In advanced semiconductor manufacturing, model-based optical proximity correction is commonly used to compensate for image errors. The final pattern is generated using correction values determined by lithography simulation. Image errors such as patterns with insufficient correction or patterns with excessive correction can be generated. These patterns with errors are called hotspots. Such errors are conventionally detected by lithography simulation of OPC patterns. When a hotspot is detected by lithography simulation, it has to be repaired manually or by repeated use of OPC tool. However, it is difficult to obtain correct pattern for a complicated shape, and the correction procedure may require a significant amount of additional processing. In order to solve this issue, we examine application of cellular automata (CA) method for hotspot correction. It is known that CA method can be used for weather or traffic analysis and prediction. In this report, we studied the CA method for deriving simple hotspot repair rule based on lattice cell-like models for light intensity distribution and OPC patterns. We will report on the results of hotspot correction technique with the OPC pattern using CA method.
We investigated the possibility of hotspot detection after lithography simulation by using Neural Networks (NN). We
applied the image recognition technique by the NN for hotspot detection and confirmed the possibility by its recognition
rate of the device pattern defects after NN learning.
Various test patterns were prepared for NN learning and we investigated the convergence and the learning time of the
NN. The compositions of the input and the hidden-layers of the NN do not have so much influence on the convergence
of NN, but the initial parameter values of weight setting have predominant effect on the convergence of the NN. There
are correlations among the learning time of the NN, the number of input samples and the number of hidden-layers, so a
certain consideration is required for NN design.
The hotspot recognition rate ranged from 90% to 42%, depending pattern type and learning sample number. Increasing
learning sample number improves the recognition rate. But learning all type patterns leads to 55% recognition, so
learning single type pattern leads to better recognition rate.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.