Speaker
Description
Analog quantum simulators have scaled rapidly to sizes that challenge classical simulation, offering a pathway to solving complex many-body physics problems. However, a paradox remains: how can we verify the quantitative accuracy of a device built to surpass our own computational capabilities? Without error correction, these NISQ devices are prone to calibration errors and environmental noise, often limiting their utility to qualitative observation rather than quantitative prediction.
In this talk, I will present a framework for Bounded-Error Quantum Simulation (BEQS), a protocol that transforms analog devices into verifiable computational tools. By combining efficient Hamiltonian and Lindbladian learning techniques with rigorous uncertainty propagation, we can derive certified confidence intervals for quantum observables directly from experimental data.
I will demonstrate the experimental implementation of this framework on a trapped-ion quantum simulator. By leveraging the method’s polylogarithmic sample complexity in system size and efficient postprocessing, we scale the approach to a 51-qubit system. We successfully reconstruct over 14,000 model parameters, revealing structural features of the simulator such as the exponential decay of long-range interactions and local versus collective dephasing. Finally, I will show how these learned models are used to generate rigorous error bounds for non-equilibrium dynamics, which are experimentally cross-validated. These results establish a blueprint for trusting the output of large-scale quantum simulators in the regime of quantum advantage.
