等离子体增强化学气相沉积氮化硅薄膜制造过程质量控制方法研究

发布时间：2018-06-02 07:41

  本文选题：PECVD + 氮化硅薄膜　； 参考：《上海交通大学》2015年博士论文

【摘要】：等离子体增强化学气相沉积(Plasma Enhanced Chemical Vapor Deposition,PECVD)的氮化硅(Si N x)薄膜(以下简称PECVD氮化硅薄膜)是提高太阳能电池光电转换效率的关键,研究PECVD氮化硅薄膜制造过程的质量控制技术能够提高产品的成品率、合格率、优质率,促进太阳能电池相关行业的快速发展,对我国太阳能电池行业有非常重要的理论和现实意义。在国内外已有的研究成果的基础上,本文围绕着PECVD氮化硅薄膜制造过程质量控制提出工艺参数具有相关性时的故障诊断、制造初期的质量控制和两种失控模式下的质量控制三个的科学问题。基于已有的氮化硅薄膜沉积技术、正交试验设计、贝叶斯理论、多元统计技术、多元诊断、计算机仿真技术等研究了PECVD氮化硅薄膜的工艺参数快速优化、基于路径图的制造过程故障诊断、基于贝叶斯理论制造过程初期的质量控制、两种失控模式下多元统计过程质量控制四个关键技术,建立了PECVD氮化硅薄膜的工艺参数快速优化模型,基于路径图的多元故障诊断模型,基于贝叶斯理论的制造过程初期质量控制模型,基于多种总体协方差矩阵估值构建的多元指数加权移动平均(Multivariate Exponentially Weighted Moving-Average,MEWMA)控制图的质量控制模型。具体研究内容从以下几个方面展开:(1)分析了PECVD氮化硅薄膜的制造过程,论述了氮化硅薄膜的关键质量特征以及影响关键质量特征的关键影响因素。阐述了PECVD氮化硅薄膜制造过程质量控制的特点,分析了基于统计过程控制的质量控制方法应用于PECVD氮化硅薄膜制造过程时存在的问题,提出了PECVD氮化硅薄膜制造过程质量控制的科学问题,提出了与PECVD氮化硅薄膜制造过程质量控制方法相关的四个关键技术,提出了PECVD氮化硅薄膜质量控制模型。(2)通过领域知识和专家意见获取氮化硅薄膜的主要影响因子和质量特性参数,利用单因素物理实验确定了各主要影响因子和质量特征值之间的关系,提出了沉积参数快速优化的一般过程,给出了基于“综合评分法”多响应优化过程的综合评分计算方法和评分规则,分析得到最佳的沉积工艺参数,从而减少了大量的实际物理试验,提出了PECVD氮化硅薄膜制造过程的工艺参数快速优化方法。(3)利用领域内专家的经验构建氮化硅薄膜故障诊断路径图,研究了基于路径图的MYT分解方法,将传统的MYT方法中所有!种可能分解方式归结成一种统一的含有个独立因式的分解方式,并在过程均值向量与协方差矩阵未知的情况下利用F分布确定各分解项的控制上限。通过已确定的各分解项的控制上限,确定各个变量对失控信号的贡献值,并且用它来确定导致制造过程失控的根本原因,实现PECVD氮化硅薄膜多元统计过程诊断,解决了在PECVD氮化硅薄膜制造过程工艺参数具有相关性及过程均值向量与协方差矩阵未知的情况下的故障诊断问题,提出了基于路径图的PECVD氮化硅薄膜制造过程故障诊断方法。(4)针对在PECVD氮化硅薄膜制造过程初期具有多批次小批量的制造特征,经常面临样本容量不足的问题,提出了基于贝叶斯理论的PECVD氮化硅薄膜制造初期的质量控制方法。利用ANOVA方差检验和Bartlett齐次检验的方法从历史批次中筛选出与当前批次类似的样本数据,利用贝叶斯理论分析历史数据提供的先验信息,引入EWMA系数来调节历史信息在总信息中所占的比重,将总体均值和总体方差的估值表示为历史批次和当前批次的总体均值和总体方差的加权平均,利用这些估值建立控制图和计算过程能力指数,构建PECVD氮化硅薄膜的制造初期的质量控制方法。(5)针对PECVD氮化硅薄膜制造过程的多输入多输出、两种失控模式的特点,提出了基于总体协方差矩阵不同估值MEWMA控制图的PECVD氮化硅薄膜制造过程的质量控制方法。Hotelling2和MEWMA控制图使用的前提都是要求总体协方差矩阵已知,但是在实际制造过程中总体协方差矩阵通常都是未知的,需要使用样本协方差矩阵作为估计值,基于不同的估计值构建控制图的监控性能有着重大的不同。本文在相关总体协方差估值研究成果的基础上提出五种总体协方差估值和MEWMA统计量的数学表达式;其次,对基于总体协方差的五种估值构建的MEWMA控制图进行了仿真研究,寻找在在均值向量发生阶跃偏移和单向持续偏移的条件下监控效果最优的MEWMA控制图;构建两种失控模式下PECVD氮化硅薄膜制造过程的质量控制方法。本文以问题为导向,企业的实际需求为依托,以统计学为理论根据,以正交试验设计、计算机仿真、物理试验为手段,研究了PECVD氮化硅薄膜制造过程质量控制方法,为氮化硅薄膜的质量控制建立了新的方法论,并结合企业实际将文中所提的技术方法应用到PECVD氮化硅薄膜制造过程质量控制当中,取得了较好的经济价值。
[Abstract]:The plasma enhanced chemical vapor deposition (Plasma Enhanced Chemical Vapor Deposition, PECVD) silicon nitride (Si N x) thin film (hereinafter referred to as PECVD silicon nitride thin film) is the key to improve the photoelectric conversion efficiency of solar cells. The study of the quality control technology of the manufacturing process of PECVD silicon nitride film can improve the product yield and the rate of qualification. The high quality rate and the rapid development of the solar cell related industries have very important theoretical and practical significance for the solar cell industry in China. Based on the existing research results at home and abroad, this paper focuses on the fault diagnosis when the process parameters of the PECVD silicon nitride thin film manufacturing process are related, and the early manufacturing period. The scientific problems of quality control and three quality control under two kinds of out of control mode. Based on the existing silicon nitride film deposition technology, orthogonal test design, Bayes theory, multivariate statistical technology, multivariate diagnosis, computer simulation technology, the rapid optimization of PECVD silicon nitride thin film is studied, based on the path map manufacturing process. Fault diagnosis is based on the quality control of the initial manufacturing process of Bayesian theory and the four key technologies of multivariate statistical process quality control under two out of control modes. The rapid optimization model of the process parameters of PECVD silicon nitride film is established, the multiple fault diagnosis model based on the path graph, and the initial quality control of the manufacturing process based on the Bayesian theory. The model is based on the multi index weighted moving average (Multivariate Exponentially Weighted Moving-Average, MEWMA) control model based on a variety of overall covariance matrix valuations. The specific research contents are carried out from the following aspects: (1) the manufacturing process of PECVD nitrogen silicon thin film is analyzed, and the correlation of silicon nitride film is discussed. Key quality characteristics and key influencing factors affecting key quality characteristics are discussed. The quality control characteristics of PECVD silicon nitride thin film manufacturing process are described. The problems existing in the application of quality control method based on statistical process control to the manufacturing process of PECVD silicon nitride thin film are analyzed. The quality control of the manufacturing process of PECVD silicon nitride thin film is proposed. Four key technologies related to the quality control method of PECVD silicon nitride thin film manufacturing process are put forward. The quality control model of PECVD silicon nitride film is proposed. (2) the main influence factors and quality characteristic parameters of the silicon nitride film are obtained by field knowledge and expert opinion. The main factors are determined by single factor physical experiment. The relationship between the influence factor and the quality characteristic value is put forward, and the general process of rapid optimization of the deposition parameters is put forward. The comprehensive scoring method and scoring rule based on the multi response optimization process based on the "comprehensive scoring method" are given. The optimum parameters of the deposition process are analyzed, and a large number of actual physical tests are reduced, and PECVD nitriding is proposed. The rapid optimization method for process parameters of silicon film manufacturing process. (3) using the experience of experts in the field to construct the fault diagnosis path map of the silicon nitride film, and study the MYT decomposition method based on the path graph, and sum up all the possible decomposition methods in the traditional MYT method into one independent factorization method, and in the process of the process. In the case of the unknown mean vector and covariance matrix, the F distribution is used to determine the upper limit of the control of each decomposition item. The contribution value of each variable to the runaway signal is determined through the upper control upper limit of the determined Decomposition terms, and it is used to determine the root cause of the loss of control in the manufacturing process, and to diagnose the multivariate statistical process of the PECVD silicon nitride film, The problem of fault diagnosis in the process of PECVD silicon nitride film manufacturing process is solved. A fault diagnosis method for PECVD silicon nitride thin film manufacturing process based on path graph is proposed. (4) there are many batches of small batch in the initial stage of the manufacturing process of PECVD silicon nitride thin film. The quality of the quantity is often faced with the problem of the shortage of sample size. The quality control method of the PECVD silicon nitride film based on Bayesian theory is proposed. The sample data similar to the current batch are screened from the historical batch using the ANOVA variance test and the Bartlett homogeneous test, and the history is analyzed by the Bayesian theory. The prior information provided by the data is introduced, and the EWMA coefficient is introduced to regulate the proportion of historical information in the total information. The estimation of the overall mean and the overall variance is expressed as the weighted average of the overall mean and the overall variance of the historical batch and the current batch, and the control charts and the calculation process capacity index are established by these valuations, and the PECVD silicon nitride is constructed. The quality control method of the early manufacturing of the film. (5) in view of the characteristics of the multiple input multiple output and two out of control modes in the PECVD silicon nitride film manufacturing process, the precondition of the quality control method.Hotelling2 and the MEWMA control chart based on the PECVD silicon nitride film manufacturing process based on the MEWMA control diagram of the overall covariance matrix is proposed. It is required that the overall covariance matrix is known, but the overall covariance matrix in the actual manufacturing process is usually unknown, and the sample covariance matrix should be used as an estimated value. The monitoring performance of the control chart based on the different estimated values is very different. This paper is based on the results of the Research on the overall covariance estimation. Five mathematical expressions of the total covariance estimations and MEWMA statistics are given. Secondly, the MEWMA control graph based on the five estimations of the total covariance is simulated to find the best MEWMA control chart under the condition of the step shift and the one-way continuous migration of the mean vector; and the construction of the two out of control mode PECV The quality control method of D silicon nitride thin film manufacturing process is based on the problem oriented, the actual demand of the enterprise, based on the statistics theory, the method of orthogonal test design, computer simulation and physical test as the means to study the quality control method of the PECVD silicon nitride thin film manufacturing process, which has established the new quality control of the silicon nitride film. The method is applied to the quality control of the PECVD silicon nitride film manufacturing process with the practice of the enterprise, and it has achieved good economic value.
【学位授予单位】：上海交通大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TB383.2;TM914.4
，

本文编号：1967969