Robust stimulated Raman shortcuts to adiabatic passage with deep learning

Abstract One of the challenging tasks in quantum control is to manipulate quantum systems with high fidelity and as fast as possible. Simulated Raman shortcuts to adiabatic passage with invariant-based optimal control is an efficient technique accurately used to transfer population between two quantum states in three-level systems. This technique requires tuning parameters continuously which results in analog quantum control. However, a digital quantum controller design is of great importance in the era of digital quantum computing. Here, we use deep reinforcement learning to obtain digital Stokes and pump fields that can realize fast and accurate population transfer between states with the same parity in the three-level Λ configuration. We find that deep reinforcement learning follows exactly theshortcuts to adiabaticity (STA) based on dynamical invariant and leads to a robust population transfer against systematic errors and dephasing. This is a promising enhancement in digital quantum information processing.