SEM graph measurement method based on AI algorithm

Abstract

The invention discloses an SEM graph measurement method based on an AI algorithm, and the method comprises the following steps: employing the AI technology to achieve the measurement of an SEM graph,and providing an AI measurement scheme which consists of three modules: 1, scale identification on an image: employing a deep learning technology to provide a pixel unit length for final measurement;2, template matching: adopting an FLANN template matching technology to accurately position a part to be measured, and obtaining a local image; and 3, measuring a specific SEM image: measuring the size of the component by adopting technologies such as binarization and edge recognition. The method increases automation levels. And the working efficiency is improved because the AI model is not tiredand people are tired, and the workload of the AI model per unit time is much higher than that of the people. And the measurement cost is reduced because an engineer is replaced by the AI model, and the working efficiency is greatly improved. Based on a real-time measurement result, an engineer can adjust process parameters in time, and the yield is improved.

Description

technical field [0001] The invention relates to the field of AI, in particular to an AI algorithm-based SEM graphic measurement method. Background technique [0002] Measurement is an important judgment basis used by semiconductor manufacturing companies to evaluate whether the process is up to standard during the manufacturing process. It includes measurement of circuit pattern size (CD), thickness measurement, alignment measurement, and other physical forms (such as measurement of graphic angle, depth, etc.). [0003] At present, the main measurement method is to design a special test pattern (Testkey) on the dicing line between chips, and then use optical or SEM imaging technology to generate the pattern and then measure it. Then, through destructive measurement (TEM) of the real chip graphics, etc., the measurement offset between the TESTKEY and the real graphics is established, so as to establish the measurement control flag of the TESTKEY to meet the process standard....

AI Technical Summary

Problems solved by technology

[0005] 1. The stability of the measurement results is very poor

[0006] 2. The measurement result is unstable and inconsistent with the offset of the real circuit

That is, the measurement results cannot reflect the real situation

[0007] 3. The number of measured graphics is too small, only tens to hundreds, which cannot represent the billions of graphics in the real circuit

[0008] 4. Most of them still use optical measurement, and the accuracy is relatively poor

Images