TWI759237B - Ingot evaluation method - Google Patents

Abstract

An ingot evaluation method is provided. A wafer image corresponding to each of the plurality of wafers cut by multiple ingots is divided into a plurality of regions. A plurality of statistical number of multiple defect types included in each region are calculated. A plurality of statistical parameters are obtained based on the plurality of statistical number of the defect types included in in each region. A regression analysis is performed using the statistical parameters and a bow value corresponding to each wafer to obtain multiple regression coefficients. A curvature prediction equation set is established based on the regression coefficients.

Description

Ingot Evaluation Method

The present invention relates to an evaluation and inspection method, and in particular, to an ingot evaluation method.

The semiconductor manufacturing industry has strict requirements on product quality, and the quality of incoming ingots has a great impact on the quality of wafers obtained after processing. Therefore, how to filter out the ingots with poor quality before processing the ingots will be one of the subjects in this field.

The invention provides a method for evaluating a crystal ingot, which can predict the quality of the crystal ingot from the evaluation piece of the crystal ingot.

The ingot evaluation method of the present invention includes: dividing respective wafer images corresponding to a plurality of wafers cut from a plurality of ingots into a plurality of regions; calculating the statistical quantity of a plurality of defect types included in each region; obtain a plurality of statistical parameters; perform regression analysis by using the statistical parameters corresponding to each wafer and the warp value to obtain a plurality of regression coefficients; and establish warpage based on the regression coefficients Degree prediction equation system.

In an embodiment of the present invention, based on the statistical number of defect types included in the area, the step of obtaining statistical parameters includes: for any defect type, performing the following steps. calculating the statistical quantity corresponding to the defect type in each area; calculating the standard deviation of the statistical quantity corresponding to the defect type in the area as one of the statistical parameters; and sorting the areas corresponding to the defect type Statistical quantity, and take out the A statistical quantity with the highest numerical value in a manner from large value to small value to calculate the average value as one of the statistical parameters, where A is a positive integer.

In an embodiment of the present invention, the defect types include threading edge dislocation (TED), threading screw dislocation (TSD) and basal plane dislocation (BPD). . The regression analysis is performed using the statistical parameters and the warp values corresponding to each wafer, and the step of obtaining the regression coefficient includes: inputting the statistical parameters and warp values corresponding to each wafer into the following formula to perform the regression analysis. Bow(k)=P1×Std(TED)+P2×Avg(TED); Bow(k)=P3×Std(TSD)+P4×Avg(TSD); Bow(k)=P5×Std(BPD)+P6×Avg(BPD). Among them, Bow(k) is the bending value of the kth wafer, P1-P6 are regression coefficients, and Std(TED), Std(TSD), and Std(BPD) are the regions corresponding to the through edge position and through The standard deviation of the statistical quantity of the three defect types of screw dislocation and basal plane dislocation, Avg(TED), Avg(TSD), Avg(BPD) are corresponding to threading edge dislocation, threading thread dislocation and basal plane position, respectively The average of the highest A statistics for the three defect types.

In an embodiment of the present invention, based on the statistical number of defect types included in the area, the step of obtaining statistical parameters includes: for any two defect types, performing the following steps. calculating the total statistical quantity of the two types of defects in each area; calculating the standard deviation of the total statistical quantity corresponding to the two types of defects in the area as one of the statistical parameters; and sorting the corresponding areas in the area From the total statistics of any two defect types, the highest A statistics are taken out to calculate the average value as one of the statistical parameters.

In an embodiment of the present invention, the defect types include threading edge dislocations, threading screw dislocations, and basal plane dislocations. Regression analysis is performed using statistical parameters and bending values corresponding to the wafer to obtain regression analysis. The step of coefficient includes: inputting the statistical parameters and bending values corresponding to each wafer into the following formula to perform regression analysis. Bow(k)=P7×Std(TED&TSD)+P8×Avg(TED&TSD);Bow(k)=P9×Std(TSD&BPD)+P10×Avg(TSD&BPD); Bow(k)=P11×Std(TED&BPD)+P12×Avg(TED&BPD). Among them, Bow(k) is the bending value of the k-th wafer, and P7-P12 are regression coefficients. Std(TED&TSD) is the standard deviation of the statistical quantity corresponding to the sum of the two defect types of threading edge dislocations and threading screw dislocations in the region, and Avg(TED&TSD) is corresponding to the two types of threading edge dislocations and threading screw dislocations. The average of the highest A statistics among the total statistics of the defect types. Std(TSD&BPD) is the standard deviation of the statistical number corresponding to the sum of the two defect types of threading dislocation and basal plane dislocation in the region, Avg(TSD&BPD) is the defect corresponding to threading dislocation and basal plane dislocation The average of the highest A statistics in the category total statistics. Std(TED&BPD) is the standard deviation of the statistical number corresponding to the sum of the two defect types of threading edge dislocations and basal plane dislocations in the region, Avg(TED&BPD) is corresponding to the two types of threading edge dislocations and basal plane dislocations The average of the highest A statistics among the total statistics of the defect types.

In an embodiment of the present invention, based on the statistical quantity of defect types included in the area, the step of obtaining statistical parameters includes: calculating the total statistical quantity of the defect types in each area; calculating that the area corresponds to the defect The standard deviation of the total statistical quantities of the types is used as one of the statistical parameters; and the statistical quantities corresponding to the total defect types in the area are sorted, and the highest A statistical quantities are taken out to calculate the average value as one of the statistical parameters, wherein A is a positive integer.

In an embodiment of the present invention, the regression analysis is performed using the statistical parameters and the warp values corresponding to each wafer, and the step of obtaining the regression coefficients includes: inputting the statistical parameters and warp values corresponding to each wafer into the following formula to execute regression analysis. Bow(k)=P13×Std(TED&TSD&BPD)+P14×Avg(TED&TSD&BPD). Among them, Bow(k) is the bending value of the k-th wafer, and P13-P14 are regression coefficients. Std(TED&TSD&BPD) is the standard deviation of the statistics corresponding to the total of three defect types of threading edge dislocations, threading screw dislocations and basal plane dislocations in the region, Avg(TED&TSD&BPD) is corresponding to threading edge dislocations, threading The average value of the highest A statistics among the total statistics of the three defect types of screw dislocations and basal plane dislocations.

In an embodiment of the present invention, after establishing the tortuosity prediction equation set based on the regression coefficients, the method further includes: using the tortuosity prediction equation set to calculate the wafer to be tested processed from the ingot to be tested If the bending value after processing is within a specification range, it is judged that the quality of the ingot to be tested is good; and if the bending value after processing is not within the specification range , it is judged that the quality of the ingot to be tested is not good.

In an embodiment of the present invention, after the step of dividing the respective wafer images of the wafers cut from the ingot into the regions, the step further includes: discarding regions located at four corners of the wafer images, Instead, the statistical number of defect types included in the remaining area is calculated.

Based on the above, a set of curvature prediction equations is established by using the processed wafers of the known ingots, thereby predicting the quality of the ingot to be tested through the set of curvature prediction equations, and then filtering out the poor processing geometry quality. The ingot can greatly improve the overall processing quality and reduce the production cost.

FIG. 1 is a block diagram of an analysis system according to an embodiment of the present invention. Referring to FIG. 1 , the analysis system includes a measuring instrument 110 and an analysis device 120 . Data transmission between the measuring instrument 110 and the analysis device 120 can be performed, for example, through wired or wireless communication.

The measuring instrument 110 is, for example, an automated optical inspection (Automated Optical Inspection, AOI for short) instrument. AOI instrument is a high-speed and high-precision optical image inspection system, including measurement lens technology, optical lighting technology, positioning measurement technology, electronic circuit testing technology, image processing technology and automation technology application, etc. It uses machine vision as the detection standard technology. The measuring instrument 110 uses an optical instrument to obtain the surface state of the finished product, and then uses computer image processing technology to detect defects such as foreign objects or abnormal patterns.

The analysis device 120 is an electronic device with computing function, which can be implemented by a personal computer, a notebook computer, a tablet computer, a smart phone, etc. or any device with computing function, and the present invention is not limited thereto. The analysis device 120 receives the measurement data (ie, the coordinate position and the type of the defect) of a plurality of known wafers from the measurement instrument 110, thereby performing training to obtain a prediction model (a set of curvature prediction equations), For the subsequent use of the measurement data of the wafer under test to obtain the quality of the wafer under test after being processed into a wafer.

2 is a schematic diagram of an ingot according to an embodiment of the present invention. Referring to FIG. 2, after the ingots 20-1 to 20-N are processed, a wafer is taken from each ingot to obtain a plurality of wafers 21-1 to 21-N. For cutting, grinding and polishing, the present invention is not limited thereto. The measuring instrument 110 is used to perform optical inspection one by one to detect whether the plurality of coordinate positions included in each of the wafers 21 - 1 to 21 -N are defective, and record the coordinate positions with defects and the types of defects. The defect types include threading edge dislocation (TED), threading screw dislocation (TSD) and basal plane dislocation (BPD), among which multiple wafers 21-1 ~21-N can be wafers obtained after processing at any position on the ingot. In some preferred embodiments, the plurality of wafers 21-1~21-N are the first wafers close to the ingots 20-1~20-N. The wafer obtained after processing at the positions of one end E1 and the second end E2 (both ends of the head and tail) can be processed at positions outside the regions of the head and tail ends of the ingots 20-1 to 20-N in other embodiments. The obtained wafer is not limited in the present invention.

FIG. 3 is a flowchart of an ingot evaluation method according to an embodiment of the present invention. Referring to FIG. 3 , in step S305 , the respective wafer images corresponding to the plurality of wafers 21 - 1 to 21 -N cut by the plurality of ingots 20 - 1 to 20 -N are divided into a plurality of regions. FIG. 4 is a schematic diagram of a wafer image according to an embodiment of the present invention. Referring to FIG. 4 , the wafer image 400 is divided into a plurality of regions C1 - C36 . The number of cuttings here is only one example. In other embodiments, the wafer image 400 can be divided into any number of regions according to requirements, and the present invention is not limited thereto. Each of the wafers 21 - 1 to 21 -N has a corresponding wafer image similar to the wafer image 400 .

Next, in step S310, the statistical number of a plurality of defect types included in each area is calculated. For example, the wafer 21-1 will be described, and the wafer image 400 will be described as the wafer image of the wafer 21-1. The analysis device 120 calculates the TED values included in the regions C1 to C36 of the wafer image 400 corresponding to the wafer 21-1 according to the measurement data corresponding to the wafer 21-1 (ie, the coordinate position and the type of the defect) of the wafer 21-1. Statistics, TSD statistics, and BPD statistics. The same applies to the other wafers 21-2 to 21-N.

In addition, since the proportion of the circle W in the four corners of the wafer image 400 is relatively low, in the step of counting defect types, the regions C33 to C36 located at the four corners of the wafer image 400 can be discarded, and the calculation The statistical number of defect types included in the remaining areas C1 to C32. For example, Table 1 describes the statistical numbers of various defect types in the regions C1 to C32 of the N wafers 21 - 2 to 21 -N.

Table 1 wafer area Defect type total number 21-1 C1 TED C _1-1-TED TSD C _1-1-TSD BPD C _1-1-BPD … … … C32 TED C _1-32-TED TSD C _1-32-TSD BPD C _1-32-BPD … … … … 21-N C1 TED _CN-1-TED TSD _CN-1-TSD BPD _CN-1-BPD … … … C32 TED _CN-32-TED TSD _CN-32-TSD BPD _CN-32-BPD

Next, in step S315, a plurality of statistical parameters are obtained based on the statistical number of defect types included in the area. Here, the calculated standard deviation and the mean value can be used as statistical parameters.

For example, for each defect type, the standard deviation and average value of the statistical quantities of a plurality of regions divided on the same wafer are calculated. If there are 3 types of defects, 6 statistical parameters can be obtained. For the wafer 21-1 in Table 1, the standard deviation Std (TED (TED) of the TED corresponding to the wafer 21-1 is calculated based on the statistical quantities C _1-1-TED to C _1-32-TED of the regions C1 to C32 . ). The standard deviation Std(TSD) of the TSD corresponding to the wafer 21-1 is calculated by using the statistical quantities _C1-1-TSD to _C1-32-TSD of the regions C1-C32. The standard deviation Std(BPD) of the BPD corresponding to the wafer 21-1 is calculated by using the statistical quantities _C1-1-BPD to _C1-32-BPD of the regions C1 to C32. And, sort C _1-1-TED to C _1-32-TED to take out the highest A (for example, 5), and calculate the average value Avg(TED). Sort C _1-1-TSD to C _1-32-TSD to take the top 5 and calculate the average Avg(TSD). Sort C _1-1-BPD to C _1-32-BPD to take the top 5, and calculate the average Avg(BPD).

After that, in step S320 , regression is performed using the statistical parameters corresponding to each wafer (21-1 to 21-N) and the bending (BOW) value of each wafer (21-1 to 21-N) after the processing process. analysis to obtain multiple regression coefficients. For example, regression analysis is performed by inputting statistical parameters and bending values into the following equations (1) to (3), and a plurality of regression coefficients P1 to P6 are obtained. Among them, Bow(k) is the bending value of the kth wafer after the processing process. (1) Bow(k)=P1×Std(TED)+P2×Avg(TED); (2) Bow(k)=P3×Std(TSD)+P4×Avg(TSD); (3) Bow(k)=P5×Std(BPD)+P6×Avg(BPD).

Finally, in step S325, a set of tortuosity prediction equations is established based on the regression coefficients. Based on the obtained regression coefficients P1 to P6, the following set of tortuosity prediction equations (A) are established: Bow1=P1×T _Std(TED) +P2×T _Avg(TED) ; Bow2=P3×T _Std(TSD) + P4×T _Avg(TSD) ; Bow3=P5×T _Std(BPD) +P6×T _Avg(BPD) ; Pre_Bow=(Bow1+Bow2+Bow3)/3.

In the prediction equation group (A), the standard deviation T _Std(TED) corresponding to the TED in the wafer to be tested and the average value T _Avg(TED) are used to obtain the curvature Bow1, and the standard deviation T corresponding to the TSD in the wafer to be tested is used. _Std(TSD) and the average T _Avg(TSD) to obtain the bow2, and use the standard deviation T _Std(BPD) corresponding to the BPD in the wafer to be tested and the average T _Avg(BPD) to obtain the bow3. After that, the bending degrees Bow1 , Bow2 , and Bow3 are averaged as the post-processing bending value Pre_Bow.

Then, it is judged whether the post-processing bending value Pre_Bow is within the specification range. If the bending value Pre_Bow after processing is within the specification range, it is determined that the quality of the ingot to be tested corresponding to the wafer to be tested is good. If the bending value Pre_Bow after processing is not within the specification range, it is determined that the quality of the ingot to be tested corresponding to the wafer to be tested is not good. The wafer to be tested obtained through the processing process.

In addition, the standard deviation and the average value are calculated for each defect type in addition to the above. The total statistical quantity can also be calculated for every two defect types, every three defect types, etc., thereby obtaining the corresponding standard deviation and average value.

For example, for any two defect types, the following steps are performed. calculating the total statistical quantity of the two types of defects in each area; calculating the standard deviation of the total statistical quantity corresponding to the two types of defects in the area as one of the statistical parameters; and sorting the corresponding areas in the area From the total statistics of any two defect types, the highest A statistics are taken out to calculate the average value as one of the statistical parameters.

Table 2 describes the statistical numbers of any two types of defects in the regions C1 to C32 of the N wafers 21-2 to 21-N.

Table 2 wafer area Defect type total number 21-1 C1 TED+TSD C _1-1-TED&TSD TSD+BPD C _1-1-TSD&BPD TED+BPD C _1-1-TED&BPD … … … C32 TED+TSD C _1-32-TED&TSD TSD+BPD C _1-32-TSD&BPD TED+BPD C _1-32-TED&BPD … … … … 21-N C1 TED+TSD _CN-1-TED&TSD TSD+BPD _CN-1-TSD&BPD TED+BPD _CN-1-TED&BPD … … … C32 TED+TSD _{CN-32-TED&TSD} TSD+BPD _CN-32-TSD TED+BPD _{CN-32-TED&BPD}

For the _{wafer 21-1} in Table 2, the standard deviation Std ( TED&TSD). And, sort C _1-1-TED&TSD to C _1-32-TED&TSD to take out the highest A (for example, 5), and calculate the average Avg(TED&TSD). By analogy, the standard deviation Std(TSD&BPD) and the mean value Avg(TSD&BPD) corresponding to TSD and BPD, and the standard deviation Std(TED&BPD) and mean value Avg(TED&BPD) corresponding to TED and BPD are obtained. The above-mentioned statistical parameters and bending values are input into the following equations (4) to (6), and regression analysis is performed to obtain a plurality of regression coefficients P7 to P12. Among them, Bow(k) is the bow value of the kth wafer. (4) Bow(k)=P7×Std(TED&TSD)+P8×Avg(TED&TSD); (5) Bow(k)=P9×Std(TSD&BPD)+P10×Avg(TSD&BPD); (6) Bow(k )=P11×Std(TED&BPD)+P12×Avg(TED&BPD).

Table 3 describes the statistics of the three types of defects in the regions C1 to C32 of the N wafers 21 - 2 to 21 -N.

table 3 wafer area Defect type total number 21-1 C1 TED+TSD+BPD C _{1-1-TED&TSD&BPD} … … … C32 TED+TSD+BPD C _{1-32-TED&TSD&BPD} … … … … 21-N C1 TED+TSD+BPD _{CN-1-TED&TSD&BPD} … … … C32 TED+TSD+BPD _{CN-32-TED&TSD&BPD}

For the wafer 21-1 in Table 3, the total of TED, TSD and BPD corresponding to the wafer 21-1 is calculated based on the statistical quantities _{C1-1-TED&TSD&BPD} to _{C1-32-TED&TSD&BPD} of the regions C1-C32. Standard Deviation Std (TED&TSD&BPD). And, sort C _{1-1-TED&TSD&BPD} to C _{1-32-TED&TSD&BPD} in a manner from the largest value to the smallest value to take out the highest A (for example, the top 5 highest in value), and calculate the average value Avg(TED&TSD&BPD). The above-mentioned statistical parameters and bending values were input into the following equation (7) to perform regression analysis, and regression coefficients P13 to P14 were obtained. Among them, Bow(k) is the bow value of the kth wafer. (7) Bow(k)=P13×Std(TED&TSD&BPD)+P14×Avg(TED&TSD&BPD).

As far as the defect types include TED, TSD and BPD, 14 statistical parameters can be used, including: the standard deviation and average value of TED, TSD and BPD, a total of 6 statistical parameters, and the standard deviation and average value of any two defect types. There are 6 statistical parameters for the average value and 2 statistical parameters for the standard deviation and the average value of the three defect types.

Table 4 shows the 14 statistical parameters corresponding to each wafer. A plurality of regression coefficients P1 to P14 are obtained by performing regression analysis through the above equations (1) to (7).

Table 4 Statistical parameters Wafer 21-1 Wafer 21-2 Wafer 21-3 … Std (TED) 666.812567 845.660688 1555.3768 … Std(TSD) 356.080047 82.8070045 301.826109 … Std(BPD) 960.610223 418.983293 1136.59931 … Avg(TED) 1413.93069 1738.43033 2795.59296 … Avg(TSD) 623.201412 157.225952 530.207507 … Avg(BPD) 1590.54079 843.729815 2150.92073 … Std(TED&TSD) 712.081456 846.248781 1567.21313 … Std(TSD&BPD) 1008.85827 874.634781 1587.50496 … Std(TED&BPD) 919.302453 418.580936 1131.86793 … Avg(TED&TSD) 1535.54079 1743.3072 2830.61124 … Avg(TSD&BPD) 1965.97558 1849.94595 3139.78662 … Avg(TED&BPD) 1608.47132 853.920371 2168.88912 … Std(TED&TSD&BPD) 985.160901 875.104565 1596.44449 … Avg(TED&TSD&BPD) 1995.02882 1854.52959 3163.80467 …

Based on the obtained regression coefficients P1 to P14, the following set of tortuosity prediction equations (B) are established: Bow1=P1×T _Std(TED) +P2×T _Avg(TED) ; Bow2=P3×T _Std(TSD) + P4×T _Avg(TSD) ; Bow3=P5×T _Std(BPD) +P6×T _Avg(BPD) ; Bow4=P7×T _Std(TED&TSD) +P8×T _Avg(TED&TSD) ; Bow5=P9×T _{Std (TSD&BPD)} +P10×T _Avg(TSD&BPD) ; Bow6=P11×T _Std(TED&BPD) +P12×T _Avg(TED&BPD) ; Bow7=P13×T _{Std(TED&TSD&BPD)} +P14×T _{Avg(TED&TSD&BPD)} ; Pre_Bow= (Bow1+Bow2+Bow3+Bow4+Bow5+Bow6+Bow7)/7.

In the prediction equation group (B), the standard deviation T _Std(TED) corresponding to the TED in the wafer to be tested and the average value T _Avg(TED) are used to obtain the curvature Bow1, and the standard deviation corresponding to the TSD in the wafer to be tested is used. T _Std(TSD) and the average T _Avg(TSD) to obtain the bow2, and use the standard deviation T _Std(BPD) corresponding to the BPD in the wafer to be tested and the average T _Avg(BPD) to obtain the bow3. And, use the standard deviation T _Std(TED&TSD) , T _Std(TSD&BPD) , T _Std(TED&BPD) and the average value T _Avg corresponding to any two defect types of TED+TSD, TSD+BPD and TED+BPD in the wafer to be tested _(TED&TSD) , T _Avg(TED&TSD) , T _Avg(TSD&BPD) to obtain the curvature Bow4, Bow5, Bow6. Bow7 is obtained by using the standard deviation T _{Std (TED&TSD&BPD)} and the average value TAvg _{(TED&TSD&BPD)} corresponding to the three defect types of TED+TSD+BPD in the wafer to be tested. After that, the mean values of the degrees of curvature Bow1 to Bow7 were taken as the post-processing curvature value Pre_Bow.

To sum up, the present invention uses the processed wafers of the known ingots to establish the tortuosity prediction equation set, whereby the quality of the ingot to be tested can be predicted through the tortuosity prediction equation set, and then the quality of the ingot to be tested can be filtered out Processing ingots with poor geometric quality can greatly improve the overall processing quality and reduce production costs.

110: Measuring instruments 120: Analysis device 20-1～20-N: Ingot 21-1～21-N, W: Wafer 400: Wafer Image C1～C36: Area E1: first end E2: second end S305～S325: Steps of Ingot Evaluation Method

FIG. 1 is a block diagram of an analysis system according to an embodiment of the present invention. 2 is a schematic diagram of an ingot according to an embodiment of the present invention. FIG. 3 is a flowchart of an ingot evaluation method according to an embodiment of the present invention. FIG. 4 is a schematic diagram of a wafer image according to an embodiment of the present invention.

S305~S325: Steps of Ingot Evaluation Method

Claims (9)

An ingot evaluation method, comprising: Divide the respective wafer images corresponding to the plurality of wafers cut by the plurality of crystal ingots into a plurality of regions; calculating the statistical quantity of a plurality of defect types included in each of the plurality of regions; obtaining a plurality of statistical parameters based on the statistical quantities of the defect types included in the regions; performing a regression analysis using the statistical parameters and a warp value corresponding to each of the wafers to obtain a plurality of regression coefficients; and A tortuosity prediction equation system is established based on the regression coefficients. The ingot evaluation method according to claim 1, wherein based on the statistical quantity of the defect types included in the regions, the step of obtaining the statistical parameters includes: For any of these defect types, calculating a statistical number of any defect type in each of those areas; calculating the standard deviation of the statistical quantities of the regions corresponding to any one of the defect types as one of the statistical parameters; and Sorting the statistical quantities corresponding to any of the defect types in the areas, and extracting the A statistical quantities with the highest numerical values in a manner from a large value to a small value to calculate an average value as one of the statistical parameters, where A is a positive integer. The ingot evaluation method of claim 2, wherein the defect types include threading edge dislocations (TED), threading screw dislocations (TSD), and basal plane dislocations dislocation, BPD), using the statistical parameters corresponding to each of the wafers and the bending value to perform the regression analysis, and the steps of obtaining the regression coefficients include: The regression analysis is performed by inputting the statistical parameters and the warp value corresponding to each of the wafers into the following formula: Bow(k)=P1×Std(TED)+P2×Avg(TED); Bow(k)=P3×Std(TSD)+P4×Avg(TSD); Bow(k)=P5×Std(BPD)+P6×Avg(BPD); Among them, Bow(k) is the bending value of the kth wafer, P1-P6 are the regression coefficients, and Std(TED), Std(TSD), and Std(BPD) are the regions corresponding to the through-blade shape, respectively. The standard deviation of the statistical quantity of the three defect types of dislocation, threading dislocation and basal plane dislocation, Avg(TED), Avg(TSD), Avg(BPD) are corresponding to threading edge dislocation, threading thread dislocation, respectively and the average of the highest A statistics for the three defect types of basal plane dislocations. The ingot evaluation method according to claim 1, wherein based on the statistical quantity of the defect types included in the regions, the step of obtaining the statistical parameters includes: For any two of these defect types, calculating a statistical number of the sum of any two defect types in each of those areas; calculating the standard deviation of the statistical quantities of the regions corresponding to the sum of any two defect types as one of the statistical parameters; and Sort the statistical quantities corresponding to the sum of any two defect types in the regions, and take out the highest A statistical quantities to calculate an average value as one of the statistical parameters. The ingot evaluation method of claim 4, wherein the defect types include threading edge dislocations, threading screw dislocations and basal plane dislocations, using the statistical parameters corresponding to each of the wafers and the warp value to perform the regression analysis, and the steps of obtaining the regression coefficients include: The regression analysis is performed by inputting the statistical parameters and the warp value corresponding to each of the wafers into the following formula: Bow(k)=P7×Std(TED&TSD)+P8×Avg(TED&TSD);Bow(k)=P9×Std(TSD&BPD)+P10×Avg(TSD&BPD);Bow(k)=P11×Std(TED&BPD)+P12×Avg(TED&BPD); Among them, Bow(k) is the bending value of the kth wafer, P7-P12 are the regression coefficients, Std(TED&TSD) is the standard deviation of the statistical quantity corresponding to the sum of the two defect types of threading edge dislocations and threading dislocations in these regions, Avg(TED&TSD) is corresponding to both threading edge dislocations and threading threading dislocations The average value of the highest A statistical quantities among the total statistical quantities of the defect types, Std(TSD&BPD) is the standard deviation of the statistical number of these regions corresponding to the sum of the two defect types of threading dislocation and basal plane dislocation, Avg(TSD&BPD) is the two defects corresponding to threading dislocation and basal plane dislocation The average of the highest A statistics in the category total statistics, Std(TED&BPD) is the standard deviation of the statistical quantity corresponding to the sum of the two defect types of threading edge dislocation and basal plane dislocation in these regions, and Avg(TED&BPD) is corresponding to the two types of threading edge dislocation and basal plane dislocation The average of the highest A statistics among the total statistics of the defect types. The ingot evaluation method according to claim 1, wherein based on the statistical quantity of the defect types included in the regions, the step of obtaining the statistical parameters includes: calculating the aggregated statistics of the defect types in each of the areas; calculating the standard deviation of the statistical quantities of the regions corresponding to the total defect types as one of the statistical parameters; and Sort the statistical quantities corresponding to the total of the defect types in the regions, and extract the highest A statistical quantities to calculate an average value as one of the statistical parameters, where A is a positive integer. The ingot evaluation method of claim 6, wherein the defect types include threading edge dislocations, threading screw dislocations and basal plane dislocations, using the statistical parameters corresponding to each of the wafers and the warp value to perform the regression analysis, and the steps of obtaining the regression coefficients include: The regression analysis is performed by inputting the statistical parameters and the warp value corresponding to each of the wafers into the following formula: Bow(k)=P13×Std(TED&TSD&BPD)+P14×Avg(TED&TSD&BPD); Among them, Bow(k) is the bending value of the kth wafer, P13-P14 are the regression coefficients, Std(TED&TSD&BPD) is the standard deviation of the statistical number corresponding to the total of three defect types of threading edge dislocation, threading screw dislocation and basal plane dislocation in these regions, Avg(TED&TSD&BPD) is corresponding to threading edge dislocation, threading The average value of the highest A statistics among the total statistics of the three defect types of screw dislocations and basal plane dislocations. The ingot evaluation method according to claim 1, wherein after establishing the tortuosity prediction equation system based on the regression coefficients, further comprising: Using the set of curvature prediction equations to calculate a post-processing curvature value of the wafer to be tested obtained by processing the ingot to be tested; Determine whether the bending value after processing is within a specification range; If the bending value after processing is within the specification range, the quality of the ingot to be tested is determined to be good; and If the bending value after processing is not within the specification range, it is determined that the quality of the ingot to be tested is not good. The ingot evaluation method according to claim 1, wherein after the step of dividing the wafer images corresponding to the wafers cut from the ingot into the regions, further comprising: The regions located at the four corners of the wafer image are discarded, and the statistics of the defect types included in the remaining regions are calculated.

Images