Adiabatic reverse annealing with dephasing

Adiabatic Reverse Annealing (ARA) is a variant of quantum annealing (QA) exploiting path modification to perform a local search in the solution space of optimization problems. Starting from a classical configuration supposedly close to the target solution, in ARA quantum fluctuations are nonmonotonically varied so as to improve the quality of the trial solution. Compared to Iterated Reverse Annealing (IRA), ARA can exponentially speed up QA by avoiding first-order quantum critical points in the phase diagram of the optimization problem.

Early works concerning the topic showed that ARA in a unitary setting is efficient in this regard for the fully connected ferromagnetic p-spin model, an exactly solvable model frequently used to benchmark QA [1,2]. In our work, we ask how this scenario changes in the presence of dissipation. Using a Markovian master equation, we show that dephasing can improve certain success metrics such as the ground state probability and the time to solution. Furthermore, we argue that collective dephasing is more beneficial compared to individual dephasing according to these metrics.

[1] PRA 98, 022314 (2018)
[2] PRA 100, 052321 (2019)