Manufacturing must provide products for customers that meet a certain quality standard. Even under the same process conditions, the basic characteristics and the influence on the process of each product could be different. This leads to a distribution...
Manufacturing must provide products for customers that meet a certain quality standard. Even under the same process conditions, the basic characteristics and the influence on the process of each product could be different. This leads to a distribution of characteristics in the final products, resulting in some degree of deviation. Therefore, to maintain a certain quality, it is necessary to manage process conditions and control deviations in product characteristics.
In this study, we analyzed the wafer limit control management method and machine learning. We analyzed and compared the correlations between semiconductor manufacturing processes such as etching, lithograph, etc. The effect of wafer limit management was reliable when there was a high correlation between processes. Through highly correlated processes, abnormal wafers could be detected using a new indicator based on standard deviation. Compared to abnormal wafer detected through machine learning, it showed the same level of detection. The abnormal wafers detected by the new indicator could undergo additional care items, such as full wafer measurement, process compensation, and full wafer screening. With each wafer limit management rather than process limit management, even on abnormal wafer could be detected early and, as a result, quality cost could be minimized.