CASUS Institute Seminar, Silvio Fuchs, Friedrich-Schiller-Universität Jena
In contrast to two- or three-dimensional phase reconstruction problems as it occurs in lensless imaging techniques like Coherent Diffraction Imaging (CDI) or ptychography, one-dimensional phase retrieval (PR) is by far the hardest to solve. Because of missing symmetries the solution is typically not unique. We developed a step-wise one-dimensional phase retrieval (PR) algorithm, which solves a special class of problems emerging in spectral interferometry measurements in the extreme ultraviolet range. The underlying technique is called eXtreme ultraviolet coherence tomography (XCT) and enables nanoscale depth-resolved imaging of e.g. semiconductor devices.
Recently, a laboratory-based experimental setup for XCT was built using XUV radiation from a laser-driven high harmonic source. In the XCT setup, the broadband reflected spectrum of the sample is directly recorded without a distinct reference wave. Therefore, the signal only contains the autocorrelation of the sample’s structure.
Our new algorithm succeeded in reconstructing the axial structure of silicon-based samples without any autocorrelation artifacts. It uses step-wise relaxing constraints together with GS and HIO approaches to ensure convergence.The algorithm is surprisingly stable and converges to the right physical solution. In contrast to other one-dimensional algorithms our version does not use sparsity constraints but still finds the right solution. The reason for that actually unexpected behavior shall be discussed. The reconstructed phase information can further be used to even extract material and other structural information like roughness or layer thicknesses.