On an analysis approach for an improved linear overlay control

Overlay control is performed using linear static models of the exposure tool with the controlled variables being the parameters of the linear exposure tool model. The CIM interface of the exposure tool also typically only offers the parameters of those linear models as the manipulated input variables. Nevertheless, the real distribution of the overlay error over the wafer surface can be nonlinear (e.g. due to processes upstream to the exposure step). While, in the past, nonlinear residuals could be tolerated and were contained well within the wide overlay specification boundaries, with shrinking feature size and ever tightening overlay budget these nonlinear residuals will become more relevant in the future. An optimization approach capable of maximizing the number of overlay targets that will reside within some inner specification limits that are relevant to chip performance is described. The key benefit of this approach consists in providing a smaller overlay within the wafer centre region. As the wafer centre is also less vulnerable to other process problems induced by upstream (like CMP) or downstream (like etch) processes, this approach has the potential to yield more high performance dies even with the linear control methodology, exposure tools, and tool interfaces used throughout the fabs. It´s not necessary that the exposure tool or it´s CIM interface is able to handle nonlinear manipulated variables in order for this approach to be deployed.