测量时延在线估计与批间控制器协同设计

Online estimation of time-varying metrology delay and run-to-run control co-design

中文关键词  批间控制  双指数加权滑动平均  时延  贝叶斯理论

英文关键词  run-to-run control (R2R)  double exponent weighted moving average (d–EWMA)  metrology delay  Bayesian theory

中文摘要

      批间控制是半导体批次生产过程中常用算法, 其关键问题在于能够及时获取上一批次的制程输出, 受测量 手段及其成本限制, 实际的生产制程很难满足这一要求. 为此, 本文提出一种基于贝叶斯统计分析的测量时延估计 算法. 在分析晶圆质量与实测时延、估计时延、以及制程漂移之间的逻辑关系的基础上, 并将晶圆的质量信息按加 工时间顺序划分两个相邻的滚动时间窗口. 基于贝叶斯后验概率函数, 及时捕获后一个滚动时间窗口内过程输出 发生漂移的概率, 从而判断是否有测量时延发生, 并估算该时延大小. 在此基础上, 给出批间控制器的测量时延补 偿策略, 及时调整制程的控制量, 提高晶圆的加工品质. 仿真结果验证所提出算法的有效性.

英文摘要

      Run-to-run (R2R) control has been widely used in the semiconductor manufacturing processes. The core of the algorithm is to obtain the process output at previous runs. Restricted by the measurement tool and cost, it is impossible to obtain the quality data timely. We propose an estimation algorithm for estimating the time-varying metrology delay by using the Bayesian statistical analysis. Firstly, we analyze the logical relationship among the wafer quality, the real measurement time-delay, the estimated measurement time-delay and the drift of the process. The wafer quality data is divided into two parts and put in two adjacent moving windows according to the wafer manufacturing sequence. Then, the occurrence probability of the time-varying metrology delay is determined based on the value of Bayesian posterior probability in the second moving window. The delay value is calculated by a trial and error method. On this basis, the delay compensation value for the R2R controller is determined, the control value of the manufacturing process is adjusted accordingly, and the quality of product is improved. The simulation example validates the efficacy of the proposed algorithm.