Fast and Unconditional All-Microwave Reset of a Superconducting Qubit

Active qubit reset is a key operation in many quantum algorithms, and particularly in quantum error correction. Here, we experimentally demonstrate a reset scheme for a three-level transmon artificial atom coupled to a large bandwidth resonator. The reset protocol uses a microwave-induced interaction between the $|f,0⟩$ and $|g,1⟩$ states of the coupled transmon-resonator system, with $|g⟩$ and $|f⟩$ denoting the ground and second excited states of the transmon, and $|0⟩$ and $|1⟩$ the photon Fock states of the resonator. We characterize the reset process and demonstrate reinitialization of the transmon-resonator system to its ground state in less than 500 ns and with 0.2% residual excitation. Our protocol is of practical interest as it has no additional architectural requirements beyond those needed for fast and efficient single-shot readout of transmons, and does not require feedback.

Figure 1

(a) Simplified schematic of the experimental setup. A transmon (orange) is coupled to two Purcell-filtered resonators. The readout resonator (green) is connected to room temperature electronics (description in the main text), while the reset resonator (blue) is connected to two $50\mathrm{\Omega }$ loads thermalized at base temperature. (b) Jaynes-Cummings ladder diagram of the transmon-reset resonator energy levels. The purple and light blue arrows represent the $e\text{−}f$ and $f0\text{−}g1$ pulsed coherent drives, respectively, and the black arrow labeled $\kappa$ illustrates the resonator decay process. (c) Illustration of the pulse schemes used to test the reset protocol. We initialize the qutrit to its ground state passively or optionally with an unconditional reset, then prepare the desired state $|g⟩$, $|e⟩$, or $|f⟩$ with control pulses (labeled ${\pi }_{ge}$ and ${\pi }_{ef}$). We reset the qutrit by simultaneously applying flattop $e\text{−}f$ (purple) and $f0\text{−}g1$ (light blue) pulses for a reset time $t_r$. The resulting qutrit state is then measured by applying a microwave tone to the readout resonator (green).

Figure 2

(a) Population $P_g$ vs the frequency ${\nu }_{f0g1}$ of a flattop $f0\text{−}g1$ pulse, of amplitude $V_{f0g1}$, applied to the qutrit initially prepared in $|f,0⟩$. (b) Measured ac Stark shifts ${\overline{\mathrm{\Delta }}}_{f0g1}$ and ${\overline{\mathrm{\Delta }}}_{ef}$ of the $f0\text{−}g1$ (blue diamonds) and $e\text{−}f$ (purple triangles) transitions, vs amplitude $V_{f0g1}$ of the f0-g1 drive. The solid lines are quadratic fits to the data. (c) Population $P_e$ vs frequency ${\nu }_{ef}$ of a flattop $e\text{−}f$ $\pi$ pulse applied on the qutrit, initially prepared in state $|e,0⟩$, in the presence of a continuous $f0\text{−}g1$ drive of amplitude $V_{f0g1}$. (d) Population $P_f$ vs duration $t$ of a resonant flattop $e\text{−}f$ pulse, of amplitude $V_{ef}=8\mathrm{mV}$. (e) Extracted Rabi rates ${\mathrm{\Omega }}_{ef}$ and $\tilde{g}$, of the $e\text{−}f$ (purple triangles) and $f0\text{−}g1$ (blue diamonds) drives versus their amplitude, $V_{ef}$ and $V_{f0g1}$. The solid lines are linear fits. (f) Population $P_f$ vs duration $t$ of a resonant square $f0\text{−}g1$ pulse, of amplitude $V_{f0g1}=444\mathrm{mV}$. The pulse schemes used to acquire the data shown in panels (a), (c), (d), and (f) are shown as insets, with the $f0\text{−}g1$ and $e\text{−}f$ pulse envelopes represented in blue and purple, respectively. The solid lines in (a) and (c) are fits to Gaussians. The solid lines in (c) and (f) are fits to Rabi oscillation models described in Ref. [34].

Figure 3

(a) Calculated reset rate $\mathrm{\Gamma }/\kappa$, vs Rabi rates $\tilde{g}/\kappa$ and ${\mathrm{\Omega }}_{ef}/\kappa$. The overdamped parameter region is hatched. The red line shows the values of ${\mathrm{\Omega }}_{ef}$ maximizing $\mathrm{\Gamma }$ as a function of $\tilde{g}$, and corresponds to the optimal branch where it is solid. (b) Maximized reset rate ${\mathrm{\Gamma }}_{\max }/\kappa$ vs $\tilde{g}/\kappa$ [we follow the red line from (a)]. In (a) and (b), the parameter configurations $A$, $B$, and $C$ at which the reset dynamic was probed (see main text and Fig. 4) are indicated with colored symbols and the exceptional point is represented by a black cross.

Figure 4

Qutrit populations $P_{g,e,f}$ vs reset time $t_r$ with reset parameters in configuration $A$ (see main text), and (a) system initialized in $|e,0⟩$ or (b) in $|f,0⟩$. The solid lines in (a) and (b) are calculated from Eq. (2). (c) Excited population $P_{\mathrm{exc}}$ as a function of reset time $t_r$, when the qutrit is initialized in $|e,0⟩$, shown for reset parameter configurations $A$, $B$, and $C$. The solid lines are calculated from a master equation simulation.