Stability And Performance Analysis Of Semiconductor Manufacturing Process With Exponentially Weighted Moving Average Controllers

Semiconductor manufacturing industry has elevated cost in productions. Improvement of production efficiency is always an important goal for manufacturers. Because of the high capital costs associated with the process equipments, it is a common practice in today's semiconductor manufacturing to have many different products and processes run on each processing tool, i.e., high-mix manufacturing; However, in the area of semiconductor manufacturing process control, most of researches have been based on the assumption that there is only a single product in the manufacturing line, i.e., single product process, which is far from reality. Therefore, this thesis investigates the control methods for mixed product production. The main work for high-mix manufacturing can be summarized as follows:1. We study a model with two products manufactured on the same tool with the same production schedule, i.e., in each production cycle, the same kind of product will be manufactured for the same batches, and propose "product-based" double exponentially weighted moving average ("product-based" dEWMA) approaches into the system which undergoes the disturbance that follows the first order integrated moving average (IMA(1,1)) with deterministic drift, and find that if the break length of the product is large, even if the dEWMA controller is adopted, the output of the system is also deviated far from the specification.2. In order to improve the tool efficiency, in the real manufacturing processes, many different kinds of products are usually manufactured on the same tool. Therefore we study a more complicated case which assumes that a number of different kinds of products are manufactured on the same tool with variable manufacturing cycles, and the campaign length and break length of each cycle are also variable. We find that for mixed product drifted process, if the break length of a product is large, then at the beginning runs of each cycle, the process output will far deviate from the target value. We propose cycle resetting algorithm for discount factor of EWMA (CR-EWMA), and cycle forecasting EWMA (CF-EWMA) to reduce those large deviations; we also propose a discount factor resetting fault tolerant (RFT) approach and fault tolerant cycle forecasting EWMA (FTCF-EWMA) algorithm to handle the step fault, which is cause by the maintenance of the tool, of the system.Because in the semiconductor manufacturing batch processes, each step is a complicated physiochemical batch process, so generally it is difficult to perform measurements on-line or carry out the measurement for each run, and those combined with the fact that many process tools are not designed for the addition of in situ sensor, resulted in measurement taken less frequently than every run, or at stochastic runs. Therefore there will be delays in the feedback of the system. The effect of the delay on the stability of the system is an important issue which needs to be understood. Unfortunately, in the field of semiconductor manufacturing run-to-run control, few works are available for this issue, and therefore in this thesis, we investigate the stability of the system with metrology delay. The main works for delay system are concluded as follows:1. For the single product process, we propose two kinds of controllers, EWMA-â… and EWMA-â…¡controllers, and the stabilities of systems with both controllers which undergo different kinds of metrology delays are investigated. Necessary and sufficient conditions for the stochastic stability are established.2. For the mixed product process, EWMA-â… and EWMA-â…¡controllers are also proposed, and we extend the theorems of single product process to analyze mixed product process. The stability conditions for the mixed product process are also obtained.