基于人工神经网络权值优化的投影光刻机像质校正灵敏矩阵的计算方法  

Calculating Method of Image Quality Correction Sensitive Matrix of Lithographic Projection System Based on Weight-Optimizing of an Artificial Neural Network

作　　者：施伟杰[1] 王向朝[1] 张冬青[1] 王帆[1] 

机构地区：[1]中国科学院上海光学精密机械研究所信息光学实验室

出　　处：《中国激光》2006年第4期516-520,共5页Chinese Journal of Lasers

摘　　要：像质校正灵敏矩阵是像质校正算法中由像质参数计算像质校正参数的过渡矩阵。矩阵中的元素表征了像质参数随像质校正参数变化的灵敏程度。像质校正灵敏矩阵是投影光刻机像质校正算法中重要的参数集合。提出了一种投影光刻机像质校正灵敏矩阵的原位测量方法,该方法利用人工神经网络(ANN)对像质参数和像质校正参数之间的依赖关系进行建模,通过网络自学习能力优化网络的连接权值使其逐渐逼近像质校正灵敏矩阵的数值,从而实现像质校正灵敏矩阵的测量。实验结果表明该方法可以高精度、有效地获得投影光刻机像质校正灵敏矩阵。Image quality correction sensitive matrix is a transfer matrix, by which the image quality correction parameters can be calculated from the image quality parameters. The elements of the matrix describe the relations between the image quality correction parameters and the image quality parameters and the matrix is an importance parameter set during the image quality correction of the proiection lithographic tool. In this paper, a calculating method of the image quality correction sensitive matrix is presented. In this method, an artificial neural network （ANN） is introduced to model the relationship between the image quality parameters and the image quality correction parameters. With the training of the neural network, the connection weights of the network are optimized towards the image quality correction sensitive matrix. Experiment results show that the image quality correction sensitive matrix can be obtained in situ by this method with a high accuracy.

关 键 词：测量 人工神经网络 像质校正灵敏矩阵 投影光刻机 

分 类 号：TN305.7[电子电信—物理电子学]