纳米级集成电路计算光刻技术研究

发布时间：2018-04-16 10:18

本文选题：分辨率增强技术 + 光学邻近校正　；参考：《浙江大学》2015年博士论文

【摘要】：随着芯片的集成度越来越高,特征尺寸越来越小,集成电路已经进入纳米级时代。这必将对集成电路的制造技术提出更高的要求。由于光源技术发展的滞后,193纳米波长光刻仍然是纳米级集成电路制造的主要选择。当集成电路发展到90纳米及以下工艺节点时,使用193纳米波长的光源来生产集成电路,会存在严重的光学邻近效应。业界提出了大量的分辨率增强技术(Resolution Enhancement Techniques, RETs)来弥补光学邻近效应所带来的版图失真,如离轴照明技术、光学邻近校正技术、移相掩模技术以及双重图形技术等。但是随着集成电路工艺节点发展到45nm以下,传统的分辨率增强技术遇到了很大的挑战。计算光刻技术(Computational lithography, CL),作为一种全新的分辨率增强技术,成为22nm以下技术节点光刻工艺的解决方案之一。目前较为主流的计算光刻算法大致分为两类,基于梯度法的计算法光刻算法和基于水平集的计算光刻算法。由于基于水平集的计算光刻算法目前为业界和学术界广泛研究,本文主要围绕基于水平集的计算光刻技术展开以下几个方面的研究工作：规则化水平集计算光刻算法。计算光刻技术又被称为基于点的光学邻近校正技术,其对版图的修改拥有更大的灵活性,可以获得更好的校正结果。但是计算光刻技术得到的掩模图案一般都过于复杂。这些复杂的图形给掩模板的生产制造带来了巨大挑战。针对这一问题,本文提出了一种规则化的水平集反向光刻算法。通过TV和拉普拉斯算子的引入,该算法在优化掩模形状的同时,较好地抑制了不规则图形的产生,其将掩模的复杂度平均降低了近40%,提高了掩模板的可制造性。用于增强工艺鲁棒性的水平集计算光刻算法。在受制造工艺参数变化干扰的条件下,掩模图形还需要有较高的图形保真度。本文提出了一种用于增强工艺鲁棒性的水平集计算光刻算法,以解决上述问题。新算法,通过工艺变化带的目标函数的引入,在优化掩模图形的同时对制程变化可能带来的影响也给予充分地考虑。这样在制造过程中,无论是成像平面的偏移,还是曝光能量的微变,在非标准光刻工艺条件下,掩模优化结果都具有较好的图形保真度。同普通的水平集计算光刻算法相比,该算法将制程制造指数(Process Manufacturability Index,PMI)平均减小了41.37%。基于混合共轭梯度的水平集计算光刻算法。大多数计算光刻算法都非常耗时,在优化过程中的收敛速度都比较慢。为了更好的解决这一问题,我们提出了一种基于混合共轭梯度的水平集计算光刻算法。该方法较好的克服了FR方法和PRP方法在应用中的不足。同传统的最速下降法相比,新的算法将掩模优化时间平均减少了36.5%,同时得到的掩模结果还具有较好的图形保真度,以及较强的制程鲁棒性。
[Abstract]:With the increasing integration of chips and smaller feature size, integrated circuits have entered the nanoscale era.This is bound to put forward higher requirements for IC manufacturing technology.Because of the development of light source technology, the lagged 193 nm wavelength lithography is still the main choice in the manufacture of nanoscale integrated circuits.When the integrated circuit is developed to 90 nm and below the process node, the optical proximity effect will be serious when the light source of 193 nm wavelength is used to produce the integrated circuit.A large number of resolution Enhancement techniques (rets) have been proposed to compensate for the distortion caused by optical proximity, such as off-axis illumination, optical proximity correction, phase shift mask and dual graphics.However, with the development of IC process node below 45nm, the traditional resolution enhancement technology has met a great challenge.Computational lithography, as a new resolution enhancement technology, has become one of the solutions of 22nm technology node lithography.At present, the popular computational lithography algorithms can be divided into two categories, one is based on gradient method and the other is based on horizontal set.Because the level set based computing lithography algorithm is widely studied by the industry and academic circles, this paper mainly focuses on the level set based computing lithography technology in the following aspects: regularized level set calculation lithography algorithm.Computational photolithography is also known as point-based optical proximity correction technology, which has more flexibility to modify the layout and can obtain better correction results.However, the mask patterns obtained by calculating photolithography are generally too complicated.These complex graphics pose a great challenge to the manufacture of mask templates.In order to solve this problem, a regularized level set reverse lithography algorithm is proposed in this paper.With the introduction of TV and Laplace operator, the algorithm not only optimizes the mask shape, but also restrains the production of irregular graphics. It reduces the complexity of mask by nearly 40 percent on average, and improves the manufacturability of mask.Level set calculation lithography algorithm for enhancing process robustness.Under the interference of manufacturing process parameters, the mask graphics also need to have a high graphic fidelity.In this paper, a level set calculation lithography algorithm for enhancing process robustness is proposed to solve the above problems.The new algorithm, by introducing the objective function of the process change band, optimizes the mask pattern and gives full consideration to the possible influence of the process change.In the process of fabrication, both the offset of the imaging plane and the micro-variation of the exposure energy, the optimized results of mask have good graphic fidelity under the condition of non-standard lithography process.Compared with the ordinary level set calculation lithography algorithm, this algorithm reduces the process Manufacturability index (PMI) by 41.37 on average.A level set calculation lithography algorithm based on mixed conjugate gradient.Most computational photolithography algorithms are time-consuming and slow to converge in the optimization process.In order to solve this problem better, we propose a level set calculation lithography algorithm based on mixed conjugate gradient.This method overcomes the deficiency of FR method and PRP method in application.Compared with the traditional steepest descent method, the new algorithm reduces the average mask optimization time by 36.5, and the obtained mask results also have good graphic fidelity and strong process robustness.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TN305.7

【相似文献】