Analysis and Optimization of Lithographic Performance on Massively-Parallel Electron-beam Systems

Abstract

While electron-beam (e-beam) lithography is widely used in transferring fine-feature patterns onto a substrate, its major drawback is the low throughput, especially for large-scale patterns. The massively-parallel electron-beam systems (MPES) were developed to increase the writing throughput and demonstrated to be able to write large-scale patterns significantly faster compared to conventional single-beam systems. However, the MPES comes with several constraints, such as abnormal beams (e.g., permanently on or off, spatial and temporal beam current fluctuation, and beam-positioning error), relatively large beam size, the realization of non-uniform dose distribution with a sub-beam-size resolution, and optimizing one performance metric affecting other metrics. To address these issues and maximize the efficiency of MPES, this study has several objectives. Firstly, the effects of abnormal beams on performance metrics are analyzed through simulation, comparing different writing methods, Single-row writing I, Single-row writing II, Multi-row writing, to suggest ways of reducing their negative effects. Secondly, the multi-row writing and multi-pass writing methods are compared in reducing the adverse effects of abnormal beams. Thirdly, a shape and dose control procedure with a sub-beam-size resolution for the proximity effect correction (PEC) is developed. Fourthly, two different methods for increasing beam utilization and reducing exposing time are investigated, i.e., lowering the dose difference among regions of a feature during the PEC and a new writing method for non-uniform distribution by adjusting the deflection angle of beams. Finally, an adaptive optimization method is designed that can handle any combination of performance metrics in a cost function for both a single feature and a large-scale pattern. Overall, this study aims at maximizing the efficiency of MPES by addressing various constraints, improving performance metrics individually, and optimizing multiple performance metrics simultaneously.