Failure of random matrix theory to correctly describe quantum dynamics.
Consider a classically chaotic system that is described by a Hamiltonian H(0). At t=0 the Hamiltonian undergoes a sudden change (H)0-->H. We consider the quantum-mechanical spreading of the evolving energy distribution, and argue that it cannot be analyzed using a conventional random-matrix theory (RMT) approach. Conventional RMT can be trusted only to the extent that it gives trivial results that are implied by first-order perturbation theory. Nonperturbative effects are sensitive to the underlying classical dynamics, and therefore the Planck's over 2 pi-->0 behavior for effective RMT models is strikingly different from the correct semiclassical limit.