智能优化技术在CMP铜抛光材料与工艺参数优化中的应用研究

发布时间：2018-08-10 19:46

【摘要】：随着GLSI的高度集成化和立体化,互连线的特征尺寸越来越小,对平坦化的要求越来越高,在多层铜布线的化学机械抛光(Chemical Mechanical Polishing,CMP)过程中,抛光液和抛光工艺是影响抛光后晶圆的表面质量和平坦化性能的两个关键因素。抛光液中各成分含量的比例不同,工艺不同,均会导致不同的抛光速率和平坦化效果。因此如何获得抛光液中各成分含量的比例和抛光工艺与抛光速率和平坦化效果之间的映射关系,并根据该映射关系得到满足工业要求的抛光液各组成成分的最佳配比和工艺条件,是目前研究的一个前沿课题。针对该问题,本课题采用我校微电子所自主研发的以FA/O型螯合剂为主要成分的多层铜布线碱性抛光液,通过计算机智能优化技术来分析优化该抛光液中各成分含量的比例和抛光工艺,从而提高抛光速率和平坦化性能。首先,作为一个跨学科的研究课题,在研究了以化学作用为主的CMP碱性技术的反应机理的基础上,通过对计算机智能优化技术中的误差反向传播(Back Propagation,BP)神经网络和人工蜂群(Artificial Bee Colony,ABC)算法的深入学习和研究,将BP神经网络和ABC算法两者相结合,针对现有建模方法BP-ABC算法预测能力不佳的问题,提出了改进后的建模方法:Back Propagation Niche Crowding Artificial Bee Colony算法,简称BP-NCABC算法。通过仿真实验证明BP-NCABC算法在建模能力和泛化能力方面均优于BP-ABC算法。其次,利用BP-NCABC算法对CMP实验数据进行仿真实验,建立抛光液各成分与抛光速率之间的关系模型,实现了对抛光速率的预测。同时利用该方法,针对300mm铜光片的实验数据建立了抛光液各成分和抛光工艺与抛光速率和片内非均匀性(With-In-Wafer-Non-Uniformity,WIWNU)之间的关系模型,实现了对抛光速率和WIWNU的同时预测,大大提高了在CMP工艺和材料的优化过程中的效率。最后,利用统计学中的敏感性分析方法对以上两个模型进行敏感性分析,使抛光工艺和抛光液各组成成分及其交互作用对抛光速率的影响程度和对WIWNU的影响程度得到了量化。根据量化分析结果,结合实际实验验证发现:在工艺条件为:压力1.2psi,流量300ml/min,抛光机转速为87rpm/min,抛光液各组成成分的配比为:磨料浓度为7vol.%,氧化剂浓度为0.5 vol.%,FA/O型螯合剂浓度为2.5 vol.%,活性剂浓度为3vol.%时,抛光速率为907.39nm/min,WIWNU为2.92%,图形片上抛光前后台阶高低差消除了大约3100?,抛光后表面状态良好,粗糙度为0.386nm,缺陷明显减少,实现了低压下的良好的平坦化性能,满足了工业发展对CMP技术的低压低磨料的要求。
[Abstract]:With the highly integrated and three-dimensional GLSI, the characteristic size of the interconnect becomes smaller and smaller, and the requirement of planarization is higher and higher. In the (Chemical Mechanical polishing process of multilayer copper wiring, Polishing fluid and polishing process are two key factors that affect the surface quality and flatness of polished wafer. Different proportion of each component in polishing liquid and different technology will result in different polishing rate and flattening effect. Therefore, how to obtain the mapping relationship between the content of each component in the polishing liquid and the polishing process, the polishing rate and the flattening effect, According to the mapping relation, the optimum proportion and process conditions of the components of polishing liquid which meet the industrial requirements are obtained, which is a forward research topic at present. In order to solve this problem, we use the FA/O type chelating agent as the main component of the multilayer copper wiring alkaline polishing liquid, which is developed by our institute of microelectronics. In order to improve the polishing rate and flatting performance, the proportion of each component in the polishing solution and the polishing process are analyzed and optimized by computer intelligent optimization technology. First of all, as an interdisciplinary research topic, the reaction mechanism of CMP alkalinity technology with chemical action is studied. Through the in-depth study and research on the Back propagation BP neural network and the artificial bee colony (Artificial Bee algorithm in the computer intelligent optimization technology, the BP neural network and the ABC algorithm are combined. In order to solve the problem of poor prediction ability of existing modeling methods, BP-ABC algorithm, an improved modeling method called BP-NCABC algorithm, is proposed in this paper. Simulation results show that BP-NCABC algorithm is superior to BP-ABC algorithm in modeling ability and generalization ability. Secondly, the BP-NCABC algorithm is used to simulate the CMP experiment data, and the relationship model between the polishing liquid components and the polishing rate is established, which realizes the prediction of the polishing rate. At the same time, based on the experimental data of 300mm copper wafer, the relationship model between polishing liquid composition and polishing process, polishing rate and in-chip non-uniformity (WIWNU) is established, and the simultaneous prediction of polishing rate and WIWNU is realized. The efficiency of CMP process and material optimization is greatly improved. Finally, the sensitivity analysis of the above two models is carried out by using the sensitivity analysis method in statistics, which quantifies the influence of polishing process and polishing liquid components and their interaction on polishing rate and WIWNU. According to the results of quantitative analysis, The experimental results show that when the process conditions are as follows: pressure 1.2psi, flow rate 300ml / min, polishing machine rotating speed 87rpm / min, the composition ratio of polishing liquid is: abrasive concentration is 7vol.slug, oxidant concentration is 0.5 vol.r / FAO type chelator concentration is 2.5 vol.and active agent concentration is 3vol.percent, The polishing rate is 907.39 nm / min WIWNU is 2.92, the difference between the steps before and after polishing on the graphics chip is eliminated, the surface condition is good, the roughness is 0.386 nm, the defect is reduced obviously, and the flatting performance is realized at low pressure. Meet the industrial development of CMP technology low-pressure abrasive requirements.
【学位授予单位】：河北工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TN305.2

【相似文献】