Use of neural networks in modeling relations between exposure energy and pattern dimension in photolithography process [MOS ICs]

Abstract

The photolithography process is one of the most complex operations in semiconductor production. Exposure energy definition is particularly critical because it strongly affects the operation results. Very complex links exist between exposure energy, pattern critical dimensions, photo resist thickness, and resistivity. At present, the wafer test experimental procedure is used in order to define suitable exposure energy. With the aim of finding a less expensive control criterion of exposure operation in the photolithography process, a neural network has been developed that is able to model the relation between exposure energy and pattern dimensions measured in different positions on the wafer. As a result, the neural network model developed has been found to perform as well as the very expensive test wafer procedure and constitutes a good alternative to this one, allowing for a remarkable cost reduction.