Just-in-time adaptive disturbance estimation for run-to-run control in semiconductor processes
As run-to-run control has become more widely used throughout the semiconductor industry, it has become apparent that some of the unique characteristics of discrete parts manufacturing are driving the need for enhanced algorithm development. One such trait is the high mix of products made in a single factory (such as an ASIC fab or foundry). Variations in product quality often are functions of the product being produced as well as the manufacturing tools being used. Therefore, control systems are often designed to use only feedback data from lots that are the same product and have experienced the same upstream process flow as the lot currently being processed. In high-mix fabs with many products, some of the feedback loops may operate with hours or even days between data points in the feedback loop. This long delay results in a loss of information about the process tool’s contribution to the variance in that specific product. Solving the problem of product quality dependence on manufacturing context for photolithography overlay is much more difficult as there are many more factors which may contribute to variation. Because overlay is a relative measurement between two processing layers, overlay control systems generally segregate the control loops not only by tool and product, but also by substrate characteristics (reference layer tool, reticle, etc.). It can be easily imagined how a high-volume, high-mix fab operating in a mix-and-match lithography mode can generate hundreds of combinations of manufacturing contexts that result in a scarcity of feedback information for any given lot. Just-in-time Adaptive Disturbance Estimation (JADE) uses recursive weighted least-squares parameter estimation to identify the contributions to variation that are dependent upon manufacturing context such as tool, product, reference tool, and reference reticle. In using JADE, the run-to-run controller may use all available feedback data and does not ignore the fact that many lots may have been run since the last time a particular product was processed. The application of JADE, as well as traditional control techniques, is demonstrated on lithography overlay and chemical mechanical planarization (CMP) data taken from high-mix production facilities. The strengths and weaknesses of the JADEalgorithm are demonstrated on a series of test cases developed to separate the various disturbances and processing issues a control system would be expected to encounter. Although JADEhas been applied to an overlay process and a CMP process, this estimation technique should be applicable to any process area in the fab.