Highly sensitive inspection system for lithography-related faults in agile-fab - detecting algorithm for monitoring and evaluation of yield impact

摘要

The authors performed automatic detection of lithography-related faults by characteristic factors calculated from fail bit count (FBC) data. They used the autocorrelation function to define a characteristic factor for a certain lithography-related failure mode. The frequency of lithography-related faults was monitored as time series data. Also, components of the characteristic factor quantified the direction of the inclination of lithography-related patterns. They changed according to time. The origin of faults varies with patterns. Thus, it is presumed that their causes differed with each period. Fault sourcing for a failure mode was carried out by correlating machine data and the characteristic factor with Χ2 test. Since the characteristic factor included only one failure factor, this test could be performed with high sensitivity. The authors classified lithography-related faults and evaluated their yield impact from their frequency and yield loss automatically.