Berlin 2024 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
QI: Fachverband Quanteninformation
QI 23: Quantum Control
QI 23.1: Talk
Thursday, March 21, 2024, 09:30–09:45, HFT-FT 131
Neural-network-supported preparation of cat states in Jaynes-Cummings model — •Pavlo Bilous1, Hector Hutin2, Benjamin Huard2, and Florian Marquardt1, 3 — 1Max Planck Institute for the Science of Light, Staudtstr. 2, 91058 Erlangen, Germany — 2Ecole Normale Supérieure de Lyon, CNRS, Laboratoire de Physique, 69342 Lyon, France — 3Department of Physics, Friedrich-Alexander-Universität Erlangen-Nürnberg, 91058 Erlangen, Germany
We present a neural-network (NN) based approach for control optimization in the Jaynes-Cummings model allowing for high-fidelity preparation of cat states ψcat(α) in the cavity. The NN is first trained on a random selection of α-values sampled from a region of interest and can be afterwards applied to any α from this region for construction of the ψcat(α) state. The data processing pipeline consisting of the NN and a Schrödinger equation solver ensures the construction of the proper fields driving the qubit and the cavity at the training stage. For each training point α, the controls are optimized to minimize the loss function defined as the infidelity of the resulting state ψ(α) with respect to the target state ψcat(α). We search for the control fields as an expansion in a so called B-spline basis used extensively in computational atomic physics. This approach reduces significantly the number of parameters needed to characterize the driving signals and ensures their well-behaved shape feasible for experimental implementation. We generalize our approach for the construction of other quantum states described by one or a few parameters.
Keywords: Jaynes-Cummings model; Cat states; Neural networks; Machine learning; B-splines