Tue September 9, 2025 4:00 pm
CUA Seminar: Kostyantyn Kechedzhi – Beyond-classical quantum simulation
Location:MIT 4-270
Kostyantyn Kechedzhi, Quantum AI @ Google
Ten Minute Talk:"EngageCUA: Fall Open House + Outreach Opportunities" by
Haley Nguyen,
Harvard
Simulation of quantum phenomena in nature is a promising early application of quantum computers due to their ability to efficiently process highly entangled wave functions that otherwise require exponentially large classical memory and/or processing times. However, in a typical many-body evolution growth of entanglement is associated with thermalization that quickly erases fine-grained details of the evolution whereas the remaining macroscopic features can often be described by an effective (approximate) classical theory. A natural question is whether there is information about such a many-body quantum system that can be learned with a quantum computer but not by classical means. In this talk we provide theoretical and experimental analysis of this question by considering moments of correlation operators, a new type of correlation functions, that exploit efficient time inversion to extract classically inaccessible information. We measure the 2nd and 4th moments of the correlation operator, a generalization of out-of-time order correlators. These moments reveal macroscopic features of the spectrum of the correlation operator and serve as a diagnostic of quantum chaos. Unlike standard correlation function measurements, even moments of the correlation operator remain remarkably sensitive to the underlying dynamics at long times, allowing us to characterize the circuit close to the ergodic regime. We present measurements of the 4th moment of the correlation operator using a Willow superconducting processor for circuits sufficiently large such that the equivalent computation on a classical supercomputer requires significantly longer runtime using best known classical algorithms. Our results pave a path toward using near-term quantum processors for Hamiltonian learning.