Fast High-Fidelity Readout of a Single Trapped-Ion Qubit via Machine-Learning Methods

Abstract

In this work, we introduce machine learning methods to implement readout of a single qubit on 171 Yb + trapped-ion system. Different machine learning methods including convolutional neural networks and fully-connected neural networks are compared with traditional methods in the tests. The results show that machine learning methods have higher fidelity, more robust readout results in relatively short time. To obtain a 99% readout fidelity, neural networks only take half of the detection time needed by traditional threshold or maximum likelihood methods. Furthermore, we implement the machine learning algorithms on hardware-based field-programmable gate arrays and an ARM processor. An average readout fidelity of 99.5% (with 10 5 magnitude trials) within 171 µs is demonstrated on the embedded hardware system for 171 Yb + ion trap. * Program available on GitHub: https://github.com/quantumiracle/On board FNN qubit discrimination †

Scalable hyperfine qubit state detection via electron shelving in the 2D5/2 and 2F7Edmunds1, Tan2, Milne3 et al. 2021Phys. Rev. A16070View full textAdd to dashboardCite

“…V B). (b) Comparison of results for 171 Yb + , with red circle markers representing standard detection techniques that can be straightforwardly implemented without advanced detector technology [21,28,34,35,39]. The shelving technique introduced in this work similarly requires no additional detector hardware.…”

“…Wölk et al [34] analyzed the time-resolved detection methods for the case of 171 Yb + , which had been experimentally demonstrated for optical qubits by Myerson et al [25] and hyperfine qubits by Hume et al [32] and Hemmerling et al [33]. Other software-based approaches include recent work by Ding et al [35] investigating the use of machine-learning methods for state estimation, implemented in hardware on a field-programmable gate array with a single 171 Yb + qubit; they achieved results similar to those of Seif et al [28], who applied machine-learning methods to the time-resolved readout from a PMT array in postprocessing.…”

“…The inset shows decay channels from 2 D 5/2 , with simplified hyperfine notation. use of new complex imaging and detection hardware [27][28][29][30][31] or software-based data processing of time-resolved information [25,28,[32][33][34][35]. As qubit numbers are increased, however, the overhead for detection hardware and software can become limiting, motivating an exploration of complementary physics-based schemes to improve measurement fidelity in hyperfine qubits.…”

Characterizing mid-circuit measurements on a superconducting qubit using gate set tomography

“…The reflected cavity signal was downconverted and integrated using a matched kernel filter [49], and binned resulting in the well-separated readout histograms shown in Figure 5b. Integrating and taking the difference of these histograms yields a fidelity F = 96.35 %, while an approach using logistic regression [70] yields a fidelity F = 96.43 % ± 0.7 %.…”

Machine-learning-assisted electron-spin readout of nitrogen-vacancy center in diamond

“…This is suffered by the threshold method employed in single-shot readout systems for classifying two state distributions 12 . During the long-time laser illumination for state detection, for example, in a trapped ion or a superconducting qubit system, unpredicted state flip can happen, which could have been revealed in the time trace 13,14 . Other qubit systems where very few photons are detected in a single measurement, like room-temperature nitrogenvacancy (NV) center in diamond, require statistical readout other than single shot.…”