Rydberg-Mediated Entanglement in a Two-Dimensional Neutral Atom Qubit Array

We demonstrate high fidelity two-qubit Rydberg blockade and entanglement on a pair of sites in a large two-dimensional qubit array. The qubit array is defined by a grid of blue detuned lines of light with 121 sites for trapping atomic qubits. Improved experimental methods have increased the observed Bell state fidelity to $F_{\text{Bell}}=0.86(2)$. Accounting for errors in state preparation and measurement we infer a fidelity of $F_{\text{Bell}}^{\text{−}\mathrm{SPAM}}=0.88$. Accounting for errors in single qubit operations we infer that a Bell state created with the Rydberg mediated $C_Z$ gate has a fidelity of $F_{\text{Bell}}^{C_Z}=0.89$. Comparison with a detailed error model based on quantum process matrices indicates that finite atom temperature and laser noise are the dominant error sources contributing to the observed gate infidelity.

Figure 1

Atomic qubit array. (a) Averaged fluorescence image of 121 site array after ICA processing with the control and target sites used for the presented data labeled. (b) Loading histogram showing an average filling fraction of 55% with inset showing state detection inferred after blow away of atoms in $f=4$. Accounting for all 121 sites, the state detection error found from the overlap of Gaussians fitted to the $|0⟩$ and $|1⟩$ distributions was $\text{mean}=0.014$, $\text{median}=0.003$. (c) Single instance of atom loading. (d) Atom retention probability after measurement. The atom retention probability after two subsequent readout sequences was $\text{mean}=96.9\%$, $\text{median}=97.9\%$. The difference between mean and median is due to some sites on the edge of the array having lower retention due to optical aberrations.

Figure 2

Single site ground-Rydberg Rabi oscillations at ${\mathrm{\Omega }}_R/2\pi =2.5\mathrm{MHz}$ with negligible crosstalk to the surrounding eight sites (see Ref. [19] for analysis). All atoms in the array were prepared in $|1⟩$ directly before the targeted site was illuminated with Rydberg beams.

Figure 3

Population in $|1⟩$ after (a) control and (b) target qubit Rydberg pulses. The Rabi frequencies for the control and target pulses were ${\mathrm{\Omega }}_R/2\pi =3.6$ and 4.6 MHz, respectively, and gap time = 300 ns.

Figure 4

Eye diagram for target qubit with blockade and no-blockade curves of amplitude 0.91(6) and 0.85(3). The inset shows the target pulse sequence. The $C_Z$ gate was operated at $\theta =0.95$ radians.

Figure 5

Bell state preparation: (a) pulse sequence, (b) populations, (c) parity oscillation. The Rydberg pulses had lengths of $t_{\pi }=150\mathrm{ns}$, $t_{2\pi }=220\mathrm{ns}$, $t_{\mathrm{gap}}=300\mathrm{ns}$. The effective ground-Rydberg superposition time is $t_{\text{gr}}=t_{\pi }/2+t_{\mathrm{gap}}+t_{2\pi }+t_{\mathrm{gap}}+t_{\pi }/2=0.98\mu \mathrm{s}$. The microwave pulses are global (duration $35\mu \mathrm{s}$) for the initial and parity pulses and global pulses combined with a 459 nm Stark pulse (duration $70\mu \mathrm{s}$) for the site selected rotation.