Model-Based Optical Metrology in R: M.o.R.

Reliable optical critical dimension (OCD) metrology in the regime where the
inspection wavelength \u03BB is much larger than the critical dimensions (CDs) of the
measurand is only possible using a model-based approach. Due to the complexity of the models
involved, that often require solving Maxwell's equations, many applications use a library
based look-up approach. Here, the best experiment-to-theory fit is found by comparing the
measurement data to a library consisting of pre-calculated simulations. One problem with this
approach is that it makes the accuracy of the solution dependent on the refinement of the
grid. Interpolating between library values requires a uniform grid in most cases, and can also
be very time-consuming. We present an approach based on radial basis functions that is fast,
accurate and most importantly works on arbitrary grids. The method is implemented in a
application based on the programming language R, that additionally allows for Bayesian data
analysis, and provides multiple diagnostics.