Sample defect analysis algorithm and wafer-level defect analysis system

Abstract

A sample defect analysis algorithm with pixel-level accuracy is disclosed. The algorithm is pre-trained using a first optical image of at least one reference sample to generate a first matrix and a second matrix. A maximum value in the second matrix is ​​defined as an upper limit, and a minimum value in the second matrix is ​​defined as a lower limit. The pre-trained algorithm is adapted to analyze a second optical image of a sample under test to generate a third matrix and a fourth matrix, wherein the second optical image includes a plurality of second pixel units. When any element in the fourth matrix is ​​greater than the upper limit or less than the lower limit, the algorithm identifies that matrix element as a defect element and identifies the second pixel unit corresponding to that defect element as a defect pixel unit. A wafer-level defect analysis system is also proposed.

Claims

1. A sample defect analysis algorithm with pixel-level accuracy, wherein the algorithm is pre-trained using a first optical image of at least one reference sample to generate a first matrix and a second matrix, wherein the first optical image includes a plurality of first pixel units, the first matrix includes a plurality of first matrix elements corresponding to the plurality of first pixel units, the second matrix includes a plurality of second matrix elements corresponding to the plurality of first pixel units, the plurality of first matrix elements are defined by a first average brightness of the first optical image, the plurality of second matrix elements are the difference between the brightness of the plurality of first pixel units and the first average brightness, wherein a maximum value among the plurality of second matrix elements is defined as an upper limit, and a minimum value among the plurality of second matrix elements is defined as a lower limit. The pre-trained algorithm is adapted to analyze a second optical image of a sample to generate a third matrix and a fourth matrix, wherein the second optical image includes a plurality of second pixel units, the third matrix includes a plurality of third matrix elements corresponding to the plurality of second pixel units, and the fourth matrix includes a plurality of fourth matrix elements corresponding to the plurality of second pixel units. The plurality of third matrix elements are defined by a second average brightness of the second optical image, and the plurality of fourth matrix elements are the difference between the brightness of the plurality of second pixel units and the second average brightness. When the fourth matrix element is greater than the upper limit or less than the lower limit, the algorithm determines the fourth matrix element as a defective element and determines the second pixel unit corresponding to the defective element as a defective pixel unit.

2. The sample defect analysis algorithm as described in claim 1, wherein the first optical image and the second optical image are microscopic images generated by second harmonic signals.

3. The sample defect analysis algorithm as described in claim 1, wherein the algorithm does not include edge detection.

4. The sample defect analysis algorithm as described in claim 1, wherein the reference sample and the sample to be tested are asymmetric central crystals.

5. The sample defect analysis algorithm as described in claim 4, wherein the reference sample and the sample to be tested are two-dimensional materials.

6. The sample defect analysis algorithm as described in claim 4, wherein the reference sample and the sample to be tested comprise one of gallium nitride (GaN), aluminum gallium nitride (AlGaN), aluminum nitride (AlN), and silicon carbide (SiC).

7. The sample defect analysis algorithm as described in claim 1, wherein the algorithm further includes 8-connected markers to generate a defect distribution map.

8. A wafer-level defect analysis system, comprising: The detection module includes a computing unit, wherein the computing unit includes the sample defect analysis algorithm as described in claim 1. And a microscopic imaging module, connected to the detection module and used to provide the first optical image and the second optical image.

9. The wafer-level defect analysis system as described in claim 8, wherein the computing unit is a graphics processor or a high-performance AI chip.

10. The wafer-level defect analysis system as claimed in claim 8, wherein the microscopic imaging module is a second harmonic signal generating microscopic imaging module.

Images