Phase-Space Atlas: A Dynamically Faithful Archive for Driven Many-Body Systems with Analytic Reconstruction

We introduce the Phase-Space Atlas, a novel archival format for periodically driven many-body systems that stores only initial conditions and per-segment dynamical models, enabling exact analytic reconstruction of arbitrary observables at any time. Applied to 17 days of full-cadence SDO/AIA 171Å data (129 three-hour patches), the method compresses raw imagery exceeding 1.4~TB to 116~kB of turning geometry (plus optional 26~MB initial images for pixel recovery) while preserving all phase-mediated observables (stiff axes, basin depth, angular Lyapunov proxies, recurrence structure) losslessly. Global concatenation confirms persistent rank-3 basin structure over weeks. The protocol supports random-access latency $