Feedforward control for reduced run-to-run variation in microelectronics manufacturing.

Abstract

Increased manufacturing yields can be obtained by reducing process variation. One potential method to achieve lower process variance is through inter-process feedforward control. During feedforward control, a process recipe is adjusted to compensate for measured input deviations. The potential benefits of feedforward control include reduced run-to-run variance, rework, and scrap. However, there are two problematic issues associated with feedforward recipe adjustment: (1) there is noise in the measurement tool and adjusting for inaccurate measurements could increase the variance; and (2) it is difficult to alter one parameter in a manufacturing process without worsening other key parameters. In this paper, we will address both issues using a systems approach. Measurement noise poses a significant threat to the success of feedforward control. If the measurement noise is large enough, the variance under feedforward control could exceed the variance with no control. To address this concern, we have integrated statistics theory into the feedforward controller design. This detunes the recipe adjustment based on the confidence in the accuracy of the sensor. These algorithms have the effect of filtering out the noise from the measurement tool. In order to address the problem of altering one parameter without adversely affecting others, we have developed a feedforward controller that selects a recipe from within a pre-defined set of allowable, qualified recipes. We call this Feedforward Recipe Selection Control (FRSC). We have implemented preliminary versions of this algorithm into a GUI-based CAD environment. This interactive software package guides the engineer through the design of feedforward controllers using process data as inputs. The vehicle for demonstrating these algorithms is critical dimension (CD) control in semiconductor manufacturing. Photoresist CD deviations from lithography will be compensated by etch bias adjustment at the Reactive Ion Etcher (RIE). An interesting aspect of CD manufacturing is that when a variance reduction is realized, the process is re-centered in order to gain higher profits associated with the faster chips. These feedforward control algorithms have been tested on two sets of industrial gate-etch data. Simple etch models are used to simulate feedforward recipe adjustment control from lithography to RIE. The results showed significant reduction in post-etch CD variance.