Spin Readout in a 22 nm Node Integrated Circuit

Abstract

Constructing a quantum computer capable of broad and important applications is likely to require millions of addressable physical qubits, posing the challenge of large-scale integration of quantum systems with classical electronics. Fully depleted silicon-on-insulator CMOS technology has been used to develop a range of cryogenic electronic components for the control and readout of different qubit modalities interfaced on separate chips. However, recent measurements of quantum dots on this technology raise the tantalising prospect of realising control electronics and spin qubits on the same manufacturing platform, within a single integrated circuit (IC). Here, we demonstrate single-shot spin readout in addressable quantum dot devices within an IC fabricated using industry-standard 22 nm fully depleted silicon-on-insulator technology. We achieve spin-to-charge conversion via a ramped energy-selective measurement, detected using a radio-frequency single-electron transistor and addressed by on-chip cryogenic electronics. The observation of consistent readout visibilities exceeding 90% and millisecond spin relaxation times in two nominally identical devices within the addressable array supports the reproducibility of the unit cell. The successful observation of spin readout using this CMOS process marks a key step towards realising highly scalable and integrated spin qubits.

Figures (6)

• Figure 1: Device schematic and spin readout protocol. a) 22 nm FDSOI IC hosting 128 unit cells formed by the QD structures and the integrated multiplexing circuitry. Zoom in of the unit cells used in this experiment containing a few-electron QD with a rfSET charge sensor in parallel. The off-chip $LC$ resonant circuit contains a superconducting spiral inductor, $L =$ 18.7 nH, coupled to the transmission line via coupling capacitor $C_C =$ 0.9 pF and a parasitic capacitance of $C_P \sim$ 2.5 pF. A bias tee containing a resistor $R$ and capacitor $C_G$ is used to apply a dc source voltage and create a good rf ground. b) Charge stability diagram of the sensor-QD system in device A labelled with the electron occupation on each dot. The dotted square indicates the chosen transition to perform spin readout. Device B shows a similar stability map and single-electron charge transition chosen for spin readout. c) Charge transition $(n_\text{S},n_\text{Q}) = (N+1, 1) \rightarrow (N+2, 0)$ in b). Pulse sequence with Empty-Load-Read stages is applied to cycle the electron occupation of the qubit dot from 0 to 1. A random electron spin state is initialised in the load stage. d) Voltage pulse sequence applied to gates $G_\text{Q}$ and $G_\text{S}$ and the subsequent response of the qubit dot electrochemical potentials, $\mu_{\uparrow/\downarrow}$, and the single-shot charge sensor signal. A 'blip' in signal before a certain time threshold indicates a $\ket{\uparrow}$ electron loaded onto the dot.
• Figure 2: Single-shot spin readout. Histograms of the first tunneling out event, $t_{\text{out}}$, for 1000 single-shot traces at $B = 1.4$ T (a) and $B = 2.5$ T (b). Both distributions are fit to a generalized probability density function, $f_{\text{total}}(t)$, and the resulting individual spin state probability density functions, $f_{\uparrow / \downarrow}(t)$, are indicated. Dashed lines show the center of each spin state peak at time when $df_{\uparrow / \downarrow}(t)/dt = 0$. c) Time separation between spin state tunneling out events as a function of magnetic field for both devices. Shaded region shows bounds of the fit. Error bars plotted are given in terms of energy.
• Figure 3: Spin relaxation rate. Normalized $\ket{\uparrow}$ state fraction as a function of $t_{\text{load}}$ at three magnetic fields for device A (a) and device B (b). Each plot it fitted to an exponential decay to obtain time constant $T_1$. Residual $\ket{\uparrow}$ fraction could be due thermal spin revivals or long lifetime $\ket{\uparrow}$ states, both of which are more likely at low field. c) Relaxation rate dependence on magnetic field for both devices plotted with a logarithmic y-axis.
• Figure 4: Spin Readout Fidelity. a) Histogram of $t_{\text{out}}$ in Fig. \ref{['fig:peak sep']}b fit to the analytical model. Dashed line indicates a maximum visibility of 92% at optimal time threshold $t = 0.65$ ms and the shaded regions show the areas of error. b) Visibility as a function of magnetic field for both devices obtained from the fits to the data in Fig. \ref{['fig:peak sep']}.
• Figure 5: Reservoir temperature and qubit dot lever arm. a) Width of the thermal broadening for the qubit dot unloading signal as a function of fridge temperature. b) Qubit dot unloading signal during the diagonal ramped 'read' stage at rf power -2 dBm and -12 dBm.