Reducing Runtime Overhead via Use-Based Migration in Neutral Atom Quantum Architectures

TL;DR

This paper introduces strategies for neutral atom quantum architectures that dynamically adapt to atom loss, significantly reducing runtime and increasing the number of effective circuit executions before reloading.

Contribution

It develops methods to find all reachable qubits and divides the architecture into sections, enabling circuit resets without full reloads, thus improving efficiency and scalability.

Findings

Increases effective shots before reload by a factor of two for circuits using 30% of the architecture.

Reduces overall runtime by 50% for 30-qubit circuits through parallel execution.

Enhances robustness against atom loss in neutral atom quantum computing architectures.

Abstract

Neutral atoms are a promising choice for scalable quantum computing architectures. Features such as long distance interactions and native multiqubit gates offer reductions in communication costs and operation count. However, the trapped atoms used as qubits can be lost over the course of computation and due to adverse environmental factors. The value of a lost computation qubit cannot be recovered and requires the reloading of the array and rerunning of the computation, greatly increasing the number of runs of a circuit. Software mitigation strategies exist but exhaust the original mapped locations of the circuit slowly and create more spread out clusters of qubits across the architecture decreasing the probability of success. We increase flexibility by developing strategies that find all reachable qubits, rather only adjacent hardware qubits. Second, we divide the architecture into separate sections, and run the circuit in each section, free of lost atoms. Provided the architecture is large enough, this resets the circuit without having to reload the entire architecture. This increases the number of effective shots before reloading by a factor of two for a circuit that utilizes 30% of the architecture. We also explore using these sections to parallelize execution of circuits, reducing the overall runtime by a total 50% for 30 qubit circuit. These techniques contribute to a dynamic new set of strategies to combat the detrimental effects of lost computational space.