TWI823778B - Epitaxial silicon wafer detection method and device - Google Patents

Abstract

The invention provides an epitaxial silicon wafer detection method and device, which belongs to the technical field of semiconductor manufacturing. The epitaxial silicon wafer detection method includes: controlling visible light to illuminate the epitaxial silicon wafer to be detected, obtaining scattered light information generated by reflecting the visible light on the surface of the epitaxial silicon wafer; judging the obtained scattered light information, and outputting the epitaxial silicon wafer. The first test result of the crystalline silicon wafer, the first test result includes qualified or unqualified; when the first test result is unqualified, the epitaxial silicon wafer is visually inspected to obtain the second test result, the second test result is The test results include qualified or unqualified; when the second test result is qualified, the epitaxial silicon wafer is judged to be qualified; when the second test result is unqualified, the epitaxial silicon wafer is judged to be unqualified.

Description

Epitaxial silicon wafer detection method and device

Embodiments of the present invention belong to the field of semiconductor manufacturing technology, and particularly relate to an epitaxial silicon wafer detection method and device.

Epitaxy growth refers to the technical process of growing a single crystal thin film (the crystallographic direction is consistent with the crystallographic direction of the substrate) on a single crystal silicon substrate through epitaxy technology. The entire production process of epiwafers includes growing (silicon rods drawn from polycrystalline silicon) → shaping (slicing and grinding) → polishing (double-sided polishing) → cleaning (removing surface particles, metal ions and organic matter) → epitaxy (vapor deposition) ) five major steps, among which epitaxy is the last important step, which can improve the crystal properties, original defects, resistivity and flatness of the polished wafer.

During the epitaxial growth process, many defects will appear on the epitaxial layer, which are divided into two categories according to their location: ① surface defects; ② internal defects. Surface defects refer to defects exposed on the surface of the epitaxial layer, which can be observed with the naked eye or a metallographic microscope. The main manifestations are: cloudy surface, pyramids, scratches, stars, pits, etc. In vivo defects refer to crystal structure defects located inside the epitaxial layer, mainly including: dislocations and stacking faults. Broadly speaking, defects also include impurities such as oxygen, carbon, heavy metals, and point defects such as atomic vacancies and interstitial atoms. The presence of these defects directly affects the performance of the semiconductor.

Various defects in the epitaxial layer are not only related to the quality of the substrate and the surface condition of the substrate, but are also closely related to the epitaxial growth process itself. The haze-like surface is a defect that exists on the surface of the epitaxial layer. Halo refers to the haze that appears on the back of the silicon wafer. A visual detector is used to detect the presence of a base-like mark on the back edge of the silicon wafer, which forms The sources include: ① Trichlorosilane (TCS) enters the holes on the surface of the base from the gap between the edge of the silicon wafer and the base, and is deposited on the back of the silicon wafer at high temperature; ② The polycrystalline silicon on the surface of the base is deposited on the back of the silicon wafer at high temperature, forming Halo with pedestal imprint.

At present, the detection method of surface defects of epitaxial silicon wafers mainly uses automated detection. Although the detection efficiency is improved, it may cause over-detection or missed detection. Manual visual inspection can improve the accuracy of inspection, but it also has problems such as low efficiency.

In order to solve the above technical problems, the present invention provides an epitaxial silicon wafer detection method and device, which can improve the detection efficiency and detection accuracy of surface defects of epitaxial silicon wafers.

In order to achieve the above objects, the technical solutions adopted in the embodiments of the present invention are: A method for detecting epitaxial silicon wafers, including: Control the visible light to illuminate the epitaxial silicon wafer to be detected, and obtain the scattered light information generated by the visible light reflected from the surface of the epitaxial silicon wafer; Judge the acquired scattered light information and output a first test result of the epitaxial silicon wafer, where the first test result includes passing or failing; When the first test result is unqualified, visually inspect the epitaxial silicon wafer to obtain a second test result, where the second test result includes passing or failing; When the second detection result is qualified, the epitaxial silicon wafer is determined to be qualified; when the second detection result is unqualified, the epitaxial silicon wafer is determined to be unqualified.

In some embodiments, judging the acquired scattered light information and outputting the first detection result of the epitaxial silicon wafer includes: The acquired scattered light information is input into a pre-trained defect detection model, and the first detection result of the epitaxial silicon wafer is output. The input of the defect detection model is the scattered light information generated by the visible light reflected from the surface of the epitaxial silicon wafer, and the output This is the first test result of epitaxial silicon wafers.

In some embodiments, when the second detection result is qualified, the defect detection model is updated according to the second detection result and the scattered light information of the epitaxial silicon wafer.

In some embodiments, the epitaxial silicon wafer detection method further includes: Select a preset proportion of epitaxial silicon wafers from the epitaxial silicon wafers whose first test results are qualified, perform a visual inspection on the epitaxial silicon wafers, and obtain a third test result, where the third test result includes qualified or unqualified; When the third detection result is qualified, the selected epitaxial silicon wafer is determined to be qualified; when the third detection result is unqualified, the selected epitaxial silicon wafer is determined to be unqualified.

In some embodiments, the preset ratio is 1% to 3%.

In some embodiments, the epitaxial silicon wafer detection method further includes: when the third detection result is unqualified, updating the defect detection model according to the third detection result and the corresponding scattered light information of the epitaxial silicon wafer.

In some embodiments, the epitaxial silicon wafer detection method further includes the step of training to obtain the defect detection model. Training to obtain the defect detection model includes: Establish an initial defect detection model. The input of the initial defect detection model is the scattered light information of the epitaxial silicon wafer, and the output is the detection result of the epitaxial silicon wafer. The detection result includes qualified or unqualified; Obtain multiple sets of training data, each set of training data includes scattered light information of epitaxial silicon wafers and corresponding manual visual inspection results; The initial defect detection model is trained using the multiple sets of training data to obtain the defect detection model.

An embodiment of the present invention also provides an epitaxial silicon wafer detection device, including: The scattered light information acquisition module is used to control visible light to illuminate the epitaxial silicon wafer to be detected, and obtain the scattered light information generated by the visible light reflected from the surface of the epitaxial silicon wafer; A detection module is used to judge the acquired scattered light information and output a first detection result of the epitaxial silicon wafer, where the first detection result includes passing or failing; The re-inspection module is used to visually inspect the epitaxial silicon wafer when the first inspection result is unqualified, and obtain a second inspection result, where the second inspection result includes passing or unqualified; The determination module is used to determine that the epitaxial silicon wafer is qualified when the second detection result is qualified; and determine that the epitaxial silicon wafer is unqualified when the second detection result is unqualified.

In some embodiments, the re-inspection module is also used to select a preset proportion of epitaxial silicon wafers from the epitaxial silicon wafers with qualified first test results, and perform visual inspection on the epitaxial silicon wafers to obtain the third test. As a result, the third test result includes passing or failing; The determination module is also used to determine that the selected epitaxial silicon wafer is qualified when the third detection result is qualified; and determine that the selected epitaxial silicon wafer is unqualified when the third detection result is unqualified.

In some embodiments, the preset ratio is 1% to 3%.

The beneficial effects of the present invention are: In this embodiment, the epitaxial silicon wafer is automatically detected based on the scattered light information reflected from the surface of the epitaxial silicon wafer, and the first detection result of the epitaxial silicon wafer is output. When the first detection result is unqualified, the epitaxial silicon wafer is The wafer is visually inspected to obtain a second test result. When the second test result is qualified, the epitaxial silicon wafer is determined to be qualified; when the second test result is unqualified, the epitaxial silicon wafer is determined to be unqualified. . This embodiment can combine automated detection and manual visual inspection to realize the detection of epitaxial silicon wafer defects, and can improve the detection efficiency and detection accuracy of epitaxial silicon wafer defects.

The above features and advantages and other features and advantages of the present invention will be more apparent from the following detailed description of exemplary embodiments of the present invention in conjunction with the accompanying drawings.

In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, but are not used to limit the present invention.

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments These are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those with ordinary skill in the art without making progressive improvements shall fall within the scope of protection of the present invention.

In addition, it should be noted that in the embodiments of the present invention, the terms "first", "second", etc. are used to distinguish similar objects and are not used to describe a specific order or sequence. It is to be understood that the terms so used are interchangeable under appropriate circumstances so that embodiments of the invention can be practiced in sequences other than those illustrated or described herein, and "first" and "second" distinctions are generally intended to It is a category, and the number of objects is not limited. For example, the first object can be one or multiple.

Those of ordinary skill in the art will understand that, unless expressly stated otherwise, the singular forms "a", "an", "the" and "the" used herein may also include the plural form. It should be further understood that the word "comprising" used in the description of the present invention refers to the presence of stated features, integers, steps, operations, elements and/or elements, but does not exclude the presence or addition of one or more other features, Integers, steps, operations, elements, components and/or groups thereof. It will be understood that when we refer to an element being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Additionally, "connected" or "coupled" as used herein may include wireless connections or wireless couplings. As used herein, the term "and/or" includes all or any unit and all combinations of one or more of the associated listed items.

In the embodiment of the present invention, the term "and/or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and/or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. Three situations. The character "/" generally indicates that the related objects are in an "or" relationship.

In the embodiment of the present invention, the term "multiple" refers to two or more than two, and other quantifiers are similar to it.

“Determining B based on A” in the present invention means that the factor A should be considered when determining B. It is not limited to "B can be determined based on A alone", but also includes: "B is determined based on A and C", "B is determined based on A, C and E", "C is determined based on A, and B is further determined based on C" wait. In addition, it can also include using A as a condition for determining B, for example, "When A meets the first condition, use the first method to determine B"; another example, "When A meets the second condition, determine B", etc.; another example , "When A meets the third condition, determine B based on the first parameter" and so on. Of course, it can also be a condition that uses A as a factor to determine B, for example, "when A meets the first condition, use the first method to determine C, and further determine B based on C" and so on.

Figure 1 shows a schematic diagram of the backside Halo of an epitaxial silicon wafer; its formation source is shown in Figure 2. When performing epitaxial crystal technology, the path of trichlorosilane (TCS) flowing through the surface of the silicon wafer includes: ① TCS flows from the silicon wafer The holes on the surface of the base enter into the gap between the edge and the base, and are deposited on the back of the silicon wafer at high temperatures; ② The polycrystalline silicon on the surface of the base is deposited on the back of the silicon wafer at high temperatures, forming the Halo imprinted by the base.

At present, the detection method of surface defects of epitaxial silicon wafers mainly uses automated detection. Although the detection efficiency is improved, it may cause over-detection or missed detection. Manual visual inspection can improve the accuracy of inspection, but it also has problems such as low efficiency. 2,000 pieces of epitaxial silicon wafers were tested, and the results of automated testing and manual visual testing were analyzed. As shown in Table 1, the accuracy of manual visual testing was 100%, and the accuracy of automated testing was 97.35%. It can be seen that , the accuracy of automated detection is low.

Table 1 Test quantity Good product Backside Halo Others Yield Accuracy Manual visual inspection 2000 1925 15 60 96.25% 100.00% Automated detection 2000 1874 40 86 92.85% 97.35%

The size of the Backside Halo halo is different, and the brightness of the halo under lighting conditions is different. As the lighting angle is different, the appearance of the Backside Halo halo will also change. These special properties are quite challenging for automated detection, so misjudgments or missed determinations may occur during the automated detection process.

The invention provides a method and device for detecting epitaxial silicon wafers, which can improve the detection efficiency and detection accuracy of surface defects of epitaxial silicon wafers.

An embodiment of the present invention provides a method for detecting epitaxial silicon wafers, as shown in Figure 3, including: Step 101: Control visible light to illuminate the epitaxial silicon wafer to be detected, and obtain scattered light information generated by reflecting the visible light on the surface of the epitaxial silicon wafer; Step 102: Judge the acquired scattered light information and output a first detection result of the epitaxial silicon wafer, where the first detection result includes passing or failing; Step 103: When the first test result is unqualified, perform a visual inspection on the epitaxial silicon wafer to obtain a second test result. The second test result includes qualified or unqualified; Step 104: When the second detection result is qualified, determine that the epitaxial silicon wafer is qualified; when the second detection result is unqualified, determine that the epitaxial silicon wafer is unqualified.

In this embodiment, the epitaxial silicon wafer is automatically detected based on the scattered light information reflected from the surface of the epitaxial silicon wafer, and the first detection result of the epitaxial silicon wafer is output. When the first detection result is unqualified, the epitaxial silicon wafer is The wafer is visually inspected to obtain a second test result. When the second test result is qualified, the epitaxial silicon wafer is determined to be qualified; when the second test result is unqualified, the epitaxial silicon wafer is determined to be unqualified. . This embodiment can combine automated detection and manual visual inspection to realize the detection of epitaxial silicon wafer defects, and can improve the detection efficiency and detection accuracy of epitaxial silicon wafer defects.

This embodiment can detect surface defects including Backside Halo. The epitaxial silicon wafers that pass the detection can be processed in subsequent processes, and the epitaxial silicon wafers that fail the detection can be scrapped.

In some embodiments, judging the acquired scattered light information and outputting the first detection result of the epitaxial silicon wafer includes: The acquired scattered light information is input into a pre-trained defect detection model, and the first detection result of the epitaxial silicon wafer is output. The input of the defect detection model is the scattered light information generated by the visible light reflected from the surface of the epitaxial silicon wafer, and the output This is the first test result of epitaxial silicon wafers. The defect detection model can be used to automatically detect epitaxial silicon wafers and improve the detection efficiency of surface defects of epitaxial silicon wafers.

In some embodiments, the epitaxial silicon wafer detection method further includes: when the second detection result is qualified, updating the defect detection model based on the second detection result and the scattered light information of the epitaxial silicon wafer.

This embodiment compares the manual visual inspection results with the automatic inspection results to continuously improve the defect detection model, which can improve the automatic detection capability of epitaxial silicon wafer defects and help improve the yield rate of epitaxial silicon wafers.

In some embodiments, the epitaxial silicon wafer detection method further includes: Select a preset proportion of epitaxial silicon wafers from the epitaxial silicon wafers whose first test results are qualified, perform a visual inspection on the epitaxial silicon wafers, and obtain a third test result, where the third test result includes qualified or unqualified; When the third detection result is qualified, the selected epitaxial silicon wafer is determined to be qualified; when the third detection result is unqualified, the selected epitaxial silicon wafer is determined to be unqualified.

In this way, automated inspection and manual visual inspection can be combined to detect defects in epitaxial silicon wafers, which can improve the detection efficiency and accuracy of defects in epitaxial silicon wafers.

In some embodiments, the epitaxial silicon wafer detection method further includes: when the third detection result is unqualified, updating the defect detection model according to the third detection result and the corresponding scattered light information of the epitaxial silicon wafer.

In this way, the defect detection model can be continuously improved, the automatic detection capability of epitaxial silicon wafer defects can be improved, and the yield rate of epitaxial silicon wafers can be improved.

In some embodiments, the preset ratio may be 1% to 3%, for example, the preset ratio may be 1%, or other values, such as 2%, 3%, etc. The higher the preset ratio is set, the more conducive it is to improving the accuracy of surface defect detection, but it will increase the amount of calculation and detection cost, because the preset ratio can be set to 1% to 3%.

In some embodiments, the epitaxial silicon wafer detection method further includes the step of training to obtain the defect detection model. Training to obtain the defect detection model includes: Establish an initial defect detection model. The input of the initial defect detection model is the scattered light information of the epitaxial silicon wafer, and the output is the detection result of the epitaxial silicon wafer. The detection result includes qualified or unqualified; Obtain multiple sets of training data, each set of training data includes scattered light information of epitaxial silicon wafers and corresponding manual visual inspection results; The initial defect detection model is trained using the multiple sets of training data to obtain the defect detection model. The defect detection model can be used to automatically detect defects in epitaxial silicon wafers.

In this embodiment, the epitaxial silicon wafers in the multiple sets of training data obtained may include epitaxial silicon wafers with various surface defects. In this way, the trained defect detection model can detect epitaxial silicon wafers with various surface defects. Defects can be cloudy surfaces, pyramids, scratches, stars, pockmarks, etc.

In a specific example, the epitaxially grown silicon wafers were pre-cleaned and then entered into automated testing. The automated testing tested 100% of the epitaxial silicon wafers. Epitaxial silicon wafers containing Backside Halo were detected and were judged to be NG (unqualified). ), the epitaxial silicon wafers without Backside Halo are judged as OK (passed); the epitaxial silicon wafers judged as NG are subject to manual visual re-inspection. As shown in Figure 4, the epitaxial silicon wafer judged to be NG is placed on the visual inspection stage 3. The visual inspection stage 3 can be rotated and tilted. The epitaxial silicon wafer 1 can be rotated and tilted. Under the visible light emitted by the visible light source 2 Under irradiation, use the visual inspection stage 3 to tilt and rotate the epitaxial silicon wafer 1. If the Backside Halo cannot be seen at all, the epitaxial silicon wafer 1 will be judged as OK. If it can be seen, the epitaxial silicon wafer 1 will be judged as NG. This can prevent over-detection by automated inspection; in addition, manual visual inspection of epitaxial silicon wafers judged as OK by automated inspection, with a sampling frequency of 1%, is performed to prevent missed inspections by automated inspection. Epitaxial silicon wafers that are tested OK can be shipped, while epitaxial silicon wafers that are judged to be NG are scrapped.

In this embodiment, the manual visual inspection results are compared with the automated inspection results. If they are inconsistent, the defect detection model is updated according to the manual visual inspection results, that is, the defect detection model is trained using the manual visual inspection results, so that the defect detection can be continuously improved. model to improve the automated detection capabilities of Backside Halo. As shown in Table 2, using the technical solution of this embodiment can increase the accuracy of automated detection results to more than 99%.

Table 2 Test quantity Good product Backside Halo Others Yield Accuracy Manual visual inspection 2000 1940 10 50 97.00% 100.00% Automated detection 2000 1935 13 52 96.75% 99.74%

An embodiment of the present invention also provides an epitaxial silicon wafer detection device, as shown in Figure 5, including: The scattered light information acquisition module 21 is used to control the visible light to be irradiated on the epitaxial silicon wafer to be detected, and obtain the scattered light information generated by the visible light reflected from the surface of the epitaxial silicon wafer; The detection module 22 is used to judge the acquired scattered light information and output a first detection result of the epitaxial silicon wafer, where the first detection result includes passing or failing; The re-inspection module 23 is used to visually inspect the epitaxial silicon wafer when the first inspection result is unqualified, and obtain a second inspection result, where the second inspection result includes passing or unqualified; The determination module 24 is configured to determine that the epitaxial silicon wafer is qualified when the second detection result is qualified; and determine that the epitaxial silicon wafer is unqualified when the second detection result is unqualified.

In this embodiment, the epitaxial silicon wafer is automatically detected based on the scattered light information reflected from the surface of the epitaxial silicon wafer, and the first detection result of the epitaxial silicon wafer is output. When the first detection result is unqualified, the epitaxial silicon wafer is The wafer is visually inspected to obtain a second test result. When the second test result is qualified, the epitaxial silicon wafer is determined to be qualified; when the second test result is unqualified, the epitaxial silicon wafer is determined to be unqualified. . This embodiment can combine automated detection and manual visual inspection to realize the detection of epitaxial silicon wafer defects, and can improve the detection efficiency and detection accuracy of epitaxial silicon wafer defects.

In some embodiments, the re-inspection module is also used to select a preset proportion of epitaxial silicon wafers from the epitaxial silicon wafers with qualified first test results, and perform visual inspection on the epitaxial silicon wafers to obtain the third test. As a result, the third test result includes passing or failing; The determination module is also used to determine that the selected epitaxial silicon wafer is qualified when the third detection result is qualified; and determine that the selected epitaxial silicon wafer is unqualified when the third detection result is unqualified.

In this way, automated inspection and manual visual inspection can be combined to detect defects in epitaxial silicon wafers, which can improve the detection efficiency and accuracy of defects in epitaxial silicon wafers.

In some embodiments, the preset ratio may be 1 to 3%, for example, the preset ratio may be 1%, or other values, such as 2%, 3%, etc. The higher the preset ratio is set, the more conducive it is to improving the accuracy of surface defect detection, but it will increase the amount of calculation and detection cost, because the preset ratio can be set to 1 to 3%.

It should be noted that each embodiment in this specification is described in a progressive manner, and the same and similar parts between the various embodiments can be referred to each other. Each embodiment focuses on its differences from other embodiments. In particular, the embodiments are described simply because they are basically similar to the product embodiments. For relevant details, please refer to the partial description of the product embodiments.

In the above description of the embodiments, specific features, structures, materials or characteristics may be combined in any suitable manner in any one or more embodiments or examples.

It will be apparent to those of ordinary skill in the art that various modifications and variations can be made to the present invention without departing from the spirit and scope of the invention. In this way, if these modifications and variations of the present invention fall within the scope of the patent application of the present invention and the scope of equivalent technologies, the present invention is also intended to include these modifications and variations.

1: Epitaxial silicon wafer 2:Light source 3:Visual inspection platform 21: Scattered light information acquisition module 22:Detection module 23:Recheck module 24:Judgment module 101-104: Steps

Figure 1 shows a schematic diagram of the Halo on the back of the epitaxial silicon wafer; Figure 2 shows a schematic diagram of the path of silicon source gas flowing through the surface of the silicon wafer; Figure 3 shows a schematic flow chart of an epitaxial silicon wafer detection method according to an embodiment of the present invention; Figure 4 shows a schematic diagram of re-inspection of epitaxial silicon wafers according to an embodiment of the present invention; FIG. 5 shows a schematic structural diagram of an epitaxial silicon wafer detection device according to an embodiment of the present invention.

101-104: Steps

Claims (8)

A method for detecting epitaxial silicon wafers, including: controlling visible light to illuminate an epitaxial silicon wafer to be detected, obtaining scattered light information generated by reflecting the visible light on the surface of the epitaxial silicon wafer; judging the obtained scattered light information, and outputting The first test result of the epitaxial silicon wafer, the first test result includes qualified or unqualified; when the first test result is unqualified, the epitaxial silicon wafer is visually inspected to obtain the second test result, the The second test result includes qualified or unqualified; when the second test result is qualified, the epitaxial silicon wafer is determined to be qualified; when the second test result is unqualified, the epitaxial silicon wafer is determined to be unqualified; The epitaxial silicon wafer detection method also includes: selecting a preset proportion of epitaxial silicon wafers from the epitaxial silicon wafers with qualified first detection results, visually inspecting the epitaxial silicon wafers, and obtaining a third detection result. The three test results include qualified or unqualified; when the third test result is qualified, the selected epitaxial silicon wafer is judged to be qualified; when the third test result is unqualified, the selected epitaxial silicon wafer is judged to be unqualified. qualified. The epitaxial silicon wafer detection method as described in claim 1, wherein judging the acquired scattered light information and outputting the first detection result of the epitaxial silicon wafer includes: inputting the acquired scattered light information into a pre-trained In the defect detection model, the first detection result of the epitaxial silicon wafer is output. The input of the defect detection model is the scattered light information generated by the visible light reflected by the surface of the epitaxial silicon wafer, and the output is the first detection result of the epitaxial silicon wafer. Test results. The epitaxial silicon wafer detection method as described in claim 2, wherein when the second detection result is qualified, the defect detection model is updated according to the second detection result and the scattered light information of the epitaxial silicon wafer. The epitaxial silicon wafer detection method as described in claim 1, wherein the preset ratio is 1% to 3%. The epitaxial silicon wafer detection method as described in claim 1, the epitaxial silicon wafer detection method further includes: when the third detection result is unqualified, based on the third detection result and the scattering of the corresponding epitaxial silicon wafer Optical information updates the defect detection model. The epitaxial silicon wafer detection method as described in claim 2, wherein the epitaxial silicon wafer detection method also includes the step of training to obtain the defect detection model, and training to obtain the defect detection model includes: establishing an initial defect detection model, the initial defect detection model The input of the defect detection model is the scattered light information of the epitaxial silicon wafer, and the output is the detection result of the epitaxial silicon wafer, which includes qualified or unqualified; multiple sets of training data are obtained, and each set of training data includes the results of the epitaxial silicon wafer. Scattered light information and corresponding manual visual inspection results; use the multiple sets of training data to train the initial defect detection model to obtain the defect detection model. An epitaxial silicon wafer detection device, including: a scattered light information acquisition module, used to control visible light to illuminate the epitaxial silicon wafer to be detected, and obtain the scattered light information generated by the visible light reflected from the surface of the epitaxial silicon wafer; The detection module is used to judge the acquired scattered light information and output the first detection result of the epitaxial silicon wafer. The first detection result includes qualified or unqualified; the re-inspection module is used to perform the first detection on the obtained scattered light information. When the test result is unqualified, the epitaxial silicon wafer is visually inspected to obtain a second test result, and the second test result includes qualified or unqualified; the determination module is used to determine when the second test result is qualified. The epitaxial silicon wafer is determined to be qualified; when the second test result is unqualified, the epitaxial silicon wafer is determined to be unqualified; the re-inspection module is also used to select the epitaxial silicon wafers that are qualified as the first test result. Select a preset proportion of epitaxial silicon wafers, conduct visual inspection of the epitaxial silicon wafers, and obtain a third test result. The third test result includes qualified or unqualified; the judgment module is also used to determine the third test result. When the result is qualified, the selected epitaxial silicon wafer is determined to be qualified; when the third test result is unqualified, the selected epitaxial silicon wafer is determined to be unqualified. The epitaxial silicon wafer detection device as described in claim 7, wherein the preset ratio is 1% to 3%.

Images