基于像素表征的光刻掩模优化方法研究

发布时间：2018-06-25 15:30

本文选题：光学临近效应 + 分辨率增强技术　；参考：《华中科技大学》2015年博士论文

【摘要】：掩模优化技术是一种推进光刻分辨率极限,扩展现有光刻机服役寿命的重要分辨率增强技术。基于像素表征的掩模优化方法在实际应用中面临众多的技术难题,其中包括:(1)掩模可制造性难题,优化合成的掩模复杂度非常高,会极大的增加掩模的制造成本;(2)鲁棒性难题,掩模的光刻胶图形对生产工艺参数离焦和曝光剂量偏差的波动非常敏感,常规方法合成的掩模生产工艺窗口窄;(3)计算效率难题,一个先进的处理器芯片通常包含数十亿个晶体管,掩模图形非常密集复杂,优化的计算过程十分耗时。鉴于此,本学位论文从掩模可制造性、鲁棒性、和掩模优化计算效率三个方面开展研究,具体内容包括:提出了基于边缘距离误差的掩模成像质量评价标准,边缘距离误差具有“长度”的量纲,同时相对掩模具有解析的形式,通过它可以直观、高效的指导掩模优化。提出了基于滤波的掩模优化方法,该方法将掩模图形中的灰度、孤岛、小孔、锯齿、伸出项等复杂结构当成掩模图形噪声,然后根据掩模的可制造规则设计一个掩模滤波器,直接将滤波器集成到目标函数中的掩模上,从而来移除或阻止这些噪声在优化过程中产生。这种方法可以有效滤除掩模中细小、复杂的结构,并且保证了每次迭代合成的掩模都具有可制造性。分析了生产过程中工艺参数及其统计分布对掩模优化结果的影响。首先,推导了包含离焦和曝光剂量偏差参数的光刻系统正向模型表达式,接着建立了基于统计的鲁棒掩模优化模型,并将基于滤波的掩模可制造性增强方法推广到鲁棒的掩模优化模型中。通过仿真实验,验证了鲁棒优化方法合成的掩模对工艺参数变化具有鲁棒性,分析了离焦和曝光剂量偏差的统计分布对合成掩模图形的影响。开发了一整套的掩模优化数值求解算法。推导了掩模优化方向与迭代步长的计算公式,引入了PRP共轭梯度法和近似最优步长计算方法,提出了基于瀑布式多重网格算法的掩模优化流程。计算结果表明,相比常规计算方法,所提出的方法显著提高了计算速度,而且能够达到更小的目标函数值。本学位论文采用基准逻辑门电路结构图形对所提出的方法进行了仿真测试,并对比了目前主流的优化算法,包括罚函数方法和单一网格算法。结果表明,本学位论文所提出的方法能够优化合成一个复杂度低、成像质量好、工艺窗口宽的掩模图形,计算速度相比常规计算方法得到了显著提升。本学位论文系统的开发了一套基于像素表征的光刻掩模优化理论模型及方法,为实际应用中所面临的技术难题提供了有效的解决方案,论文所开展的研究工作丰富了现有的光刻掩模优化理论,为进一步推进光刻分辨率极限提供了一种高效的技术手段。
[Abstract]:Mask optimization is an important resolution enhancement technique to advance the resolution limit of lithography and extend the service life of existing lithography machine. The mask optimization method based on pixel representation is faced with many technical problems in practical application, including: (1) mask manufacturability problem, the complexity of optimization synthesis is very high, which will greatly increase the manufacturing cost of mask; (2) robustness problem. The photolithographic pattern of the mask is very sensitive to the fluctuation of the defocusing and exposure dose deviation of the production process parameters, and the window of the mask manufacturing process synthesized by the conventional method is narrow. (3) the calculation efficiency is difficult. An advanced processor chip usually contains billions of transistors, the mask graphics are very dense and complex, and the optimization process is time-consuming. In view of this, this dissertation studies the mask manufacturability, robustness, and mask optimization calculation efficiency. The main contents are as follows: the evaluation standard of mask imaging quality based on edge distance error is proposed. The edge distance error has the dimension of "length" and the analytic form of the mask mould. It can be used to direct the mask optimization directly and efficiently. In this paper, a filter based mask optimization method is proposed, in which the complex structures such as gray scale, isolated island, small hole, sawtooth, protruding term and so on are taken as the noise of the mask pattern, and then a mask filter is designed according to the manufacturing rules of the mask. The filter is directly integrated into the mask of the objective function to remove or prevent the noise from being generated in the optimization process. This method can effectively filter out the small and complex structure of the mask and ensure that the mask synthesized by each iteration is manufacturable. The effects of process parameters and their statistical distribution on the mask optimization results were analyzed. First, the forward model expression of lithography system including defocus and exposure dose deviation parameters is derived, and then the robust mask optimization model based on statistics is established. The filter based mask manufacturability enhancement method is extended to the robust mask optimization model. The simulation results show that the mask synthesized by the robust optimization method is robust to the change of process parameters. The influence of the statistical distribution of defocus and exposure dose deviation on the synthetic mask pattern is analyzed. A whole set of mask optimization algorithms are developed. The formulas of the optimum direction and the iterative step size of the mask are derived, the PRP conjugate gradient method and the approximate optimal step size method are introduced, and the mask optimization flow based on the waterfall multi-grid algorithm is proposed. The results show that compared with the conventional method, the proposed method can significantly improve the calculation speed and achieve a smaller value of the objective function. In this dissertation, the proposed method is simulated and tested by using the benchmark logic gate structure figure, and the current mainstream optimization algorithms, including penalty function method and single grid algorithm, are compared. The results show that the proposed method can optimize and synthesize a mask pattern with low complexity, good imaging quality and wide process window. In this dissertation, a theoretical model and method of lithographic mask optimization based on pixel representation is developed, which provides an effective solution to the technical problems in practical application. The research work in this paper enriches the existing photolithographic mask optimization theory and provides an efficient technical means to further advance the resolution limit of lithography.
【学位授予单位】：华中科技大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TN305.7

【相似文献】