CN112907542A

CN112907542A - Method for detecting defects of wafer back, storage medium and computer device

Info

Publication number: CN112907542A
Application number: CN202110205137.0A
Authority: CN
Inventors: 王泽逸; 庄均珺; 王恺; 陈旭
Original assignee: Shanghai Huali Integrated Circuit Manufacturing Co Ltd
Current assignee: Shanghai Huali Integrated Circuit Manufacturing Co Ltd
Priority date: 2021-02-24
Filing date: 2021-02-24
Publication date: 2021-06-04
Anticipated expiration: 2041-02-24
Also published as: CN112907542B

Abstract

The invention discloses a crystal back defect detection method, which comprises the following steps: step S1: obtaining a wafer back graph; step S2: filtering and differentiating the wafer back graph; step S3: distinguishing whether the wafer back graph is a baseline image or a highlight image; step S4: and identifying the defects of the crystal back through the characteristic difference of the image, and outputting alarm information. The invention can effectively identify the shapes of various defects of the wafer back by improving the accuracy of distinguishing the pattern types of the wafer back, effectively realize the distinguishing and automatic judgment among the defects of the wafer back and avoid the yield loss caused by abnormal wafer back.

Description

Method for detecting defects of wafer back, storage medium and computer device

Technical Field

The present invention relates to the field of semiconductor integrated circuit manufacturing, and more particularly, to a method for detecting defects on a back surface of a wafer, a storage medium, and a computer device.

Background

As semiconductor chip manufacturing processes are more and more precise, crystal face defects caused by crystal back influences are more and more greatly influenced, but scanning detection modes aiming at the crystal back defects still need to be judged manually, subjective influences are large, and the occurrence of online process abnormity is difficult to grasp in time.

The existing mainstream crystal back detection method mainly comprises two methods: 1. carrying out wafer back visual inspection 2, carrying out wafer back scanning; the visual inspection of the wafer back is mainly influenced by human factors, the judgment difference is large, and the wafer back scanning can effectively detect the wafer back abnormity.

However, it is far from sufficient to scan defects, the back-side scanning is different from the crystal-surface defect scanning, and the main reason is that the positioning observation of the defects cannot be performed, so the back-side defects are mainly determined by scanning maps and graphs, such as the publication CN111754480A, and a plurality of defect morphologies can simultaneously appear on the scanned graphs, as shown in fig. 1. Even if the method as in CN111754480A is adopted, the recognition rate is still low.

Disclosure of Invention

The invention aims to solve the technical problem of providing a crystal back defect detection method, which can distinguish and automatically judge defects, effectively identify various defect appearances of a crystal back and avoid yield loss caused by abnormal crystal back.

The invention provides a method for detecting defects of a wafer back, which comprises the following steps:

s1: obtaining a wafer back graph;

s2: filtering and differentiating the wafer back graph;

s3: distinguishing whether the wafer back graph is a baseline image or a highlight image;

s4: and identifying the defects of the crystal back through the characteristic difference of the image, and outputting alarm information.

Optionally, in step S3, the back of wafer pattern is distinguished from the baseline image or the highlight image according to the same judgment result for 2 or more times through N times of detection, where N is greater than or equal to 3, and a different detection rule is set for each detection. And further, N is equal to 3, three times of detection are carried out, different detection rules are set for each time of detection, and when the crystal back graph is judged to be a highlight image for at least 2 times, the crystal back graph is determined to be the highlight image.

Optionally, in step S3, a detection rule is set to identify a batch of the wafer back patterns, and if a certain wafer back pattern is determined to be a highlight image, the highlight image is defined as a contrast image; tracing the M wafer back graphics which are judged as baseline images before, comparing the wafer back graphics with the comparison images, judging the M wafer back graphics again, and if the judgment result is a highlight image, updating the judgment result to be the highlight image, wherein M is more than or equal to 30. Further, M is equal to 50.

Optionally, in step S3, a detection rule is set to identify a batch of the wafer back patterns, and if a certain wafer back pattern is determined to be a baseline image, the baseline image is defined as a comparison image; and tracing the back graph of the P wafers judged as the highlight image before, comparing the back graph with the comparison image, judging the back graph of the P wafers again, and if the judgment result is a baseline image, updating the judgment result to be the baseline image, wherein P is more than or equal to 30. Further, P is equal to 50.

Based on the same inventive concept, the present invention further provides a computer-readable storage medium, wherein computer-executable instructions are stored on the computer-readable storage medium, and when the computer-executable instructions are executed, the method for detecting the back defect of the wafer as described in any one of the above items is implemented.

Based on the same inventive concept, the invention further provides a computer device, which comprises a processor and a storage device, wherein the processor is suitable for realizing each instruction, the storage device is suitable for storing a plurality of instructions, and the instructions are suitable for being loaded by the processor and executing any one of the crystal back defect detection methods.

Compared with the prior art, the wafer back defect detection method can effectively identify various defect appearances of the wafer back by improving the accuracy of wafer back pattern type distinguishing, effectively realize the distinguishing and automatic judgment among the defects of the wafer back and avoid the yield loss caused by wafer back abnormity.

Drawings

FIG. 1 is a diagram of a back defect.

FIG. 2 is a schematic step diagram of example 1.

FIG. 3 is a graph showing test data of example 1.

FIG. 4 is a graph showing the test results of example 1.

FIG. 5 is a graph showing the test results of example 2.

FIG. 6 is a graph showing the test results of example 3.

FIG. 7 is a graph showing the test results of example 4.

FIG. 8 is a graph showing the test results of example 4.

Detailed Description

It should be noted that the greatest feature of the present invention, which is different from the prior art, is to distinguish whether the backside graph is the baseline image baseline or the highlight image highlight, so the following specific embodiment mainly illustrates how to improve the recognition rate of the baseline image and the highlight image through different rules and strategies. The wafer backside pattern and the final Defect recognition can be achieved by using existing equipment, such as a wafer image scanner, such as the Klaity Defect monitoring system, and the method disclosed in the document CN 111754480A.

Example 1:

the method for detecting defects on the back of a wafer as provided in the embodiment of fig. 2 includes:

s1: obtaining a wafer back graph;

s2: filtering and differentiating the wafer back graph;

The crystal back pattern recognition of the embodiment mainly carries out filtering and difference processing on the image with poor crystal back through the difference between the characteristic codes of the baseline image and the highlight image, highlights the detection characteristics, reduces the back bottom influence, and distinguishes the image after automatically generating rules according to the program.

To illustrate the technical effect of the present embodiment, the following is exemplified:

test data: a certain amount of basic database and a part of defect patterns needing to be alarmed are selected for testing, as shown in fig. 3.

The specific distinguishing method in step S3 is as follows: and carrying out tests on different basic databases and Test data in three times. As shown in fig. 4.

1. The base database databases are all baseline images baseline, and the Test data Test are all highlight images highlight.

2. The database databases are all baseline images baseline, and the Test data Test is all highlight images highlight and partial baseline images baseline

3. The basic database is a partial baseline image baseline and a partial highlight image highlight, the Test data is a partial baseline image baseline and a partial highlight image highlight

The detection rate of the Test result is an evaluation index for detecting highlight images in the Test data, and for example, the Test data in the first Test is 16 highlight images, and the Test data in the first Test is judged to be 13 highlight images, and 3 missing images exist, so the detection rate is 81%.

The overdetection rate in the Test result is an error detection evaluation index for the highlight image in the Test data, for example, the Test data Test in the second Test is 16 highlight images and 48 baseline images, and it is determined that 19 highlight images of the highlight image preview _ HL are detected, and 7 of the highlight images have an error detection, so the overdetection rate is 36.84%.

Example 2:

on the basis of embodiment 1, the specific discrimination method in step S3 is changed as follows:

and testing by adopting a single-chip multi-range comparison detection method, and performing tests on different basic databases and Test data three times, as shown in figure 5.

And randomly extracting 6 images from the wafer back graph of each batch as test data test 3 times, and taking the rest images as a base database. Each wafer back pattern is detected 3 times, and detection rules are set differently for 3 times. The grain back pattern judged to be the highlight image highlight twice or more out of the 3 times is finally determined to be the highlight image highlight.

From the test conclusion, it can be seen that the detection rate is effectively improved and the overdetection rate is reduced in the embodiment 2 compared with the embodiment 1.

Example 3:

the method adopts a rollback detection method for testing, compares and calculates test data and recent rollback data to obtain whether the judgment is recently and continuously generated, and comprises the following two specific methods:

the first (Appliach 1) is based on the highlight image highlight

Initially setting a detection rule to ensure that the detected highlight image highlight is certain correct and the over-detection rate is 0;

and detecting a single chip, if the chip is judged to be the highlight image highlight, tracing the graph of the first 50 chips which are judged to be the baseline image baseline, comparing each chip with the highlight image highlight, and judging whether the highlight image is missed to be detected or not secondarily.

The second (Approach2) is based on a baseline image baseline

A detection rule is initially set to ensure that the detected baseline image is correct and the detection rate is 100%;

detecting a single chip, if the chip is judged to be the baseline image baseline, tracing the graphs of the first 50 chips which are judged to be the highlight images highlight, comparing each chip with the chip baseline image baseline, and judging whether the highlight images are subjected to over-detection or not secondarily;

the test data and the test result are shown in fig. 6, the Highlight image Highlight is used as a reference, a rollback comparison method is used, the detection rate is about 96%, the over-detection rate is 17.5%, the Baseline image Baseline is used as a reference, the detection rate is about 88%, and the over-detection rate is about 28%.

Example 4:

this embodiment can be combined with the method of embodiment 3 on the basis of embodiment 2.

For example, we refer to the method of example 2 as the 2/3 method and the method of example 3 as the rollback method.

The 2/3 method is combined with the rollback method to be more suitable for the condition that the occurrence frequency of highlight images is high, such as continuity abnormity shown in fig. 3, and the detection rate can be effectively improved by increasing the detected times of each image. The test results are shown in fig. 7 and 8. Compared with the embodiment 2, the detection rate is effectively improved, and the over-detection rate is reduced.

The present invention has been described in detail with reference to the specific embodiments and examples, but these are not intended to limit the present invention. Many variations and modifications may be made by one of ordinary skill in the art without departing from the principles of the present invention, which should also be considered as within the scope of the present invention.

Claims

1. A method for detecting defects of a back of a crystal is characterized by comprising the following steps:

step S1: obtaining a wafer back graph;

step S2: filtering and differentiating the wafer back graph;

step S3: distinguishing whether the wafer back graph is a baseline image or a highlight image;

step S4: and identifying the defects of the crystal back through the characteristic difference of the image, and outputting alarm information.

2. The method of claim 1, wherein:

in step S3, the back of wafer pattern is distinguished as a baseline image or a highlight image according to the same judgment result for 2 or more times through N times of detection, where N is greater than or equal to 3, and different detection rules are set for each detection.

3. The method of claim 2, wherein:

and N is equal to 3, three times of detection are carried out, different detection rules are set for each time of detection, and when the crystal back graph is judged to be a highlight image for at least 2 times, the crystal back graph is determined to be the highlight image.

4. The method of claim 1, wherein:

in step S3, setting a detection rule, identifying a batch of the wafer back patterns, and if a certain wafer back pattern is determined to be a highlight image, defining the highlight image as a contrast image;

tracing the M wafer back graphics which are judged as baseline images before, comparing the wafer back graphics with the comparison images, judging the M wafer back graphics again, and if the judgment result is a highlight image, updating the judgment result to be the highlight image, wherein M is more than or equal to 30.

5. The method of claim 4, wherein: said M is equal to 50.

6. The method of claim 1, wherein:

in step S3, setting a detection rule to identify a batch of the wafer back patterns, and if a certain wafer back pattern is determined to be a baseline image, defining the baseline image as a comparison image;

and tracing the back graph of the P wafers judged as the highlight image before, comparing the back graph with the comparison image, judging the back graph of the P wafers again, and if the judgment result is a baseline image, updating the judgment result to be the baseline image, wherein P is more than or equal to 30.

7. The method of claim 6, wherein: said P is equal to 50.

8. A computer-readable storage medium having computer-executable instructions stored thereon, the computer-readable storage medium characterized by: the computer-executable instructions, when executed, implement the backside defect detection method of any one of claims 1 to 7.

9. A computer device comprising a processor adapted to implement instructions and a storage device adapted to store instructions adapted to be loaded by the processor and to perform the method of any of claims 1 to 7.