A two-way approach to out-of-time-order correlators
Yingfei Gu, Alexei Kitaev, Pengfei Zhang
TL;DR
This work develops a two-way mean-field framework to compute out-of-time-order correlators (OTOCs) in all-to-all interacting systems, notably the SYK model, by propagating forward and backward scrambling modes on a doubled thermofield double background. By decomposing scrambling into coherent and incoherent parts and introducing size operators, the authors derive a nonlinear OTOC equation that can be solved exactly in the large-$q$ limit, yielding a closed-form late-time expression featuring a confluent hypergeometric function. A kinetic-equation approach on the double Keldysh contour connects perturbations to a solvable Liouville-type equation, clarifying the scramblon dynamics and their black-hole analogies. The work also establishes a link between the Lyapunov exponent and high-frequency tails of the spectral function, providing a nontrivial bound that ties chaotic growth to spectral features. Overall, the paper offers a concrete, tractable route to exact OTOCs in a prototypical chaotic quantum many-body system and deepens the connection between quantum chaos and gravitational scattering pictures.
Abstract
Out-of-time-order correlators (OTOCs) are a standard measure of quantum chaos. Of the four operators involved, one pair may be regarded as a source and the other as a probe. A usual approach, applicable to large-$N$ systems such as the SYK model, is to replace the actual source with some mean-field perturbation and solve for the probe correlation function on the double Keldysh contour. We show how to obtain the OTOC by combining two such solutions for perturbations propagating forward and backward in time. These dynamical perturbations, or scrambling modes, are considered on the thermofield double background and decomposed into a coherent and an incoherent part. For the large-$q$ SYK, we obtain the OTOC in a closed form. We also prove a previously conjectured relation between the Lyapunov exponent and high-frequency behavior of the spectral function.
