Automated Tune-Up of Poor Man's Majorana's In a two-Site Kitaev Chain Using a Neural Network
ORAL
Abstract
Majorana zero modes are predicted to appear at the ends of a Kitaev chain [1], which can be implemented using quantum dots (QDs) coupled to superconductors [2]. Even a two-site chain can host “Poor Man’s Majorana’s” in a parameter sweet spot [3], which was recently demonstrated experimentally [4]. The chain’s robustness to global perturbations grows as the number of sites is increased. This involves adding QDs and tuning their interaction to the sweet spot, where the elastic co-tunneling (ECT) and crossed-Andreev reflection (CAR) rates are equal. The size of the parameter space grows rapidly, making the tune-up of longer chains both time-consuming and difficult.
Previous work has shown that a generative machine learning model can predict the ratio of CAR and ECT rates based on charge stability diagrams of interacting quantum dots in a Kitaev chain [5]. In this talk, we show that the predictions of a neural network, retrained on a small experimental data set, can be used to set gate voltages, which tune a two-site Kitaev chain into the sweet spot in realtime. We show that the algorithm is able to recognize the sweet spot and find it. In addition, the predictions of the neural network match the experimentally found values accurately. This paves the way for the automated tune-up of longer Kitaev chains.
[1] Kitaev, A. Yu. "Unpaired Majorana fermions in quantum wires." Physics-uspekhi 44.10S (2001): 131.
[2] Sau, Jay D., and S. Das Sarma. "Realizing a robust practical Majorana chain in a quantum-dot-superconductor linear array." Nature communications 3.1 (2012): 964.
[3] Leijnse, Martin, and Karsten Flensberg. "Parity qubits and poor man's Majorana bound states in double quantum dots." Physical Review B 86.13 (2012): 134528.
[4] Dvir, Tom, et al. "Realization of a minimal Kitaev chain in coupled quantum dots." Nature 614.7948 (2023): 445-450.
[5] Koch, Rouven, et al. "Adversarial Hamiltonian learning of quantum dots in a minimal Kitaev chain." arXiv preprint arXiv:2304.10852 (2023).
Previous work has shown that a generative machine learning model can predict the ratio of CAR and ECT rates based on charge stability diagrams of interacting quantum dots in a Kitaev chain [5]. In this talk, we show that the predictions of a neural network, retrained on a small experimental data set, can be used to set gate voltages, which tune a two-site Kitaev chain into the sweet spot in realtime. We show that the algorithm is able to recognize the sweet spot and find it. In addition, the predictions of the neural network match the experimentally found values accurately. This paves the way for the automated tune-up of longer Kitaev chains.
[1] Kitaev, A. Yu. "Unpaired Majorana fermions in quantum wires." Physics-uspekhi 44.10S (2001): 131.
[2] Sau, Jay D., and S. Das Sarma. "Realizing a robust practical Majorana chain in a quantum-dot-superconductor linear array." Nature communications 3.1 (2012): 964.
[3] Leijnse, Martin, and Karsten Flensberg. "Parity qubits and poor man's Majorana bound states in double quantum dots." Physical Review B 86.13 (2012): 134528.
[4] Dvir, Tom, et al. "Realization of a minimal Kitaev chain in coupled quantum dots." Nature 614.7948 (2023): 445-450.
[5] Koch, Rouven, et al. "Adversarial Hamiltonian learning of quantum dots in a minimal Kitaev chain." arXiv preprint arXiv:2304.10852 (2023).
*This work has been supported by the Dutch Organization for Scientific Research (NWO) and Microsoft Corporation Station Q.
–
Presenters
-
David van Driel
- Delft University of Technology