Wafer defect prediction device and operating method thereof
A method of predicting wafer defect information includes estimating a distribution with respect to defect occurrence time data, the defect occurrence time data including information about a time associated with a wafer defect occurrence, distinguishing a defect distribution type according to a result of the estimating the distribution, and outputting wafer defect information predicted according to the distinguished defect distribution type.
1 . A method of predicting wafer defect information, the method comprising:
estimating a distribution with respect to defect occurrence time data, the defect occurrence time data including information about a time associated with a wafer defect occurrence in a wafer;
classifying types of defects occurring in the wafer using the estimated distribution, wherein classifying the types of defects occurring in the wafer comprises:
fitting different portions of the estimated distribution using a plurality of distributions, wherein each of the plurality of distributions has a different gradient; and
for each portion of the estimated distribution, classifying the corresponding type of defect based on the gradient of the corresponding distribution;
outputting wafer defect information predicted according to the classified types of defects; and
fabricating a semiconductor device using the wafer or dispositioning the wafer depending on the wafer defect information.
2 . The method of claim 1 , wherein the estimating of the distribution comprises:
dividing the defect occurrence time data into k distributions based on a goodness of fit (GoF) between a distribution of the defect occurrence time data and a probability distribution; and
outputting the probability distribution corresponding to each of the k distributions,
wherein k is an integer greater than or equal to 1.
3 . The method of claim 1 , wherein the outputting of the wafer defect information comprises outputting wafer defect information predicted from at least one of a lifetime or an extrinsic defect rate.
4 . The method of claim 2 , wherein classifying the types of defects occurring in the wafer further comprises distinguishing the type of defect based on at least one of a shape of the k distributions or a k value.
5 . The method of claim 2 , further comprising:
predicting wafer defect information based on lifetime of a wafer with respect to a first defect distribution type, wherein k is 1.
6 . The method of claim 2 , wherein classifying the types of defects occurring in the wafer further comprises:
classifying the types of defects, in response to a case where k is greater than or equal to 2 by comparing a gradient of a first distribution with a gradient of a second distribution,
wherein the first distribution is based on defect occurrence time having at least one defect occurrence time earlier than at least one defect occurrence time of the second distribution.
7 . The method of claim 6 , further comprising:
predicting wafer defect information based on a lifetime of a wafer,
wherein the predicting the wafer defect information is with respect to a second defect distribution type in which a gradient of the first distribution is greater than a gradient of the second distribution.
8 . The method of claim 6 , further comprising:
predicting wafer defect information based on an extrinsic defect rate of a wafer,
wherein the predicting the wafer defect information is with respect to a third defect distribution type in which the gradient of the first distribution is less than the gradient of the second distribution.
9 . The method of claim 2 , wherein the probability distribution comprises a Weibull distribution.
10 . The method of claim 9 , wherein the gradient of a first distribution and the gradient of a second distribution correspond to a shape parameter of the Weibull distribution.
11 . A method of predicting wafer defect information, the method comprising:
estimating a distribution with respect to defect occurrence time data including information about a time associated with an occurrence of the wafer defect in a wafer;
classifying types of defects occurring in the wafer using the estimated distribution, wherein classifying types of defects comprises:
fitting different portions of the estimated distribution using a plurality of distributions, wherein each of the plurality of distributions has a different gradient; and
for each portion of the estimated distribution, classifying the corresponding type of defect based on the gradient of the corresponding distribution;
outputting the wafer defect information by using a wafer defect prediction model according to the classified types of defects; and
fabricating a semiconductor device using the wafer or dispositioning the wafer depending on the wafer defect information.
12 . The method of claim 11 , wherein the estimating of the distribution comprises:
dividing the defect occurrence time data into k distributions based on a goodness of fit (GoF) between a distribution of the defect occurrence time data and a probability distribution; and
outputting the probability distribution corresponding to each of the k distributions, wherein k is an integer greater than or equal to 1.
13 . The method of claim 11 , wherein the wafer defect prediction model is based on machine learning, wherein for each distribution, the machine learning is based on values of a shape parameter, a scale parameter, a weight, and a likelihood.
14 . The method of claim 11 , wherein the wafer defect prediction model predicts at least one of whether a defect occurs for each wafer or an extrinsic defect rate for each wafer.
15 . A device for predicting wafer defect information by using a processor, the device comprising:
processing circuitry configured to:
estimate a distribution with respect to defect occurrence time data including information about a time associated with a wafer defect occurrence in a wafer,
classify types of defects occurring in the wafer using the estimated distribution, wherein classifying types of defects comprises:
fitting different portions of the estimated distribution using a plurality of distributions, wherein each of the plurality of distributions has a different gradient, and
for each portion of the estimated distribution, classifying the corresponding type of defect based on the gradient of the corresponding distribution,
classify a grade of the wafer based on the classified types of defects, and
fabricate a semiconductor device using the wafer or dispositioning the wafer depending on the classified grade.
16 . The device of claim 15 , wherein the processing circuitry is further configured to divide the defect occurrence time data into k distributions based on a goodness of fit (GoF) between a distribution of the defect occurrence time data and a probability distribution, and to output the probability distribution corresponding to each of the k distributions,
wherein k is an integer greater than or equal to 1.
17 . The device of claim 15 , wherein the processing circuitry is further configured to output wafer defect information predicted from at least one of lifetime or an extrinsic defect rate.
18 . The device of claim 16 , wherein the processing circuitry is further configured to generate a wafer defect prediction model by updating parameters of a machine learning model based on training data, the training data having information about the k distributions included in the distribution estimation result as input variables, and having a wafer defect occurrence information and an extrinsic defect rate as output variables.
19 . The device of claim 18 , wherein the processing circuitry is further configured to classify the wafer to be a second grade wafer in response to the extrinsic defect rate of the wafer predicted through the machine learning model being less than or equal to a first defect rate.
20 . The device of claim 17 , wherein the processing circuitry is further configured to classify the wafer to be a first grade wafer in response to the lifetime of the wafer being greater than a first defect rate.