KR-20260067560-A - METHOD OF LEARNING TO ARTIFICIAL INTELLIGENCE FOR DETERMINE TYPES OF ARCS GENERATED IN THE SEMICONDUCTOR PLASMA PROCESS

KR20260067560AKR 20260067560 AKR20260067560 AKR 20260067560AKR-20260067560-A

Abstract

The present invention relates to an artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process, comprising: a data collection step for collecting data from a semiconductor plasma process; a first learning step for training a first model that extracts an arc signal from the data; a second learning step for training a second model that extracts arc features from the arc signal extracted by the first model; and a visualization step for reducing the arc feature vector extracted through the second learning step to a lower dimension and mapping the reduced low-dimensional arc feature vector to a spatial dimension to visualize the arc by feature. According to the present invention, artificial intelligence can identify various types of arcs that cause wafer defects and visualize the identified arcs by type. Due to the above effects, the user can easily identify the type of arc occurring, efficiently track the cause of the arc, and take appropriate measures regarding problems occurring in the semiconductor plasma process, thereby minimizing the occurrence of wafer defects.

Inventors

양원준

Assignees

주식회사 선인씨엔에스

Dates

Publication Date: 20260513
Application Date: 20241106

Claims (5)

A data collection step for collecting data from a semiconductor plasma process; A first learning step for training a first model that extracts an arc signal from the above data; A second learning step for training a second model that extracts arc features from the arc signal extracted from the first model; and An artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process, comprising a visualization step of reducing the arc feature vector extracted through the above second learning step to a lower dimension and mapping the reduced low-dimensional arc feature vector to a spatial dimension to visualize the arc by feature.
In claim 1, The above data collection step is, An artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process, characterized by classifying the above data into first data having only a periodic signal and second data having both a periodic signal and an arc signal.
In claim 2, The above-mentioned first model is, It is learned using an unsupervised learning method, and is learned to extract and restore the features of a periodic signal using the first data above. After the learning is completed, the first data and the second data are each received as input to generate third data having a periodic signal and fourth data having a periodic signal with the arc signal removed, respectively. An artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process, characterized by calculating the difference between the first data and the third data to extract a fifth data having a signal close to a zero vector, and calculating the difference between the second data and the fourth data to extract a sixth data having only an arc.
In claim 3, The above second model is, An artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process, characterized by being learned using a self-supervised learning method, and extracting an arc feature vector using the fifth data and the sixth data.
In claim 4, The above self-supervised learning method is, An artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process, characterized by being implemented using one or more of the Contrastive Learning technique and the BYOL (Bootstrap Your Own Latent) technique.

Description

Method of Learning to Artificial Intelligence for Determining Types of Arcs Generated in the Semiconductor Plasma Process The present invention relates to a technology for identifying arcs that cause wafer defects by using artificial intelligence to extract various types of arc characteristics that occur during a semiconductor process and visualizing the extracted arc characteristics to facilitate the classification of arcs by type. The semiconductor plasma process is a technology that carries out semiconductor processes by applying high current and high voltage within a vacuum chamber with various gas compositions. Since the semiconductor plasma process is a highly precise operation, irregular plasma shapes can occur due to various reasons, such as the incorporation of minute impurities, errors in components of the equipment used in the process, or aging; this leads to the generation of an arc, which is a light resembling lightning. The aforementioned arc causes various types of wafer defects, such as wafer perforation, uneven etching, and coating, thereby reducing process efficiency and productivity. Since defects caused by arcs are often discovered at the shipment stage, significant resource loss occurs during the semiconductor manufacturing process. Therefore, when such an arc occurs, technology is required to trace the cause of the arc and take appropriate measures according to the cause; however, currently, there is a complete lack of technology capable of classifying or identifying arcs by type. FIG. 1 illustrates a computing system for implementing an artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process according to an embodiment of the present invention and a method for visualizing the arc generated in a semiconductor plasma process by type. FIG. 2 is a flowchart of an artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process according to an embodiment of the present invention, and FIG. 3 illustrates various examples of first data that can be used in an artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process according to an embodiment of the present invention, and FIG. 4 illustrates an example of second data (an arc signal including a square wave periodic signal) that can be used in an artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process according to an embodiment of the present invention, and FIG. 5 illustrates first data (a) and second data (b) used in an artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process according to an embodiment of the present invention, and FIG. 6 illustrates third data (a) and fourth data (b) used in an artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process according to an embodiment of the present invention, and FIG. 7 illustrates the fifth data (a) and sixth data (b) used in an artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process according to an embodiment of the present invention, and FIG. 8 illustrates the process of the first learning step of an artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process according to an embodiment of the present invention, and FIG. 9 illustrates the process of the second learning step and the visualization step of an artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process according to an embodiment of the present invention, and FIG. 10 is an example of a visualization result according to the visualization step of an artificial intelligence learning method for determining the type of arc generated in a semiconductor plasma process according to an embodiment of the present invention. Hereinafter, some embodiments of the present invention will be described in detail with reference to exemplary drawings. It should be noted that in assigning reference numerals to the components of each drawing, the same components are given the same reference numeral whenever possible, even if they are shown in different drawings. Furthermore, in describing the embodiments of the present invention, if it is determined that a detailed description of related known configurations or functions would hinder understanding of the embodiments of the present invention, such detailed description is omitted. In addition, terms such as first, second, A, B, (a), (b), etc., may be used when describing the components of the embodiments of the present invention. These terms are used merely to distinguish the components from other components, and the essence, order, or sequence of the components is not limited by the terms used. Referri