In situ beam drift detection using a two-dimensional electron-beam position monitoring system for multiple-electron-beam–direct-write lithography

• Published in: Journal of vacuum science & technology. B, Microelectronics and nanometer structures processing, measurement and phenomena Microelectronics and nanometer structures processing, measurement and phenomena, 2011-07, Vol.29 (4), p.041607-041607-12
• Main Authors: Chen, Sheng-Yung, Tsai, Kuen-Yu, Ng, Philip C. W., Ng, Hoi-Tou, Liu, Chun-Hung, Shen, Yu-Tian, Kuan, Chieh-Hsiung, Chen, Yung-Yaw, Kuo, Yi-Hung, Wu, Cheng-Ju, Yen, Jia-Yush
• Format: Article
• Language: English
• ISSN: 1071-1023; 2166-2746; 1520-8567; 2166-2754
• DOI: 10.1116/1.3613697

Electron-beam lithography is one of the promising candidates to replace optical projection lithography due to its high resolution and maskless direct-write capability. In order to achieve the throughput requirement for high-volume manufacturing, miniaturized electro-optics elements are utilized to drive massively parallel beams simultaneously. In high-throughput multiple-electron-beam systems, beam positioning drift problems can become quite serious due to several factors such as thermal distortion and fabrication errors of electron optics. In single-beam systems, periodic recalibration with reference markers on the wafer can be utilized to achieve beam placement accuracy. This technique is not easy for multiple-beam systems. In this article, an innovative in situ two-dimensional electron-beam position monitoring system for multiple-electron-beam lithography is studied. An array of miniaturized electron detectors to measure scattered electrons from the substrate is placed above the wafer. It is assumed that the detector array signals are correlated with the distribution of electron trajectories, and the change of trajectory distortion due to the beam drift can be predicted by Monte Carlo electron-scattering simulation. A standard quadrant detection (SQD) method and a linear least-squares (LLS) method are used to estimate the beam drift from the detector array signals. Simulation results indicate that while the estimation uncertainty of both methods can be reduced substantially when the number of detected electrons is large enough. The LLS method always outperforms the SQD one regardless the detected electron numbers.