Fast, continuous and coherent atom replacement in a neutral atom qubit array

Fast, continuous and coherent atom replacement in a neutral atom qubit array demonstrates that a continuously loaded reservoir near the computation zone enables on-demand atom extraction with tweezers up to 500 times per second, establishing a complete foundation for implementing fast quantum circuits with unlimited depth.

The gist

A continuously fed atom reservoir several hundred microns from the computation zone allows for repeated extraction of atoms using optical tweezers with a characteristic loading time of only 1 ms, enabling continuous reloading at rates up to 500 cycles per second, which is an improvement of nearly two orders of magnitude over the current state-of-the-art.

Core Mechanism and Reservoir Setup

The approach relies on a dense, continuously fed atom reservoir located several hundred microns from the computation zone. Atoms are guided from a 3D magneto-optical trap (MOT) in an adjoining vacuum chamber to this reservoir using an optical dipole trap (ODT) with an in-vacuum enhancement cavity, which is driven by multiple frequency tones to partially suppress standing waves. Atoms arriving at the reservoir are stopped and cooled using a 1D optical molasses. The loading time constant of the reservoir is approximately 5 seconds, but the researchers operate in the steady-state regime.

Atom Extraction and Loading Dynamics

To load optical tweezers from this reservoir, a 2D tweezer array with 256 sites is superimposed using cross acousto-optic deflectors (AODs) operating at 488 nm. The loading dynamics are governed by the equation:

dn(t)/dt = −Γatomn(t) − Dβn(t)2 + Dϕ0/Veff, where D is the molasses duty cycle and Veff is the effective volume of the molasses. Simulations show that at a reservoir atom density of n = 4 × 1011 atoms/cm3, the predicted loading probability is 0.45 at a 1 ms loading time, which closely matches experimental values.

Qubit Initialization and Loss-Resolving Readout

To use the loaded atoms as qubits, several steps are required, including light-assisted collisions (LAC) to produce singly-occupied tweezers and initialization into the metastable 6s6p 3P0 state. The sequence for qubit preparation involves:

1. Identification image to determine initial occupancy.

2. Cooling to approximately 10 µK in 6 ms.

3. Optical pumping into the 1⟩ state using a two-photon Raman transition to the 5d6s 3D1 state, followed by depumping through the 6s6d 3D2 state (a fast decay rate of τ = 24 ns).

Coherence Preservation and Gate Operations

The work demonstrates that existing qubits are completely undisturbed by the reloading process, as evidenced by comparisons between experiments with and without reloading. The lifetime, coherence time, and single-qubit gate operations on the metastable qubits are completely unaffected by subsequent rounds of reloading. Specifically, Ramsey coherence times (T∗2) were found to be 0.69(2) s with continuous reloading compared to 0.73(2) s in the control experiment. Furthermore, coherent Rabi oscillations between 0⟩ and 1⟩ exhibit no decay beyond that imposed by the 3P0 state lifetime while reloading is occurring.

Performance Metrics

The system achieves repeated preparation and measurement of arrays of singly-occupied tweezers at rates up to 30 cycles per second, with measurements reaching 50 cycles per second by reusing qubits with non-destructive measurements. The performance is characterized by high fidelity: when preparing the atom in 1⟩, the probability to record the correct outcome is 0.9644(13), rising to 0.9927(10) when loss is not included. The reservoir remains not measurably depleted by extracting atoms at a rate of 7.3 × 104 atoms/s, even when operating at the maximum extraction rate of up to 500 cycles/s with tload = 1 ms.

Conclusion and Future Directions

The results establish a solid foundation for exploring fully fault-tolerant quantum computation with neutral atom arrays, supporting zero-deadtime quantum metrology and timekeeping. These techniques can be extended to create a fully-functional quantum processor by adding rearrangement, two-qubit gates, and mid-circuit measurement. The fast atom replacement opens the door to fusion-based strategies explored in photonic quantum computing and allows for the exploration of leakage reduction strategies to preserve logical information across many cycles of atom replacement.

Appendix A: Experimental Apparatus Highlights

The apparatus includes a 3D MOT, a Zeeman slower, an ODT with an in-vacuum enhancement cavity (using 1036 nm wavelength), and a science chamber where optical molasses is applied.

Here are the specific improvements that can be made to AI systems, drawing directly from the capabilities demonstrated by this scientific paper:

1. Improving Resource Management for Deep Quantum Circuits: The paper demonstrates a method for fast, continuous and coherent atom replacement in neutral atom qubit arrays, achieving high throughput (up to 500 cycles per second) while maintaining qubit coherence.

2. Implementing Unlimited Depth Quantum Circuit Simulation/Execution: This technique removes the bottleneck of fixed circuit depth by allowing continuous mid-circuit reloading. The improved AI system can simulate or execute quantum circuits of unlimited depth, which is crucial for exploring complex, deep quantum algorithms that are currently restricted in neutral atom systems.

3. Developing High-Speed Quantum Error Correction (QEC) Protocols: The paper shows that existing qubits remain undisturbed by the reloading process, implying a foundation for fault-tolerant computation. The AI can be trained to design and optimize QEC codes (like surface codes) specifically tailored to handle the erasure errors introduced by atom loss, potentially leading to lower error correction overhead.

4. Enabling Zero-Deadtime Quantum Metrology and Timekeeping: The fast loading/imaging cycles allow for high-speed qubit manipulation. The AI can be used to design quantum metrology protocols that leverage this speed for zero-deadtime measurements, significantly enhancing the sensitivity and precision of quantum clocks and sensors.

5. Optimizing Multi-Stage Quantum Operations (Initialization/Readout): The paper details a complex sequence involving light-assisted collisions (LAC), non-destructive imaging, cooling, and optical pumping to initialize and measure qubits. An improved AI system can be used to dynamically optimize the timing, laser parameters (detuning, intensity), and pulse sequences for these sub-cycles to maximize fidelity (e.g., maximizing the correct outcome probability from 96% to over 99%).

6. Designing Adaptive Qubit Control Schemes: The use of trap intensity modulation allows for gate operations while reloading. The AI can learn optimal control sequences that dynamically adjust trap parameters based on real-time feedback (like intensity noise measurements) to maintain qubit coherence during high-frequency operational cycles.

7. Reducing Systematic Errors via Noise Characterization: The paper analyzes and attributes loss mechanisms (e.g., technical intensity noise in the transport ODT) to specific physical sources like RIN. The AI can be deployed as a sophisticated diagnostic tool to correlate observed experimental performance degradation with specific environmental noise sources, allowing for predictive error mitigation strategies.