In the Kuramoto-Sivashinsky equation, the initial energy selects the attractor.
The Kuramoto-Sivashinsky (KS) equation is a canonical model for spatiotemporal chaos — flame fronts, thin film flows, chemical reaction fronts. At fixed parameters, it produces apparently ergodic chaotic dynamics. The standard picture: you set the viscosity, and the system settles onto its attractor regardless of how you start it.
Ashtari and Schneider (arXiv:2603.09811) show this is wrong. The state space is layered. At fixed parameters, multiple invariant sets coexist — chaotic attractors at higher energy levels, periodic orbits (traveling waves) at lower ones. Which layer the system reaches depends on the initial condition's energy. Amplify the same spatial profile by different amounts and you reach different dynamical fates.
The periodic orbits have a simple structure: their period decreases inversely with initial energy. Higher energy means faster traveling waves. In transitional parameter regions, the layers interleave — periodic dynamics at low energy coexists with strange attractors at higher energy, not as transients but as persistent, distinct dynamical regimes occupying the same parameter point.
The mechanism is symmetry. The KS equation with periodic boundaries has continuous translational symmetry, which shows up as a degenerate neutral direction in the Lyapunov spectrum. This degeneracy partitions the state space into layers that cannot communicate. The initial energy places the trajectory on a particular layer, and symmetry prevents it from transitioning to another.
The structural point: symmetry doesn't just constrain individual solutions — it organizes the entire state space into non-communicating strata. What looks like a single chaotic system is a stack of independent dynamical systems, accessed by energy.
Ashtari and Schneider, "Symmetry-driven layered dynamics in the Kuramoto-Sivashinsky equation," arXiv:2603.09811 (2026).