PROXIMITY EFFECT CORRECTION IN ELECTRON-BEAM LITHOGRAPHY - A HIERARCHICAL RULE-BASED SCHEME-PYRAMID

PYRAMID, a hierarchical, rule-based scheme for proximity effect correction in electron-beam lithography is proposed. The current implementation performs solely pattern modification, and uses a single dose for the entire circuit. Based on a digital image processing model of the physical lithographic process, a hierarchical correction procedure is employed in two parts, local correction and global correction. The local correction, fully implemented, is concerned with interactions between circuit components within a small window. The local correction itself is very systematic, using two levels of correction to minimize proximity effect caused by intrashape and intershape interactions, respectively. Rule tables are used to dictate correction modes for different situations. These tables, constructed a priori, are utilized to accelerate the correction process by minimizing the numerical calculation required during circuit correction. While the local correction ignores interactions between widely separated circuit elements, the global correction takes general characteristics of the entire circuit pattern into account to make adjustments to the local correction modes. This combination of local and global corrections permits the correction of arbitrarily sized circuit patterns in a small fraction of the time required by many previous approaches that are more computationally intensive. In addition, PYRAMID produces output circuit patterns that are decomposable into rectangles, allowing efficient representation of the circuit patterns, and maintaining compatibility with shaped electron-beam architectures. Furthermore, the hierarchical approach that has been employed permits us to develop and evaluate fast proximity correction schemes for new electron-beam architectures such as parallel beam lithography systems. Such system architectures may place severe constraints on the allowable dosage variation, e.g., constant dose, during exposure. Test patterns with a minimum feature size of 0.1-mu-m are presented along with corresponding correction times.