Long coherence silicon spin qubit fabricated in a 300 mm industrial foundry

TL;DR

This work demonstrates a long-lived singlet–triplet qubit in gate-defined silicon MOS double quantum dots fabricated in a 300 mm CMOS foundry, achieving a Hahn-echo coherence time of $T_2^{\text{Hahn}} = 4~\text{ms}$ and revealing an exceptionally quiet detuning environment with $\delta\varepsilon_{\text{rms}} = 2.2~\mu\text{eV}$. Through noise spectroscopy, the authors identify strong in-phase correlations of charge noise between neighboring dots, which render the S--T$_0$ encoding robust to common-mode fluctuations and markedly suppress dephasing. The qubit exhibits two valley configurations with distinct $\Delta g$ and spin-valley couplings, and exchange tunability up to $\sim$200 MHz, enabling dual-axis control. The results underscore the potential of silicon quantum dots to adapt qubit encodings to the microscopic noise landscape, providing a practical path toward scalable quantum information processing in industrially fabricated architectures.

Abstract

Silicon spin qubits offer long coherence times, a compact footprint and compatibility with industrial CMOS manufacturing. Here, we investigate spin qubits hosted in quantum dots fabricated in a state-of-the-art 300 mm nanoelectronics foundry and demonstrate substantially enhanced coherence, achieving a Hahn-echo time of $T_2^{\text{Hahn}} = 4\,\mathrm{ms}$ for singlet--triplet oscillations. Employing noise spectroscopy and noise correlation measurements, we identify detuning noise with an amplitude of $δ\varepsilon_{\mathrm{rms}} = 2.2\,μ\mathrm{eV}$ (integrated over 90 s) and observe strong zero-phase correlations between two spatially separated spin qubits. The singlet--triplet basis intrinsically rejects these common-mode fluctuations, yielding a pronounced suppression of dephasing. Our results suggest that exploiting the versatility of silicon quantum dots to adapt the qubit encoding to the microscopic noise landscape represents a promising strategy for advancing scalable quantum information processing.

Figures (9)

• Figure 1: Device layout, charge stability, energy diagrams and states relevant for the singlet--triplet qubit, and characterization of singlet--triplet oscillations driven by Zeeman energy differences. (a) False-color scanning electron micrograph of the device. Gold, green, and blue indicate the first, second, and third gate layers, respectively. The black crossed squares are ohmic contacts. The left single-electron transistor (SET) functions as a charge detector, while the right one serves as a reservoir for the double quantum dot (region enclosed in dark blue). The double dot is defined beneath the plunger gates P1 and P2 (black dashed circles) within the channel shaped by the screening gates (SG, outlined in dark gold). (b) Detector current $I_{\textnormal{det}}$ as a function of plunger gate voltages $V_{P1}$ and $V_{P2}$, showing the charge stability diagram of a double dot in the (1,3)-(0,4) charge regime. The black arrow marks the detuning $\varepsilon$ axis, and the white arrows indicate the pulsing sequence used to generate $\ket{S}-\ket{T_0}$ oscillations. (c) Energy diagram of the singlet--triplet qubit relevant (1,3) and (0,4) charge states as a function of detuning at finite magnetic field. The avoided crossing arises from tunnel coupling between $\ket{S(1,3)}$ and $\ket{S(0,4)}$ states. The exchange energy $J(\varepsilon)$, corresponding to the energy difference between the hybridized singlet $\ket{S}$ and triplet $\ket{T_0}$ branches, is indicated in blue. In the far-detuned regime (W point), the states converge to $\ket{\downarrow \uparrow}$ and $\ket{\uparrow \downarrow}$ respectively, split by the Zeeman energy difference $\Delta E_\textnormal{Z}$, labeled in brown. (d) Detuning pulse sequence used to observe singlet–triplet oscillations. (e) Evolution of the quantum state at detuning point W on the Bloch sphere during the pulsing procedure in (d). (f) Probability of measuring $\ket{T_0}$ state as a function of time showing long lasting coherent $\ket{S}-\ket{T_0}$ oscillations.(g) Distribution of extracted $T_2^*$ times from (f), fitted with a Gamma distribution.
• Figure 2: Excited-valley singlet--triplet qubit and $\Delta g$ anisotropy. Red (blue) denotes data or states associated with the ground (excited) valley throughout panels (a)–(c). (a) Energy levels as a function of detuning $\varepsilon$ at finite magnetic field, extending Fig. \ref{['fig:fig1']}(c) to include both ground- and first-excited-valley (1,3) states. Arrows indicate initialization in the singlet states of ground $\ket{S^{gg}(1,3)}$ and excited $\ket{S^{ge}(1,3)}$ valley configuration. (b) Fourier transform of S--T$_0$ oscillations as a function of in-plane magnetic field $B_{\textnormal{x}}$ at fixed detuning position W. Two distinct frequency branches are visible corresponding to the ground- (dominant amplitude) and excited-valley (weaker amplitude) S--T$_0$ oscillations. White dashed lines mark anticrossings from spin-valley coupling at energies corresponding to valley-splitting $E_V^{ge}$ and $E_V^{eg}$. The inset shows model fits (solid lines) to the measured data (points). (c) $\Delta g$ values as a function of in-plane magnetic field angle $\phi$ (where $\phi=0$ corresponds to the $+x$ axis along $[110]$ in Fig. \ref{['fig:fig1']}) extracted from S--T$_0$ oscillations of ground and excited valley states. Circles are data, and solid lines are model fits.
• Figure 3: Characterizing exchange interaction. (a) Detuning pulse sequence used to drive exchange oscillations. (b) Evolution of the quantum state on the Bloch sphere during exchange time $\tau_J$ following the pulsing procedure in (a). (c) Probability of measuring $\ket{T_0}$ state as a function of exchange time $\tau_J$ and detuning $\varepsilon$, showing coherent exchange oscillations. (d-f) Extracted frequency, dephasing time $T_2^*$, and quality factor Q as a function of detuning, obtained from fits to the oscillations in (c). The inset in (e) shows fit (red line) to a dephasing model based on detuning noise.
• Figure 4: Hahn-echo experiment. (a) Detuning pulse sequence used for the Hahn-echo experiment. An exchange $\pi$-pulse is applied to refocus dephasing arising from fluctuations in $\Delta E_{\textnormal{Z}}$. (b) Probability of measuring the $\ket{T_0}$ state as a function of total waiting time $2 \tau_{\textnormal{W}}$ and additional delay $\Delta t$, measured using the sequence in (a). The phase shift of the oscillations at larger total waiting times originates from a combination of finite bias-tee response time and the $\Delta g$ susceptibility to plunger gate voltages. (c) Normalized oscillation amplitude extracted from (b) as a function of total wait time. The data (green circles) are fitted with an exponential decay $\exp (-2 \tau_{\textnormal{W}} / T_2^{\textnormal{Hahn}} )$, yielding $T_2^{\textnormal{Hahn}} = 4ms$.
• Figure 5: Noise characterization. (a) Time traces of frequencies of single-spin qubit $f_{Q1}$ and $f_{Q2}$, and the exchange frequency $f_\textnormal{J}$ (blue, red and light-blue trace, respectively), shifted by arbitrary offsets for clarity. (b) Power spectral densities (PSDs) of $f_{Q1}$ (blue) and $f_{Q2}$ (red) with fits (solid lines) to a model combining $1/f^{\alpha}$ noise from weak coupling to an ensemble of two-level-fluctuators (TLFs) and a Lorentzian component from a single strongly coupled TLF. PSDs of the sum $\Sigma=f_{Q1} + f_{Q2}$ (grey) and the difference $\Delta = f_{Q1} - f_{Q2}$ (gold) are also shown. Throughout the measured frequency range, $\Delta$ --- corresponding to Zeeman energy difference --- exhibits significantly lower noise level than the individual spins. (c) Correlation strength and (d) correlation phase between noise signals $f_{Q1}$ and $f_{Q2}$ indicating strong in-phase correlations over the entire frequency range. The correlation strength peaks at the frequencies where the single TLF dominates the individual spin qubit PSDs. In panels (b)–(d), the shaded regions represent 90% confidence intervals.