Automatic field sampling for cd measurement

Abstract

A method and apparatus is provided for inspecting a semiconductor wafer for field-to-field critical dimension (CD) variations using statistical techniques, such that an optimal number of fields on the wafer under inspection are measured, thereby increasing the accuracy of the results of the inspection procedure and avoiding unnecessary sampling. Embodiments include randomly selecting a predetermined number of fields on a semiconductor wafer to be inspected, and measuring the CD of a comparable feature in each of the sample fields, as by a critical dimension scanning electron microscope (CD-SEM). A statistical function, such as an average or standard deviation, of the measured CDs is calculated. Further fields are the randomly selected, CDs measured, and the running average or standard deviation calculated after each CD is measured. If the last acquired CD does not change the average or standard deviation by a predetermined amount, the inspection procedure for the wafer under inspection is terminated. Otherwise, the random field sampling, CD measurement and running average or standard deviation calculation continues. Because the number of fields to be sampled is not fixed, sampling continues as necessary without oversampling or undersampling, thereby obtaining statistically accurate results and increasing production throughput.