Das (2602.21149) shows that classical many-body systems with local interactions have upper bounds on how chaotic they can be — and the tightest bound becomes independent of temperature in the thermodynamic limit.
Chaotic instability is quantified by the largest Lyapunov exponent: the rate at which nearby trajectories diverge. In quantum systems, the Maldacena-Shenker-Stanford bound provides a universal ceiling on chaos growth tied to temperature. No such universal bound exists for classical systems. Das provides the next best thing: explicit, Hamiltonian-specific bounds that depend on inertia and interaction geometry.
The bounds separate into two classes. Non-violable bounds are set by worst-case local curvature of the interaction potential and are insensitive to spatial structure — they hold instantaneously regardless of the system's state. Ergodic ceilings are tighter: they retain spectral information about collective modes and finite-size effects and hold under generic dynamical evolution. The ergodic ceiling captures collective suppression mechanisms that the non-violable bound misses.
For a one-dimensional coupled-rotor chain (a Josephson junction array), the ergodic ceiling has a closed analytic form. It creates a dynamically inaccessible region in the Lyapunov exponent-energy plane — a forbidden zone that the system cannot sustain chaotic growth into. Numerics confirm the prediction.
The striking result: in the thermodynamic limit, the ergodic ceiling approaches an inertial ceiling that depends only on the particle mass and interaction range, independent of temperature and coupling strength. The system's maximum chaos rate is set by its inertia — how quickly particles can respond to forces — and the geometry of their interactions. Heat the system arbitrarily and the chaos ceiling stays fixed. The memory of initial conditions has a minimum lifetime determined by mass and geometry alone.
Note (2026-11-11): checked against the full text of the paper (Das, arXiv 2602.21149). The structure is as the essay says: a strict instantaneous (non-violable) bound from local curvature and inertia, a sharper ergodic ceiling that keeps the spectrum of collective modes, a closed form for the one-dimensional rotor chain (a Josephson junction array) confirmed numerically, and a thermodynamic-limit inertial ceiling independent of temperature and coupling strength. Four points differ. (1) The essay says the inertial ceiling 'depends only on the particle mass and interaction range'. In the paper the non-violable inequality is that the largest Lyapunov exponent is at most 1 over 2m plus half the operator norm of the Hessian of the interaction potential, and for the rotor chain the ergodic ceiling is 1 over 2m times a factor that tends to 1 as the number of sites grows (the geometry-dependent correction goes as N to the power minus 2 in one dimension and N to the power minus 2 over d in d dimensions), so the limiting ceiling is set by the mass alone; I found no dependence on an interaction range. (2) The paper writes the ergodic ceiling with an approximate inequality (the rotor-chain formula carries a 'less than or approximately' sign) and describes it as holding under generic dynamical evolution, unlike the strict non-violable bound; the essay does not say the two have different standing. (3) The paper says the ceiling limits 'sustained' chaotic growth, and says classical systems do not admit a universal chaos bound; its results cover a class of natural Hamiltonian systems with local interactions, and the formula given is for nearest-neighbour cosine interactions. (4) The essay's phrase 'the memory of initial conditions has a minimum lifetime determined by mass and geometry alone' matches the paper's claim of a minimal microscopic timescale for loss of memory only for the thermodynamic limit, where the geometry term has vanished.