reflecting the intensity of phase gradient changes caused by dislocation strain on the gallium nitride epitaxial wafer surface. The orange diffraction pattern broadening curve shows relatively smooth fluctuations, with peak values ​​of approximately 0.85, 1.15, and 0.72, indicating that the broadening degree of transmission diffraction patterns varies with spatial location, reflecting the influence of dislocation strain on diffraction pattern broadening. The green strain field amplitude curve is the product of the blue and orange curves, with a peak value reaching a maximum of approximately 1.6 at 50 μm, much higher than other locations, indicating that the strain field intensity is the greatest at this location, representing the most significant spatial region of dislocation strain; the strain field amplitudes at 20 μm and 75 μm are relatively low, at approximately 0.85 and 0.58, respectively.

[0037] Example 4: In practice, the spatial map of dislocation defects in the gallium nitride epitaxial wafer to be tested is reconstructed based on the spatial distribution data of the dislocation strain field. This includes the process of extracting candidate defect points, determining dislocation core points, growing regions to generate spatial contour regions, and superimposing to generate spatial maps.

[0038] In the spatial distribution data of dislocation strain field, each spatial coordinate point It has a strain field amplitude and a strain field direction angle ,in For row-direction spatial indexing This is a column-direction spatial index. The operation to extract coordinate points whose strain field amplitude exceeds a preset threshold as candidate defect points is as follows: traverse all spatial coordinate points of the dislocation strain field spatial distribution data, and calculate the strain field amplitude of each spatial coordinate point. With preset threshold Compare. Preset threshold. The setting method is as follows: calculate the average value of the strain field amplitude at all spatial coordinate points in the spatial distribution data of dislocation strain field. , set the preset threshold Set as a fixed multiple of the average value, with a value of 2. The rationale for choosing 2 times the average value is that in gallium nitride epitaxial wafers, the strain field amplitude in dislocation-free regions is at a low level and fluctuates around the average value. Strain concentration caused by dislocations will cause the local strain field amplitude to be significantly higher than the background level. Regions reaching more than twice the background average value usually indicate the presence of strain anomalies caused by dislocations. Anything that meets the following criteria... The spatial coordinates of the points are recorded as candidate defect points, forming a set of candidate defect points.

[0039] For each candidate defect point in the candidate defect point set, calculate the rate of change of the strain field direction angle in the neighborhood of that candidate defect point. The calculation method is as follows: using the coordinates of the candidate defect point... A square neighborhood window with a predetermined side length is selected centered on the target area. The predetermined side length of the square neighborhood window, measured in pixels, is set to 5 pixels. The reason for choosing 5 pixels as the side length is as follows: when the window side length is less than 3 pixels, the number of pixels in the neighborhood is too small, resulting in insufficient statistical stability of the calculated orientation angle change rate; when the window side length is greater than 9 pixels, the window may cover the strain field influence areas of multiple dislocation defects, causing the orientation angle change rate to be smoothed and reducing the ability to distinguish dislocation cores; choosing 5 pixels ensures sufficient neighborhood pixel samples while maintaining sensitivity to orientation angle changes. The square neighborhood window contains... There are a total of 121 pixels. After removing the center pixel itself, the number of surrounding pixels is... There are 120 candidate defect points. Within a square neighborhood window, the strain field direction angle difference between the window center point and each peripheral point within the window is calculated. The absolute values ​​of all direction angle differences are then averaged to obtain the candidate defect points. Direction angle change rate The formula for its calculation is: ; in, Representing coordinates The rate of change of the orientation angle of the candidate defect point, in degrees; This represents the total number of points within the square neighborhood window, with a value of 120. and These are the row offset and column offset relative to the center coordinates, respectively, and their values ​​range from [value range missing]. to an integer, and and Not both zero; Indicates the center point The strain field direction angle at the location, in degrees; The coordinates within the square neighborhood window are: The strain field direction angle at the surrounding points, in degrees; This indicates the absolute value operation. The absolute value operation ensures that regardless of whether the strain field orientation angle increases or decreases relative to the center point in the neighborhood, it is treated as an equivalent change in orientation angle, thus capturing the abrupt torsional characteristics of the orientation angle at the dislocation core.

[0040] The calculated rate of change of direction angle With preset angle threshold Compare. Preset angle threshold. The angle is set to 15 degrees. The rationale for setting it to 15 degrees is that the lattice strain direction in dislocation-free regions is basically consistent, and the rate of change of the orientation angle is generally below 5 degrees; near the dislocation core, due to lattice distortion, the strain field direction is drastically deflected, and the rate of change of the orientation angle usually exceeds 15 degrees. When the rate of change of the orientation angle of the candidate defect point... When the degree is greater than 15 degrees, the candidate defect point is marked as a dislocation nucleus point. All marked dislocation nucleus points constitute the dislocation nucleus point set.

[0041] Using each dislocation core point in the set of dislocation core points as a seed point, region growth is performed along a path where the strain field direction angle changes continuously. The starting point of region growth is the dislocation core point, which is added to the current defect region. Starting from the boundary point of the current defect region, the eight neighboring points of the boundary point are checked to determine whether the absolute value of the difference in strain field direction angle between the neighboring point and the boundary point is less than a preset continuous angle tolerance, which is set to 10 degrees. When the absolute value of the difference in strain field direction angle between the neighboring point and the boundary point is less than 10 degrees, it is considered that the neighboring point maintains a continuous change in direction angle with the boundary point, and the neighboring point is included in the current defect region; when the absolute value of the difference is greater than or equal to 10 degrees, the neighboring point is not included. This process continues until all neighboring points of the current defect region boundary no longer meet the condition of continuous change in direction angle, at which point the region growth terminates. All spatial coordinate points included in the region growth process constitute the spatial contour region of the dislocation defect corresponding to the dislocation core point. The preset continuous angle tolerance is set to 10 degrees based on the following considerations: the strain field direction inside the dislocation defect has the characteristic of continuous spatial evolution, and the difference in direction angle between adjacent points exceeding 10 degrees usually means that it has crossed the strain influence range of the dislocation.

[0042] A region growth operation is performed on each dislocation core point in the set of dislocation core points to obtain the spatial contour regions of all dislocation defects. The spatial contour regions of each dislocation defect are then superimposed with the strain field amplitude distribution of the corresponding region in the spatial distribution data of the dislocation strain field. Specifically, the strain field amplitude of each dislocation defect is retained within the spatial contour region, while the strain field amplitude outside the spatial contour region is set to zero. The superimposed spatial contour regions of all dislocation defects and their internal strain field amplitude distributions are then synthesized in the same spatial coordinate system to generate a spatial map of the dislocation defects in the gallium nitride epitaxial wafer under test. The location, boundary shape, and internal strain field amplitude gradient distribution of each dislocation defect are all presented in the spatial map.

[0043] Example 5: In specific implementation, please refer to Figure 3 To identify the dislocation type of the gallium nitride epitaxial wafer under test based on the clustering region of the singularity feature points of the strain field in the spatial spectrum, it is necessary to extract the edge contour points where the strain field amplitude shows a step change, take the area enclosed by the edge contour points as the clustering region, calculate the rotational cumulative amount of the strain field direction angle within the clustering region, determine the dislocation spin direction based on the sign of the rotational cumulative amount, and determine the dislocation type by combining the aspect ratio of the shape of the clustering region.

[0044] In the spatial map, for each reconstructed spatial contour region of a dislocation defect within the spatial map, the abrupt change characteristics of the strain field amplitude are examined point-by-point along the boundary of the spatial contour region. Taking a point on the boundary as the center, the difference in strain field amplitude between the boundary point and the points in the non-defect region immediately outside the boundary point is compared. When the absolute value of this amplitude difference exceeds a preset step threshold, the boundary point is marked as an edge contour point. The preset step threshold is set to three times the standard deviation of the strain field amplitude of all points within the spatial contour region where the boundary point is located. The basis for setting three times the standard deviation is that the strain field amplitude fluctuates inside the dislocation defect, but the strain field amplitude at the boundary usually drops by an order of magnitude from the defective region to the non-defective region. Three times the standard deviation can distinguish between internal fluctuations and boundary abrupt changes. All the marked edge contour points on the boundary of the spatial contour region are connected sequentially according to their spatial position, and the enclosed closed area is determined as the cluster region.

[0045] For each cluster region, calculate the cumulative rotation of the strain field direction angle within the cluster region. Read the strain field direction angle values ​​point by point along the boundary of the cluster region in a counter-clockwise direction. Assume the boundary of the cluster region is formed by... Composed of edge contour points, the first in the counterclockwise direction The strain field direction angle at each edge contour point is denoted as... ,in From 1 to Integers, and supplement the definition To form a closed loop. The strain field direction angle difference between two adjacent edge contour points is obtained by calculation. The original difference, if the original difference is greater than Then subtract the original difference. If the original difference is less than Then add the original difference value This makes the processed difference located at Within the interval. The first... The first to the second The difference in strain field direction angle between each edge contour point after processing is denoted as . Rotational cumulative amount The calculation formula is: ; in, This represents the cumulative rotation of the strain field direction angle in the clustered region, in degrees. This represents the total number of edge contour points on the boundary of the clustered region. This indicates the index number of the edge contour point in the counterclockwise direction. Take 1 to Integers; Indicates from the first The edge contour point to the first The strain field direction angle difference between edge contour points after angle normalization, in degrees.

[0046] Based on rotational cumulative amount The sign of a dislocation determines its spin direction: when When the value is positive, the spin direction of the dislocation corresponding to the cluster region is determined to be right-handed; when... When the value is negative, the spin direction of the dislocation corresponding to the cluster region is determined to be left-handed.

[0047] The aspect ratio of the clustered region is obtained by finding its minimum bounding rectangle. The ratio of the longer side to the shorter side of the minimum bounding rectangle is calculated and used as the aspect ratio of the clustered region. The aspect ratio is combined with the dislocation spin direction to determine the dislocation type: a preset aspect ratio threshold of 3 is set; if the aspect ratio of the clustered region is less than 3 and the cumulative rotation... The absolute value is less than When the dislocation defect corresponding to the clustered region is identified as an edge dislocation, the shape aspect ratio of the clustered region is greater than 3 and the cumulative rotation amount is... The absolute value is greater than When the condition is met, the dislocation defect corresponding to the clustered region is classified as a screw dislocation; dislocation defects corresponding to clustered regions that do not meet the above two conditions are classified as mixed dislocations. The preset aspect ratio threshold of 3 is selected based on the following: the strain field contour lines caused by edge dislocations tend to be isotropic, the clustered region is nearly circular or square, and the aspect ratio is usually below 3; the strain field caused by screw dislocations has the characteristic of spiral distribution along the dislocation line, the clustered region is elongated, and the aspect ratio is usually above 3; absolute value threshold and These correspond to the characteristics of weak rotational effects near edge dislocations and significant rotational effects near screw dislocations, respectively.

[0048] In practice, the dislocation density distribution of each depth layer in the gallium nitride epitaxial wafer under test is determined based on the dislocation type and the attenuation curve of the strain field gradient amplitude in the spatial map. Along the thickness direction of the gallium nitride epitaxial wafer under test, multiple depth layers are divided at equal intervals, with each interval being one-tenth of the total thickness of the epitaxial wafer, resulting in ten depth layers. For each depth layer, the strain field gradient amplitude at the corresponding spatial location is extracted. The extraction method is as follows: all spatial coordinate points of the depth layer are located in the spatial map; the directional derivative of the strain field amplitude along the thickness direction at each spatial coordinate point is calculated as the strain field gradient amplitude at that point; and the arithmetic mean of the strain field gradient amplitudes at all spatial coordinate points within the depth layer is taken as the representative value of the strain field gradient amplitude for that depth layer. The representative values ​​of the strain field gradient amplitudes of the ten depth layers are plotted as curves in order of their center depths, with the horizontal axis representing depth and the vertical axis representing the representative values ​​of the strain field gradient amplitude, resulting in an attenuation curve of the strain field gradient amplitude changing with depth.

[0049] The decay curve is piecewise linearly fitted, with each adjacent depth layer forming a local segment. The fit proceeds from the surface to the bottom layer, and the data points within each segment are fitted using the least squares method to form a straight line segment. Adjacent segments are not required to be continuous at their boundary depths. The number of dislocation nuclei per unit volume within the corresponding depth layer is calculated based on the slope of each segment. A calibration table or curve relating slope values ​​to dislocation densities is pre-established using standard gallium nitride epitaxial wafers with known dislocation densities. The slope values ​​of each segment are then interpolated within this calibration table to obtain the dislocation density value for the corresponding depth layer. The dislocation density values ​​for all depth layers are arranged in depth order, thus yielding the dislocation density distribution of each depth layer in the gallium nitride epitaxial wafer under test.

[0050] In practice, after obtaining the dislocation density distribution at each depth layer, a cumulative distribution curve of dislocation density as a function of depth is generated based on the dislocation density distribution at each depth layer. With depth as the horizontal axis and the cumulative dislocation density from the epitaxial wafer surface to the current depth as the vertical axis, the cumulative dislocation density is the sum of the dislocation densities of all depth layers from the surface to that depth, and the cumulative distribution curve is plotted. On the cumulative distribution curve, the rate of change of dislocation density corresponding to each depth layer is calculated. The rate of change of dislocation density is the dislocation density value of the current depth layer divided by the depth interval between the current depth layer and the previous layer. Each rate of change of dislocation density is compared with a preset rate of change threshold, which is set to three times the average rate of change of dislocation density across all depth layers. The basis for selecting three times the average value is that the dislocation density changes relatively gently along the depth in uniformly distributed regions, with the rate of change close to the average value; in heterogeneous interfaces or defect-rich regions, the dislocation density increases sharply, with the rate of change much greater than the average change level. Depth locations where the rate of change of dislocation density exceeds three times the average value are extracted, and these depth locations are marked as interface defect-rich layers.

[0051] For the marked interface defect enrichment layer, determine the peak dislocation density in the interface defect enrichment layer, and the average dislocation density of the two depth layers that are adjacent to the interface defect enrichment layer above and below and are not marked as interface defect enrichment layers. Calculate the defect enrichment factor of the interface defect enrichment layer. The defect enrichment factor is the ratio obtained by dividing the peak dislocation density by the average dislocation density of the adjacent depth layers.

[0052] See Figure 5 In the figure, the horizontal axis represents the thickness depth of the gallium nitride epitaxial wafer, ranging from 0 to 10 micrometers, and the vertical axis represents the dislocation density, in cm. - ². The figure shows the curves of dislocation density as a function of depth for three types of dislocations: edge dislocation density, screw dislocation density, and mixed dislocation density.

[0053] The edge dislocation density curve shows a monotonically decreasing trend, starting at approximately 1.0 × 10^6 cm. -², gradually decreasing with increasing depth, dropping to approximately 0.05 × 10^6 cm at a depth of 10 micrometers. - ² shows that the edge dislocation density is high on the surface of the epitaxial wafer and decreases significantly along the thickness direction, which is consistent with the characteristics of the dislocation density distribution calculated based on the strain field gradient amplitude decay curve in Example 5.

[0054] The screw dislocation density curve exhibits a distinct peak distribution, with the peak value being approximately 3.4 × 10^6 cm. - ², appearing in the depth range of 3.0 to 3.5 micrometers. The peak position corresponds to the typical characteristics of the interface defect enrichment layer in Example 5, indicating that the screw dislocation density increases concentratedly at this depth, and then the density rapidly decreases to a lower level, reflecting the spatial distribution characteristics of the screw dislocation enrichment layer.

[0055] The mixed dislocation density curve generally shows a trend of first rising and then falling, reaching a peak in the depth range of approximately 4.0 to 5.0 micrometers, with a peak value of approximately 1.8 × 10^6 cm. - ², this region has a higher dislocation density than the surface and deep layers, exhibiting characteristics of a defect-rich area with mixed dislocation types. The broad peak shape of the curve indicates that the mixed dislocations are relatively uniformly distributed and extend over a wide range.

[0056] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers, characterized in that, The method includes: A dual-wavelength laser beam is incident on the surface of the gallium nitride epitaxial wafer to be tested, and the reflection interference pattern and transmission diffraction pattern on the surface of the epitaxial wafer are acquired simultaneously. The interference phase gradient distribution is extracted from the reflection interferogram, and the diffraction pattern broadening factor is extracted from the transmission diffraction pattern. The spatial distribution of the dislocation strain field of the gallium nitride epitaxial wafer to be tested is constructed by merging the interference phase gradient distribution and the diffraction pattern broadening coefficient according to spatial coordinates after alignment with spatial coordinates. This includes: multiplying each spatial coordinate point in the interference phase gradient distribution with the broadening value of the corresponding spatial coordinate point in the diffraction pattern broadening coefficient to obtain the strain field amplitude of each coordinate point; using the interference gradient direction angle of each coordinate point as the strain field direction angle, and combining it with the strain field amplitude to generate the strain field vector of each coordinate point; and arranging the strain field vectors of all coordinate points according to spatial coordinates to form the spatial distribution of the dislocation strain field. Based on the spatial distribution data of the dislocation strain field, the spatial map of the dislocation defects in the gallium nitride epitaxial wafer to be detected is reconstructed. Based on the clustering regions of strain field singularity feature points in the spatial spectrum, the dislocation type of the gallium nitride epitaxial wafer to be tested is identified; Based on the dislocation type and the attenuation curve of the strain field gradient amplitude in the spatial spectrum, the dislocation density distribution of each depth layer in the gallium nitride epitaxial wafer to be tested is determined.

2. The non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers according to claim 1, characterized in that, The process of incidenting a dual-wavelength laser beam onto the surface of the gallium nitride epitaxial wafer to be tested, and simultaneously acquiring the reflection interference pattern and transmission diffraction pattern of the epitaxial wafer surface, includes: A first wavelength laser beam is used to illuminate the upper surface of the gallium nitride epitaxial wafer to be tested at a first incident angle, and a reflection interference pattern formed by the reflection from the upper surface is collected. A second wavelength laser beam is irradiated at a second incident angle onto the lower surface of the gallium nitride epitaxial wafer to be tested, and diffraction patterns transmitted through the lower surface are collected. The difference between the first wavelength and the second wavelength is between 20 nanometers and 50 nanometers, and the first incident angle and the second incident angle are respectively set as complementary angles on both sides of Brewster's angle.

3. The non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers according to claim 1, characterized in that, The interference phase gradient distribution is extracted based on the reflection interferogram, and the diffraction pattern broadening coefficient is extracted based on the transmission diffraction pattern, including: The interference intensity values ​​of each pixel in the reflected interferogram are subjected to phase unwrapping processing to obtain the interference phase distribution matrix; The phase change rate is calculated along the horizontal and vertical directions of the interference phase distribution matrix to obtain the interference phase gradient distribution. The full width at half maximum (FWHM) of each order of diffraction spots in the transmission diffraction pattern is measured, and the difference between the FWHM of each order of diffraction spots and the corresponding order of diffraction spots in a defect-free standard epitaxial wafer is determined as the diffraction pattern broadening factor.

4. The non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers according to claim 1, characterized in that, Based on the spatial distribution data of the dislocation strain field, the spatial map of the dislocation defects in the gallium nitride epitaxial wafer to be tested is reconstructed, including: Coordinate points whose strain field amplitude exceeds a preset threshold are extracted from the spatial distribution data of the dislocation strain field as candidate defect points; Calculate the strain field direction angle change rate in the neighborhood of each candidate defect point, and determine the candidate defect point whose direction angle change rate is greater than the angle threshold as the dislocation core point; Using each fault core point as a seed point, a region is grown along a path where the strain field direction angle changes continuously to obtain the spatial contour region of each fault defect. The spatial map is generated by superimposing the spatial contour regions of each fault defect and their corresponding strain field amplitude distributions.

5. The non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers according to claim 4, characterized in that, The calculation of the strain field direction angle change rate in the neighborhood of each candidate defect point, and the determination of candidate defect points with the direction angle change rate greater than an angle threshold as dislocation nuclei, includes: A square neighborhood window with a predetermined side length is selected centered on each candidate defect point; Within the square neighborhood window, the strain field direction angle difference between the center point of the window and each peripheral point within the window is calculated, and the absolute value of all direction angle differences is taken and averaged to obtain the direction angle change rate of the candidate defect point. The rate of change of the orientation angle is compared with a preset angle threshold. When the rate of change of the orientation angle is greater than the preset angle threshold, the candidate defect point is marked as the dislocation core point.

6. The non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers according to claim 1, characterized in that, The step of identifying the dislocation type of the gallium nitride epitaxial wafer to be tested based on the clustering region of strain field singularity feature points in the spatial spectrum includes: Extract edge contour points where the strain field amplitude exhibits a step change in the spatial map, and use the area enclosed by the edge contour points as the clustering region. Calculate the cumulative rotation of the strain field direction angle within the aggregation region, and determine the dislocation spin direction based on the sign of the cumulative rotation. Based on the aspect ratio of the aggregated region and the dislocation spin direction, the dislocation type is determined to be an edge dislocation, a screw dislocation, or a mixed dislocation.

7. The non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers according to claim 6, characterized in that, Calculating the cumulative rotation of the strain field direction angle within the accumulation region, and determining the dislocation spin direction based on the sign of the cumulative rotation, includes: The strain field direction angle values ​​were read point by point along the boundary of the aggregation region in a counterclockwise direction; The cumulative rotation amount is obtained by summing the strain field direction angle differences between adjacent boundary points. When the cumulative rotation amount is positive, the dislocation rotation direction is determined to be right-handed; when the cumulative rotation amount is negative, the dislocation rotation direction is determined to be left-handed.

8. The non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers according to claim 1, characterized in that, Based on the dislocation type and the attenuation curve of the strain field gradient amplitude in the spatial spectrum, the dislocation density distribution of each depth layer in the gallium nitride epitaxial wafer to be tested is determined, including: Along the thickness direction of the gallium nitride epitaxial wafer to be tested, multiple depth layers are divided at equal intervals; For each depth layer, the strain field gradient magnitude at the corresponding spatial location of that depth layer is extracted, and the attenuation curve of the strain field gradient magnitude as a function of depth is plotted. The attenuation curve is piecewise linearly fitted, and the number of dislocation core points per unit volume within the layer at that depth is calculated based on the slope value of each segment to obtain the dislocation density distribution.

9. The non-destructive testing method for dislocation defects in gallium nitride epitaxial wafers according to claim 1, characterized in that, After determining the dislocation density distribution of each depth layer in the gallium nitride epitaxial wafer to be tested based on the dislocation type and the attenuation curve of the strain field gradient amplitude in the spatial spectrum, the process further includes: The cumulative distribution curve of dislocation density as a function of depth is generated based on the dislocation density distribution of each depth layer; Extract the locations of abrupt changes where the rate of change of dislocation density exceeds a threshold from the cumulative distribution curve, and mark these locations as interface defect enrichment layers. The defect enrichment factor of the interface defect enrichment layer is calculated based on the ratio of the peak dislocation density of the interface defect enrichment layer to the average dislocation density of the adjacent depth layer.

Citation Information

Patent Citations

  • Gallium nitride device and gallium nitride packaging structure

    CN118866823A

  • Method and system for optimizing light spot uniformity of high-precision VCSEL (Vertical Cavity Surface Emitting Laser) device

    CN120951680A