随着集成电路特征尺寸进入2Xnm及以下节点，光源与掩模联合优化(SMO)成为了拓展193 nm ArF浸没式光刻工艺窗口、减小工艺因子的重要分辨率增强技术(RET)之一。提出了一种基于随机并行梯度速降(SPGD)算法的SMO方法，通过随机扰动进行梯度估计，利用估计梯度来迭代更新光源与掩模，避免了求解梯度解析表达式的过程，降低了优化复杂度。对周期接触孔阵列及十字线、密集线三种掩模图形的仿真验证表明，三种掩模图形误差（PE）值分别降低了75%、80%与70%，该方法较大程度地提高了光刻成像质量。

As the critical dimension in integrated circuit fabrication moving toward nodes-2Xnm and below, source and mask optimization (SMO) has been one of the most effective solutions of resolution enhancement techniques (RETs) to extend the process window and decrease process factor of 193 nm ArF lithography. We propose an efficient SMO method based on stochastic parallel gradient descent (SPGD) algorithm. The gradients of the objective function are estimated by random disturbance and utilized to guide the optimization, which avoids to calculate the analytic expression of the gradients. The complexity of optimization is reduced. The proposed SMO method is demonstrated using three mask patterns, including a periodic array of contact holes, a cross gate and dense lines. Three kinds of mask pattern error (PE) are reduced by 75%, 80% and 70% respectively. The numerical results show that our method can provide great improvements in printed image quality.