Speaker
Description
Validation and calibration are essential for the development of quantum computers and simulators, yet they remain challenging. Tomography provides complete information about a device but quickly becomes impractical, whereas randomized benchmarking offers a cheaper alternative that relies on simplifying assumptions about the underlying noise. These tools were developed primarily with quantum computation in mind, and dedicated protocols for quantum simulation remain scarce. At the heart of the difficulty lies noise, an unavoidable feature of quantum hardware that limits performance and is itself hard to characterize.
In our work, we introduce a validation principle designed to exploit noise when it cannot be avoided. The approach harnesses the many-body quantum Mpemba effect, in which a careful choice of initial state can substantially accelerate or slow down relaxation towards a steady state. Nonequilibrium many-body dynamics are extraordinarily sensitive to imperfections in state preparation and control, and the Mpemba effect acts as a dynamical magnifying glass that converts otherwise hidden sources of error into measurable signatures. By preparing fast-relaxing and slow-relaxing states and tracking their trajectories, we obtain a stringent benchmark of device performance together with an on-chip calibration routine, both exploiting collective effects beyond the reach of single- and two-qubit diagnostics.
We demonstrate the protocol on two neutral-atom processors, QuEra's Aquila and Pasqal's Fresnel, implementing a dissipative six-atom Ising ring. Using only measurements in the computational basis, we benchmark and compare the two devices and identify complementary systematic errors, an amplitude calibration error on Aquila and a detuning drift on Fresnel. This constitutes the first observation of the many-body quantum Mpemba effect in an open quantum system.
| I am the presenting author | Yes |
|---|