Wafer map analyzer, method for analyzing wafer map using the same and method for manufacturing semiconductor device

Abstract

A method for analyzing a wafer map using a wafer map analyzer includes generating first wafer maps each displaying characteristics of a first wafer for a corresponding channel of a plurality of channels. The first wafer maps are auto-encoded together to extract a first feature. The method also includes determining whether the first feature is a valid pattern, classifying the type of the first feature based on unsupervised learning when the first feature is a valid pattern and extracting a representative image of features classified into the same type as the first feature.

What is claimed is: 
     
         1 . A wafer map analyzer, comprising:
 a storage device which stores a wafer map of a semiconductor wafer; and   a processor connected to the storage device,   wherein the processor extracts features from the wafer map, determines validity of the features, clusters the features to classify the features into multiple classified types, generates a feature having a center value for each of the classified types, and reconstructs the feature into a wafer map to generate a representative image of the type, and   the storage device stores the representative image for each type.   
     
     
         2 . The wafer map analyzer of  claim 1 , wherein a plurality of wafer maps having different characteristics corresponds to a single wafer map. 
     
     
         3 . The wafer map analyzer of  claim 2 , wherein the processor simultaneously auto-encodes all the plurality of wafer maps corresponding to the single wafer to extract one feature for each wafer. 
     
     
         4 . The wafer map analyzer of  claim 1 , wherein the processor determines the validity, using a reconstruction error of the feature. 
     
     
         5 . The wafer map analyzer of  claim 4 , wherein the wafer map comprises a plurality of units having good values or bad values, and
 wherein the processor determines the validity considering the reconstruction error and a number of units having the bad values.   
     
     
         6 . The wafer map analyzer of  claim 1 , wherein the storage device assigns a code to the representative image for each type and stores the code. 
     
     
         7 . The wafer map analyzer of  claim 1 , wherein the processor clusters the features to classify the features into multiple classified types based on unsupervised learning. 
     
     
         8 . A wafer map analyzer, comprising:
 a non-volatile memory which stores a program and a wafer map of a semiconductor wafer;   a volatile memory to which the program is loaded;   a processor which executes the program; and   a bus which connects the processor, the non-volatile memory and the volatile memory,   wherein the program includes:   an auto encoder which automatically encodes a wafer map to extract a feature,   a feature filter which determines validity of the feature for clustering and excludes the feature when the feature is not valid,   a clustering machine which performs clustering of the feature into a group and generates a center feature of the group according to the clustering;   a code allocator which allocates a code to a representative image corresponding to the center feature and stores the code in the non-volatile memory.   
     
     
         9 . The wafer map analyzer of  claim 8 , wherein the auto encoder reconstructs the center feature to generate the representative image, and the code is searchable to identify the semiconductor wafer mapped by the wafer map. 
     
     
         10 . The wafer map analyzer of  claim 8 , further comprising:
 an anomaly pattern detector which determines whether an invalid feature is an anomaly pattern.   
     
     
         11 . The wafer map analyzer of  claim 10 , wherein the auto encoder reconstructs the anomaly pattern to generate a reconstructed wafer map anomaly pattern. 
     
     
         12 . The wafer map analyzer of  claim 11 , wherein the code allocator allocates a code to the reconstructed wafer map anomaly pattern, and stores the code in the non-volatile memory. 
     
     
         13 . A method for analyzing wafer maps of semiconductor wafers, the method comprising:
 forming a plurality of wafer maps for a plurality of wafers, respectively;   auto-encoding the plurality of wafer maps to extract a plurality of features corresponding to the plurality of wafers;   excluding invalid features not valid for classification among the plurality of features;   classifying valid features among the plurality of features into a plurality of types using unsupervised learning;   generating a plurality of center features corresponding to respective centers of the plurality of types; and   reconstructing the plurality of center features to output a representative image.   
     
     
         14 . The method of  claim 13 , wherein a plurality of wafer maps having different characteristics corresponds to a single wafer map. 
     
     
         15 . The method of  claim 14 , wherein extracting the plurality of features comprises:
 simultaneously auto-encoding all the plurality of wafer maps corresponding to the single wafer to extract one feature for each wafer.   
     
     
         16 . The method of  claim 13 , wherein excluding the invalid features among the plurality of features comprises:
 determining validity, using a reconstruction error of each feature.   
     
     
         17 . The method of  claim 13 , wherein the wafer map comprises a plurality of units having good values or bad values, and
 determining wherein excluding the invalid features among the plurality of features includes considering the reconstruction error and a number of units having the bad values.   
     
     
         18 . The method of  claim 13 , further comprising:
 determining whether the invalid feature among the plurality of features is an anomaly pattern and storing the invalid feature in a storage device.   
     
     
         19 . The method of  claim 18 , wherein determining whether the invalid feature is the anomaly pattern comprises:
 comparing an anomaly pattern sample stored in advance with an invalid feature among the plurality of features.   
     
     
         20 . The method of  claim 13 , further comprising:
 specifying a code of the representative image and storing the code in a storage device.