Accurate and Efficient Proximity Effect Correction for Electron Beam Lithography Based on Distributed Parallel Computing

Haojie Zhao,Wenze Yao,Hongcheng Xu,Siyuan Zhang,Yujie Yang,Xin Zhang,Jie Liu

DOI: https://doi.org/10.1109/iwaps57146.2022.9972245

2022-01-01

Abstract:This paper proposes an efficient proximity effect correction (PEC) method for electron beam lithography $(E B L)$ based on distributed parallel computing. To facilitate PEC calculations of large-scale layout, this method splits the exposure layout into multiple sub-layouts, and distributes them to different computer nodes for calculation through the message passing interface (MPI) technology. Under the premise of ensuring accuracy of PEC of the same layout, compared to the calculation time of 1 computing node (56 CPU cores per node), the acceleration ratios using $N_{\text {node }}=2,4$ and 6 nodes are $N_{\mathrm{ratio}}=2.28,5.29$, and 8.31, respectively. Given the same PEC calculation case with $N_{\text {pixel }}$ pixels, when $N_{\text {node }}$ is increased, the number of pixels calculated by one node $\left(n=N_{\text {pixel }} / N_{\text {node }}\right)$ decreases, leading to $N_{\text {ratio }} / N_{\text {node }}\gt1$ due to the $O(n \times \log (n))$ time complexity of the fast Fourier transform (FFT) in convolution computation. The proposed distributed parallelization scheme has been implemented in the second version of the $H N U-E B L$ software (http://www.ebeam.com.cn/).