Towards scaling-up single-electron currents: multiplexed operation of quantum-dot pumps with accelerated characterisation using machine learning

The last decade saw a rapid progress in the development of semiconductor-quantum-dot single-electron pumps, devices that transfer electrons one by one. The error rate of single-electron transfer is approaching 1 part in 10^7 [1,2], raising expectations to be used as primary standards directly following the definition of SI ampere. However, the magnitude of the generated current is limited to ~100 pA in the present device technology, which may ultimately limit the uncertainty of the measurements. In this presentation we introduce multiplexer device technology that can potentially break the 10^7 barrier. This technology allows us to control a large number of devices with a limited number of control gates. We have fabricated devices with 64 single-electron pumps on one chip and characterised them in one single cool down in a dilution refrigerator [3]. We demonstrate a statistical study of how the operating parameters of GaAs-based single-electron pumps vary between devices. We also show that the multiplexer technology can be used to parallelise the electron pump current, generating a scaled-up current that is not achievable with a single device. To overcome scalability limitations of established traditional characterization routines we present a novel active learning sparse measurement framework to accelerate the pump characterization process with machine learning [4]. Compared to traditional approaches it reduces the number of required measurements by about an order of magnitude.

[1] F. Stein et al., Metrologia 54, S1 (2017); https://doi.org/10.1088/1681-7575/54/1/S1.
[2] S. P. Giblin et al., Metrologia 60, 055001 (2023); https://doi.org/10.1088/1681-7575/ace054.
[3] S. Norimoto et al., Appl. Phys. Lett. 125, 113501 (2024); https://doi.org/10.1063/5.0225998.
[4] N. Schoinas et al., Appl. Phys. Lett. 125, 124001 (2024); https://doi.org/10.1063/5.0221387.