Variational method for learning Quantum Channels via Stinespring Dilation on neutral atom systems

Research output: Working paperPreprintProfessional

49 Downloads (Pure)


The state |ψ(t)⟩ of a closed quantum system evolves under the Schrödinger equation, where the reversible evolution of the state is described by the action of a unitary operator U(t) on the initial state |ψ0⟩, i.e.\ |ψ(t)⟩=U(t)|ψ0⟩. However, realistic quantum systems interact with their environment, resulting in non-reversible evolutions, described by Lindblad equations. The solution of these equations give rise to quantum channels Φt that describe the evolution of density matrices according to ρ(t)=Φt(ρ0), which often results in decoherence and dephasing of the state. For many quantum experiments, the time until which measurements can be done might be limited, e.g. by experimental instability or technological constraints. However, further evolution of the state may be of interest. For instance, to determine the source of the decoherence and dephasing, or to identify the steady state of the evolution. In this work, we introduce a method to approximate a given target quantum channel by means of variationally approximating equivalent unitaries on an extended system, invoking the Stinespring dilation theorem. We report on an experimentally feasible method to extrapolate the quantum channel on discrete time steps using only data on the first time steps. Our approach heavily relies on the ability to spatially transport entangled qubits, which is unique to the neutral atom quantum computing architecture. Furthermore, the method shows promising predictive power for various non-trivial quantum channels. Lastly, a quantitative analysis is performed between gate-based and pulse-based variational quantum algorithms.
Original languageEnglish
Number of pages11
Publication statusPublished - 19 Sept 2023


Dive into the research topics of 'Variational method for learning Quantum Channels via Stinespring Dilation on neutral atom systems'. Together they form a unique fingerprint.

Cite this