Sparsity-driven reconstruction of molecular orbitals from angle-resolved photoemission spectroscopy

Recently, iterative phase retrieval has successfully been used to reconstruct spatial profiles of molecular orbitals measured with angle-resolved photoemission spectroscopy (ARPES). So far, phase retrieval algorithms for this technique, dubbed 'orbital imaging', have made use of the support constraint, in which case the spatial extent and shape of the orbital must be approximately known. In this article, we present a sparsity-driven approach to phase retrieval, which uses only the number of non-zero pixels in the orbital. No knowledge of the orbital shape is required, and the sparsity number can easily be determined from the measured data. Using the relaxed averaged alternating reflections algorithm, this sparsity constraint enables fast and reliable phase retrieval for noise-free as well as noisy simulated ARPES data. Furthermore, we extend the phase retrieval approach to include symmetry. We apply the resulting phase retrieval algorithms successfully to both simulated and experimental ARPES data. The experimental data are acquired using a so-called 'momentum microscope', which uses time-of-flight-based detection to measure the full momentum and energy-dependent spectrum in a single measurement. This enables us to image multiple molecular orbitals simultaneously. Finally, we discuss the validity of reconstructed orbitals depending on various aspects of the experiment.