Smoother starts and finishes cut errors in adiabatic quantum state preparation

[Attention Griffin]

Researchers at the University of Copenhagen and CWI in Amsterdam have found a simple way to reduce errors when preparing complex quantum states slowly. Quantum Zeitgeist covered the paper, which is published in an American Physical Society journal. The process is called adiabatic state preparation, where a system is gently moved from an easy starting state to a harder target state. If the change happens too quickly, errors creep in

The key finding is that the beginning and the end of the process matter most. If the system's Hamiltonian changes smoothly at both ends, with its rate of change easing to zero, the errors fall much faster as the total time grows. With n vanishing derivatives at each end, the infidelity drops as 1 over T to the power n plus 1, where T is the evolution time. In plain terms, easing in and easing out pays off

The authors also tightened an earlier theoretical error bound by one order. There is a catch, though. The improved scaling only dominates when the evolution is slow enough. For faster runs, a different exponential behaviour takes over, set by the size of the system and the smallest energy gap along the way

They tested the idea on a one dimensional mixed field Ising model using matrix product state simulations. They also tried the schedules on an 11 atom chain on QuEra's Aquila neutral atom machine. In the noiseless simulations, the smoother schedules cut the final infidelity by orders of magnitude. The paper is by Emil Pedersen, Freek Witteveen, Klaus Mølmer and Matthias Christandl

This is the kind of practical theory that quantum simulators can use right away. It does not need new hardware, just a better choice of how to ramp the controls. Analogue machines like Aquila could benefit straight away, and so could anyone using adiabatic methods for optimisation

Have you seen schedule shaping make a difference in your own experiments? And do results like this make analogue quantum simulators more attractive?

Error In Quantum State Prep Drops As Evolution Gets Smootherquantumzeitgeist.com