High Fidelity Qubit Control in a Natural Si-MOS Quantum Dot using a 300 mm Silicon on Insulator Wafer

TL;DR

This work addresses the challenge of achieving high-fidelity single-qubit gates in natural silicon MOS quantum dots by combining fast ESR drive with real-time qubit-frequency feedback on a 300 mm industrial wafer. The authors demonstrate a 99.5% ± 0.3% average single-qubit fidelity via randomized benchmarking, supported by a driven coherence time $T^{Rabi}$ ≈ 11 μs at a Rabi frequency around 5 MHz and a Rabi Q-factor exceeding 50. Low-frequency magnetic and charge noise are characterized (σ_f ≈ 0.41 MHz; PSD ~ 1/f^{0.47}), yet the feedback and pulse-area calibration enable performance near the Rabi-limited bound $F_{limit} ≈ 1 - 1/(4 Q_{Rabi})$ ≈ 99.54%. The results highlight the feasibility of scalable quantum control in industrial silicon technology and point to routes for further improvements via faster feedback and reduced charge noise.

Abstract

We demonstrate high-fidelity single qubit control in a natural Si-MOS quantum dot fabricated in an industrial 300 mm wafer process on a silicon on insulator (SOI) wafer using electron spin resonance. A relatively high optimal Rabi frequency of 5 MHz is achieved, dynamically decoupling the electron spin from its 29-Si environment. Tracking the qubit frequency reduces the impact of low frequency noise in the qubit frequency and improves the $T^{Rabi}$ from 7 to 11 $μ$s at a Rabi frequency of 5 MHz, resulting in Q-factors exceeding 50. Randomized benchmarking returns an average single gate control fidelity of 99.5 $\pm$ 0.3%. As a result of pulse-area calibration, this fidelity is limited by the Rabi Q-factor. These results show that a fast Rabi frequency, low charge noise, and a feedback protocol enable high fidelity in these Si-MOS devices, despite the low-frequency magnetic noise.

Figures (9)

• Figure 1: Single qubit control a) Scanning electron micrograph of the device with the quantum dot indicated by the red dot, next to the single electron transistor, b) Pulse sequence for single qubit control, c) Charge stability diagram of the P1 quantum dot indicating the initialize (I), control (C), and measure (M) positions for the ST and P1 gates, d) Rabi oscillation at a Rabi frequency of 5 MHz with a $T^{Rabi}$ of 6.6 $\mu s$
• Figure 2: Single qubit control using qubit frequency feedback (a) The pulse sequence for the calibration of the qubit frequency. (b) Example of the contrast signal ($Measure_2- Measure_1$) around resonance, with a straight line fit to locate zero detuning. (c) Rabi chevron at a Rabi frequency of 5 MHz. (d) $1/{T^{Rabi}}$ as a function of detuning from the qubit frequency. The line is a fit to Eq. \ref{['eq:T2rabi']} with a detuning uncertainty of 0.4 MHz. Hence $1/T^{Rabi}$ can be described by the uncertainty in the effective Rabi frequency. At zero-detuning, the effective Rabi frequency is a minimum and the quality factor of the Rabi oscillation is limited by the accuracy of setting zero detuning. (e) Rabi oscillations at 5 MHz on resonance, with and without feedback during the measurement. Note that the feedback reduces the Rabi frequency, and improves $T^{Rabi}$ from 7 $\mu s$ to 11 $\mu s$.
• Figure 3: Power dependence of Rabi damping a) $T^{Rabi}$ as a function of Rabi frequency to study the microwave power dependence. b) Q-factor ($Q_{Rabi}=f_{Rabi}T^{Rabi}$) as a function of Rabi frequency. The shaded regions show one standard deviation around the average of 10 measurements of a Rabi oscillation of 600 shots. The results with and without feedback are compared and show that feedback improves the $T^{Rabi}$ for Rabi frequencies below 6 MHz, and gives more consistent measurements.
• Figure 4: Pulse calibration and randomized benchmarking (a) Pulse sequence for $\pi/2$ pulse calibration. (b) An example of the contrast signal ($Measure_2- Measure_1$) from the $\pi/2$ pulse calibration experiment around the matching pulse amplitude with a straight line fit to locate the optimal pulse amplitude. (c) Distribution of the optimized pulse amplitudes for a $\pi/2$ with a gaussian fit in purple. In (c,e) The mean is shown as the violet line, and the expected pulse amplitude distribution is shown as a grey dotted gaussian based on the standard deviation of the measured qubit frequency from data of Figure \ref{['fig:fig5']}. (d) Pulse sequence for $\pi$ pulse calibration. (e) Distribution of the optimized pulse amplitudes for a $\pi$ with a gaussian fit in purple. (f) Randomized benchmarking results, using the optimized pulse amplitudes and qubit frequency feedback, with an average single gate fidelity fitted to $99.5\pm0.3\%$ fidelity.
• Figure 5: Qubit low frequency noise analysis (a) The qubit frequency distribution measured from the qubit frequency feedback protocol during the measurement. (b) Power spectral density of the qubit frequency showing a $1/f^{0.47}$ dependence.