A universal neutral-atom quantum computer with individual optical addressing and non-destructive readout

TL;DR

A universal neutral-atom quantum computer with gate rates limited by optical switching times, rather than shuttling, by individually addressing tightly focused laser beams at an array of single atoms is demonstrated.

Abstract

Quantum computers must achieve large-scale, fault-tolerant operation to deliver on their promise of transformational processing power [1-4]. This will require thousands or millions of high-fidelity quantum gates and similar numbers of qubits [5]. Demonstrations using neutral-atom qubits trapped and manipulated by lasers have shown that this modality can provide high two-qubit gate (CZ) fidelities and scalable operation [6-13]. However, the gates in these demonstrations are driven by lasers that do not resolve individual qubits, with universal computation enabled by physical mid-circuit shuttling of the qubits. This relatively slow operation may greatly extend runtimes for useful, large-scale computation. Here we demonstrate a universal neutral-atom quantum computer with gate rates limited by optical switching times, rather than shuttling, by individually addressing tightly focused laser beams at an array of single atoms. We achieve CZ fidelity of 99.35(4)% and local single-qubit RZ gate fidelity of 99.902(8)%. Moreover, we demonstrate non-destructive readout of alkali-atom qubits with 0.9(3)% loss, which boosts operational speed. This technique also enables us to measure a state-of-the-art CZ fidelity of 99.73(3)% when excluding atom-loss events, which may be mitigated through erasure conversion. Our results represent a critical step towards large-scale, fault-tolerant neutral-atom quantum computers that can execute computations on practical timescales.

Figures (10)

• Figure 1: System overview.a, Image of array of Cs atoms, trapped in a glass cell. b, Locally addressed $R_Z(\phi)$ (beam shown in blue) and CZ (beams show in purple) gates are applied via beams propagating perpendicular to the atom plane. Microwave-based global $R(\theta,\phi)$ gates (labeled GR) address the entire array. c, Layout of key lasers implementing trapping and gates. Rydberg excitation is implemented by counter-propagating 459 nm (blue) and 1040 nm (red) lasers. Rydberg illumination for control and target qubits are sourced from separate pulse shaping systems (labeled A and B). An optical tweezer array (1064 nm) is generated via AODs and combined with the 1040 nm light. Readout fluorescence is collected and imaged on an electron-multiplying CCD (EMCCD). d, Atomic levels and addressing fields, including lasers for two-photon Rydberg excitation, readout, and optical pumping (OP). Rydberg beams are circularly polarized ($\sigma^+$ and $\sigma^-$ for 459 nm and 1040 nm, respectively). Rydberg blockade shift denoted by $V$.
• Figure 2: Single-qubit gates.a, Rabi oscillations driven by a global microwave field. b, Ramsey oscillations from local qubit $R_Z(\phi)$ control. Error bars in (a, b) represent projection noise. c, d, Randomized benchmarking results for global $R(\theta,\phi)$ and local $R_Z(\phi)$ gates, respectively. Each point is an average over multiple random circuits (10 per depth for $R(\theta,\phi)$ and 5 per depth for $R_Z(\phi)$ ), with error bar given by standard error of the mean. e, Randomized benchmarking results, plotted as cumulative distribution function (CDF) of 24-qubit (or 1-qubit) results. For 24-qubit results, gates are applied to each of the 24 qubits in each circuit. For the single qubit results, the reported uncertainty is the benchmarking fit uncertainty. For the 24-qubit results, the reported uncertainty is the ensemble standard deviation.
• Figure 3: Characterization of CZ gates.a, Rabi oscillations between $\ket{1}$ and $\ket{R}$, at a rate of $2 \pi \times 3$ MHz. b, Rydberg spectroscopy, measured while a neighboring atom is already in the Rydberg state. Decreased qubit $\ket{1}$ population indicates simultaneous excitation to the Rydberg state, which is only possible when the excitation is detuned (x-axis) by the Rydberg blockade. Error bars in (a, b) represent projection noise. c, Phase profile of 1040 nm optical pulse used in the CZ gate, measured using optical heterodyne interferometry. d, Circuit used as cost function for CZ gate optimization. Global $R(\theta,\phi=0)$ gates denoted as $X_\theta$. e, Circuits used for characterization of CZ gates include a main block of $(N-3)$ pairs of alternating Haar-random global SU(2) gates and CZ gates, as well as an inversion operation (consisting of three CZ gates and four global rotation gates $R_i$) and a padding circuit that reduces the spurious impact of $R(\theta,\phi)$ errors on extracted CZ fidelity. For circuits with $N=0$ the "Main" and "Inversion" blocks are omitted entirely. f, Observed probability to retain both atoms as a function of the number of CZ gates ($N$) in the circuit, along with the fit used to extract the retention probability. Decay is fit to an asymptote of zero. g, Observed probability to obtain the expected state-selective measurement outcome, given that both atoms were retained, as a function of the number of CZ gates in the circuit, along with the extracted cycle polarization. Decay is fit to an asymptote of 25%. Data in f, g are obtained from sequential NDSSR (g) and occupancy readout (f) measurements on the same quantum state. Each point is obtained as an average over between 18 and 48 circuits (183 total circuits in dataset), with an error bar corresponding to standard error of the mean.
• Figure 4: Non-destructive state-selective readout. NDSSR signal distribution from 43 repetitions of SPAM characterization experiment. Each experiment executes two circuits - empty circuit and circuit with single $R(\pi,0)$ gate - with approximately 600 shots each, resulting in approximately 25800 camera frames for each circuit. The turquoise-colored distribution describes the signals where atoms are expected to be in a bright state ($|F=4\rangle$) and the purple-colored distribution describes the signals for a dark state ($|F=3\rangle$). The samples for each distribution come from the deterministically loaded diagnostic site pair used for CZ gate fidelity benchmarking. Photo-electron counts for each site were calculated from the raw EMCCD camera analog-digital units using nominal EM gain and pre-amplifier gain. Each quadrant of the plane defined by NDSSR and post-NDSSR photo-electron counts contains two text labels showing total counts from $|F=4\rangle$ preparation and measurement experiment (turquoise) and counts from $|F=3\rangle$ preparation and measurement experiment (purple). a, Sampled probability distribution of photo-electron counts for bright and dark states. The dashed black line denotes the discrimination threshold that yields discrimination fidelity of 99.6(2)%. The shaded areas under the distributions on both sides of the discrimination threshold denote signals associated with state detection error. Depumping and atom loss during the NDSSR process and microwave pulse infidelity during state transfer from the stretched state to the bright qubit state give rise to bright state detection error of 3.1(7)%, which is reduced to 2.6(7)% when atom loss is post-selected out. The dark state detection error of 1.6(5)% is believed to be a combination of (partial) state preparation error and $F=3 \rightarrow F=4$ repumping during imaging from residual leakage of repump laser intensity. The middle inset shows a single NDSSR measurement of a four-qubit GHZ state realized in the 24-qubit array. b, NDSSR and post-NDSSR occupancy readout photo-electron counts. The bright state atom loss rate is 1.0(4)% and the dark state atom loss rate is 0.8(3)%. The atom loss rates and state detection errors and their uncertainties provided are the means and standard deviation of their measured values for the 43 SPAM characterization experiments that comprise the dataset.
• Figure 5: 3D Model of the Quantum Processing Unit (QPU), highlighting the dual-source beam steering modules for both Rydberg wavelengths, a microwave horn for global $R(\theta,\phi)$ gates, AODs for generating a 2D array of 1064 nm optical tweezers, the electron-multiplying CCD (EMCCD) for imaging, the optics supporting non-destructive state-selective readout (NDSSR), the ultra-high-vacuum glass cell, and the high-NA lenses for focusing addressing beams and collecting readout light.