Hundred-hertz quantum circuit iteration rate in a reusable neutral-atom array

AI summary

Overview

• Research area: Quantum physics / quantum computing hardware — neutral-atom array processors, non-destructive qubit readout, and system throughput.
• Technical level: Intermediate (assumes familiarity with qubits, fluorescence readout, and randomized benchmarking, but the argument is hardware-architecture driven).
• Scope: An experimental demonstration that a 10-atom rubidium array with chip-based photonic readout and atom reuse can execute complete quantum circuits at a raw rate of 101 Hz and a normalized Fisher information rate of 57.7 Hz.

What This Paper Is About

Neutral-atom processors can trap many atoms and operate with high gate fidelity, but each experiment is slow: loading, imaging, rearranging, and destructive readout mean that a full circuit (initialization to readout) has typically been repeated fewer than ten times per second for the past decade. The authors attack both bottlenecks at once — array preparation and state readout — by using non-destructive, site-resolved detection through an integrated photonic chip so that the same atoms can be reused for one circuit after another. The goal is to raise the quantum circuit iteration rate (qCIR) and, more importantly, the actual information gained per unit time.

Key Contributions

1. Demonstration of a high-throughput neutral-atom processor built around a laser-written "volcano" optical channel mapping (OCM) chip, which creates a one-to-one correspondence between each atom site and an independent single-photon detector.
2. Experimental realization of non-destructive readout with atom reuse on a 10-qubit array, achieving a median single-shot retention probability of 99.7% and a median readout fidelity of 93.9% at an optimized readout duration of 1.5 ms.
3. Achievement of a raw qCIR of 101 Hz and a post-selected qCIR of 74.8 Hz, plus verification that repeated cycles accumulate Fisher information linearly (consistent, unbiased sampling).
4. Experimental validation of a previously proposed throughput-optimization methodology, quantified by a normalized Fisher information rate of 57.7 Hz — more than an order of magnitude better than a destructive-readout approach, even one with perfect readout fidelity (which would reach only 3.37 Hz).

Main Findings

• Readout fidelity versus atom survival is a tunable trade-off: Scanning the readout duration maps array-averaged readout infidelity against single-shot loss probability; the authors pick 1.5 ms deliberately, sacrificing some fidelity for higher overall throughput, and show a model curve from independently calibrated parameters (initial temperature, trap depth, scattering rate, collection efficiency) that matches the data.
• High retention over 100 cycles: In the benchmark sequence (200 ms loading, 45 ms existence detection, 7 ms per cycle, 100 cycles), the median retention probability is 99.7% and median readout fidelity is 93.9% across the 10 atoms.
• Raw and post-selected iteration rates: The raw qCIR is 101 Hz; applying a conservative post-selection (keeping only runs where both initial and final existence checks confirm the atom) gives 74.8 Hz using R_pos = R_raw × P_ret^100.
• Repeated sampling is statistically consistent: Rabi oscillations reconstructed from only the 25th, 50th, 75th, and 100th repetitions agree in bounds, frequency, and phase; combining all 100 repetitions within a 0.99 s sequence reduces error bars by an order of magnitude; the inverse population variance grows linearly with the number of repetitions included.
• Fast randomized benchmarking: Single-qubit RB using the gate set {X_{π/2}, X_{−π/2}, Y_{π/2}, Y_{−π/2}} yields an average gate fidelity of 99.93%, with each data point averaged over 10 atoms and sampled over 593 random sequences, taking 13.2 minutes instead of the several hours reported previously.
• Throughput figure of merit: At the chosen operating point (F = 93.9%, R_pos = 74.8 Hz), the normalized Fisher information rate Q = R_pos(2F−1)² is 57.7 Hz, versus 3.37 Hz for the same system without atom reuse even at ideal readout.

Methodology in Plain English

Ten single rubidium-87 atoms are held in a one-dimensional line of optical tweezers, with qubit states encoded in two hyperfine levels of the 5S₁/₂ manifold and manipulated by global microwaves. Rather than imaging the whole array with a camera, the authors collect the light from each atom into a separate waveguide inside a laser-written three-dimensional photonic chip: the waveguides are densely packed (50 μm pitch) at the atom end, matching the 4.5 μm atom spacing after optical conversion, and sparsely packed (127 μm pitch) at the other end, where they butt-couple to a standard fiber array feeding individual single-photon counters. This gives every atom its own detector channel, read out in real time by an FPGA.

Because detection is site-resolved and fast, the atoms do not have to be thrown away and reloaded after each shot. During readout the trap depth is raised to 5 mK and the trap and readout beams are alternated at 1 MHz with a 0.5 duty cycle to limit heating, so atoms survive many consecutive circuits. The team then characterizes the fidelity–survival trade-off by varying the readout duration, repeats a full initialize–operate–read cycle 100 times per loaded array, and checks consistency by fitting Rabi oscillations and randomized benchmarking curves built from those repeated samples.

Why This Matters

The paper reframes neutral-atom performance in terms of information per second rather than fidelity alone, and shows experimentally that atom reuse plus parallel single-photon detection delivers a step change in throughput without changing the underlying physics of the qubits.

• Quantum sensing and metrology: Ramsey/Rabi-type measurements benefit directly from a 57.7 Hz Fisher information rate, since frequency estimation precision scales with accumulated information rather than with single-shot fidelity.
• Algorithmic sampling and variational workloads: Workflows that need many circuit repetitions — optimizer loops, parameter scans, sampling-based algorithms — are limited by iteration rate, which is exactly the quantity raised here.
• Benchmarking and calibration in the lab: Randomized benchmarking at 13.2 minutes per point instead of hours makes routine, high-statistics device characterization practical.
• Mid-circuit measurement and feed-forward: Site-resolved non-destructive readout is a prerequisite for dynamical circuits with real-time feedback and for erasure/leakage detection.

Industry relevance: The approach uses off-the-shelf fiber arrays, single-photon detectors, and FPGA control, and is compatible with the existing "volcano" addressing architecture, so it is a hardware-level upgrade path for neutral-atom computing rather than a redesign. Higher throughput translates directly into lower cost per useful measurement for any commercial or cloud-accessed neutral-atom processor.

Future Directions

• Reduce the 7 ms cycle time (about 5 ms cooling plus 1.5 ms readout) toward roughly 0.1 ms each by improving photon collection efficiency and adopting readout schemes that suppress recoil heating, aiming at kilohertz qCIR.
• Extend the scheme from the current one-dimensional 10-atom array to two-dimensional qubit arrays, exploiting the volcano architecture's site-selective addressing.
• Implement dynamical quantum circuits with mid-circuit measurement and real-time feedback, enabled by the non-destructive, individually resolved readout.
• Push the throughput optimization further: the paper notes that over an order of magnitude improvement in Q was achieved purely by adjusting experimental parameters and sequence strategy, implying more gain is available from better operating-point selection.

Target Audience

Experimentalists building neutral-atom quantum processors and their supporting photonics; quantum-optics researchers interested in non-destructive readout and atom-loss budgets; engineers working on optical packaging, fiber-array coupling, and FPGA-based real-time control; and quantum-computing researchers who care about throughput and sampling rate rather than gate fidelity alone. Readers seeking a first introduction to neutral-atom hardware will find the introduction accessible, but the results assume comfort with fluorescence readout, Fisher information, and randomized benchmarking.

Authors’ abstract

Neutral-atom quantum processors have rapidly advanced in scale and coherence, yet their practical performance remains constrained by limited quantum circuit iteration rates (qCIRs) and information throughput. Here we experimentally demonstrate a high-throughput neutral-atom system based on non-destructive readout and atom reuse. By integrating a chip-based photonic interface with a 10-qubit array, we implement non-destructive readout with a retention probability of 99.7%, and further achieve a raw qCIR of 101Hz and a post-selected qCIR of 74.8Hz. More importantly, we verify a general throughput optimization methodology and obtain a normalized Fisher information rate of 57.7Hz, improving the achievable throughput by more than one order of magnitude compared with conventional methods. Our results establish a practical route toward high-throughput neutral-atom quantum processors.