Interpreting X-ray absorption spectra of Vanadyl Phthalocyanines Spin Qubit Candidates using a Machine Learning-Assisted Approach

Vanadyl Phthalocyanines (VOPc) molecules, with spin 1/2 and long coherence times up to room temperature, are promising candidates for molecular spin qubits[M. Atzori et al., J. Am. Chem. Soc. 138, 2154 (2016).]. To optimize their quantum coherence, it is crucial to investigate orbital splitting on surfaces and interactions between molecular spins and substrates[I. Cimatti et al., Nanoscale Horiz. 4, 1202 (2019).]. We employ X-ray absorption spectroscopy (XAS), machine-learning-assisted multiplet calculations, and Density Functional Theory (DFT) to explore the 3d orbital structure of VOPc on Ag (100) and TiOPc/Ag (100)[K. Noh et al., Nanoscale Horiz. 8, 624 (2023).]. XAS measurements were conducted at the EPFL-PSI Xtreme beamline, and simulated spectra from atomic multiplet calculations were optimized with a Bayesian Optimization (BO) algorithm. Our combined BO-multiplet approach retrieves input parameters for crystal field and intra-atomic interactions efficiently. We confirm VOPc's spin 1/2 in all configurations, with orbital energies varying based on surface and TiOPc interlayer interactions[J. H. Lee et al., Phys. Rev. B, 109, 235427 (2024).]. Finally, we demonstrate this method's effectiveness in uncovering magnetic interactions in metal-organic frameworks on surfaces, such as in the case of supramolecular metal complexes on Niobium diselenide[V. Vaňo et al., Phys. Rev. Lett., Accepted (2024).].