A system for auto defect screening using adaptive machine learning includes an adaptive model controller, a defect/nuisance library and a module for executing data modeling analytics. The adaptive model controller has a feed-forward path for receiving a plurality of defect candidates in wafer inspection, and a feedback path for receiving defects of interest already screened by one or more existing defect screening models after wafer inspection. The adaptive model controller selects data samples from the received data, interfaces with scanning electron microscope (SEM) review/inspection to acquire corresponding SEM results that validate if each data sample is a real defect or nuisance, and compiles model training and validation data. The module of executing data modeling analytics is adaptively controlled by the adaptive model controller to generate and validate one or more updated defect screening models using the model training and validation data according to a target specification.

Upper and lower limits of predetermined characteristic values of products contained in a plurality of product lots are stored in accordance with the product standard for a target product. An average value of standard deviations in the characteristic values is calculated based on a control chart for the product lots. An average value of the characteristic values is calculated, and an upper limit and a lower limit of an average value of the characteristic values in a 95% confidence interval is calculated. A measurement standard deviation representing a variation in a measuring instrument with regard to the characteristic values is estimated. One of an upper limit and a lower limit of the average value of the characteristic values in the confidence interval is updated as an average value of the characteristic values. A standard deviation in the characteristic values of the product is estimated, and an upper defect rate and a lower defect rate are calculated, so that a yield rate is calculated.

A system for auto defect screening using adaptive machine learning includes an adaptive model controller, a defect/nuisance library and a module for executing data modeling analytics. The adaptive model controller has a feed-forward path for receiving a plurality of defect candidates in wafer inspection, and a feedback path for receiving defects of interest already screened by one or more existing defect screening models after wafer inspection. The adaptive model controller selects data samples from the received data, interfaces with scanning electron microscope (SEM) review/inspection to acquire corresponding SEM results that validate if each data sample is a real defect or nuisance, and compiles model training and validation data. The module of executing data modeling analytics is adaptively controlled by the adaptive model controller to generate and validate one or more updated defect screening models using the model training and validation data according to a target specification.

Upper and lower limits of predetermined characteristic values of products contained in a plurality of product lots are stored in accordance with the product standard for a target product. An average value of standard deviations in the characteristic values is calculated based on a control chart for the product lots. Using data acquired by measuring the samples, an average value of the characteristic values is calculated, and an upper limit and a lower limit of an average value of the characteristic values in a 95% confidence interval is calculated. A measurement standard deviation representing a variation in a measuring instrument with regard to the characteristic values is estimated, and one of an upper limit and a lower limit of the average value of the characteristic values in the confidence interval, which corresponds to either one of the input upper and lower limits that is closer to the calculated average value of the characteristic values, is updated as an average value of the characteristic values. A standard deviation in the characteristic values of the product is estimated, and an upper defect rate and a lower defect rate are calculated, so that a yield rate is calculated.