Preparation, characterization, and certification of quantum states in Rydberg atom arrays
Edison Carrera Unaucho
Abstract
Programmable Rydberg atom arrays provide a versatile platform for generating, controlling, and measuring interacting many-body quantum states. This thesis investigates which states can be explored using global controls, whether selected states can be prepared efficiently, how robust their useful properties are to imperfections, and how these properties can be certified from finite experimental data. First, we study ensembles generated by random global pulse sequences and compare their statistical properties with those of Haar-random states within the dynamically accessible symmetry sector. Weak interactions generate correlations slowly, whereas strong Rydberg blockade confines the dynamics to a restricted set of configurations. At intermediate interaction strengths, the system develops Haar-like statistical signatures on experimentally relevant timescales, revealing a compromise between interaction-induced exploration and blockade-induced confinement. However, quantum optimal-control calculations show that highly entangled target states are systematically more difficult to prepare than weakly entangled ones, demonstrating that expressibility does not imply trainability. Second, we use gradient-based quantum optimal control to prepare spin-squeezed states. For small systems, the optimized protocols approach the finite-size lower bound of the Wineland parameter and produce substantially stronger squeezing than uncontrolled quantum-quench dynamics. Pulses optimized for N=10 can be transferred without re-optimization to larger arrays, reaching a Wineland parameter of 0.227 for N=50. The robustness analysis further shows that the states with the strongest squeezing under unitary dynamics are not necessarily the most resilient to dephasing: states with a larger mean-spin length can retain greater metrological usefulness in the presence of noise. Finally, we formulate spin-squeezing certification with finite statistics as a hypothesis-testing problem. We derive rigorous non-asymptotic bounds on the corresponding p-value, thereby controlling the probability that finite-sample fluctuations lead to a false certification of squeezing. Together, these results show that reliable quantum-state engineering requires not only broad state-space exploration, but also efficient preparation, robustness to imperfections, and statistically controlled certification.