Use of the response surface method in IC manufacturing

Process, device and circuit simulation is in wide and general use and its impact on process/circuit developments is certainly very high. Due to the increasing complexity of the process architectures and, moreover, to the increasing need of reducing the development cost and time, new methods and new tools are now necessary to successfully carry out optimization, statistical analysis and design centering tasks. For both process and circuit design the problems to be faced are generally characterized by a high number of input variables and by a high number of constraints/targets to be simultaneously satisfied/achieved. Moreover, the design sensitivity to process perturbations has to be minimized and taken into account in order to maximize the circuit parametric yield. In order to efficiently manage problems of such a complexity. methods based on experimental design [l-3], and regression analysis [4] have been proposed (see for example [S-lo]). The central point of these methods is the minimization