Surface Reconstruction for the Determination of Nanotopography in Makyoh

The integration calculation method that was first introduced for Makyoh data evaluation is proven to have high error and bad performance on noisy data. It ws shown that especially for the reconstruction of the Nanotophography on semiconductor wafer surfaces, it exhibits faulty data. New algorithms were introduced based on Fourier and Cosine transforms, they are Frankot-Chellappa, Poisson Solver, α-surface, M-estimator, Regularization and Diffusion.

Frankot-Chellappa is favored, since it has only less than 20 lines of program code, it is simple and fast. It is based on the Fourier Transform. hence, the reconstructed image filters out of the larger spatial wavelenght regions. It is also shown that it is not robust against noise and cannot preserve the edge.

Poisson is an alternative option and proves to be very good in reconstructing the syntheti topographies. It is also robust against noise. For real nanotophography measurement data, this algorithm show good calculation results in peak-to-valley variation profile.

M-estimator is almost similar with Poisson solver. It is robust against noise and shows good calculation results. But the compuational process is iterative and complicated. Regularization and Diffusion is not robust against noise but work very well in reconstructing smooth surface. α-surface reconstructs noisy image very well but the deviation from the ground truth is large.

Foir the validity of the reconstructed Makyoh data, further attempts need to be done on performing reference measurements. That is the nanotopography profile reconstructed by algorithms introduced in this thesis needs to be compared with results obtained from other measurment methods, i.e. by Stylus Profiler or Interferometer.