The purpose of statistical process control is the identification of abnormal variations in the materials, equipment, parameters or procedures used as inputs for particular processes (see Table 3.1). This is usually accomplished by measuring physical objects on the output, such as linewidths or overlay structures. When processes require test wafers, which often occurs during process development or the initiation of a new manufacturing process, the output of the process is decoupled from the input, and statistical process control cannot fulfill its primary purpose.
This situation is shown schematically in Fig. 3.1. The inputs collectively comprise the process. A single wafer is taken from a lot of wafers and processed through the lithography operation. After the processing is complete, this test wafer is measured for parameters of interest, such as linewidths or overlay. From the values of these measurements, in comparison to the process targets, the remaining wafers in the lot are processed through the lithography operation with adjusted process parameters. For example, the exposure dose might be adjusted to bring linewidths to the process target. Changes in the dose might be required to compensate for drift in the stepper's dose control system or the changes in the resist process. Measurements of the linewidths of the lot, except for the test wafer, will not reveal a drift in the stepper's dose control system or a change in the resist process, because the exposure dose has been adjusted to compensate for these instabilities. In order for statistical process control to reveal variations and instabilities in the inputs, it must be applied to a simple process, where the input variables are directly coupled to the measurable output. In this chapter a process control methodology applicable to situations in which test wafers are used is presented
Online access to SPIE eBooks is limited to subscribing institutions.