Quantum computing promises to revolutionize various fields by tackling problems intractable for classical computers. However, realizing this potential requires overcoming significant challenges, primarily the high error rates in quantum operations. For practical quantum computation, achieving sufficiently low error rates is essential, particularly for two-qubit gates, which are fundamental building blocks of quantum algorithms. Error rates below 1% (fidelities above 99%) are needed to surpass the thresholds for quantum error correction, a crucial technique for building fault-tolerant quantum computers. Moreover, scaling these systems to a large number of qubits while maintaining low error rates, high parallelism, and strong connectivity is paramount. Various platforms are being explored for quantum computation, each with unique advantages and challenges. Superconducting circuits and trapped ions have demonstrated impressive progress, achieving high-fidelity entangling operations on a limited number of qubits. Scaling these systems, however, presents significant technical hurdles. Neutral atom arrays, on the other hand, offer the potential for scalability and high connectivity due to their inherent flexibility and addressability of individual atoms. They are also capable of maintaining coherent control over large numbers of qubits, suitable for both analogue and digital quantum simulations. Although neutral atoms have shown promising results, the challenge has been to improve the fidelity of two-qubit gates significantly beyond the previously demonstrated ~97.5%. This paper aims to address this challenge by presenting an experimental realization of high-fidelity entangling gates on a neutral atom quantum computer, pushing the boundaries of fidelity and scalability in this promising platform.

Previous work in quantum computing has demonstrated impressive progress in different platforms. Superconducting circuits and trapped ions have achieved high-fidelity entangling operations but face scalability challenges. Neutral-atom arrays, while promising for scalability and high connectivity, lagged behind in two-qubit gate fidelity, with previously demonstrated values around 97.5%. This paper builds on prior research in neutral-atom quantum computing, particularly work on Rydberg-mediated interactions and parallel gate implementations, but seeks to significantly improve gate fidelity to surpass the error-correction threshold and open avenues for large-scale quantum computations.

The researchers employed neutral rubidium atoms trapped in individual optical tweezers, encoding quantum information in long-lived hyperfine qubits. High-fidelity single-qubit rotations were achieved using Raman laser pulses. Entangling operations were performed in parallel by arranging atoms into designated gate sites and employing Rydberg-blockade interactions via state-selective excitation into highly excited Rydberg states. The core innovation is the use of fast, single-pulse entangling gates optimized through numerical methods. These optimal control schemes, inspired by previous work on time-optimal gates, utilized a continuous phase profile for a single laser pulse, eliminating the need for discrete phase jumps. A family of single-pulse gates was designed with tunable parameters, including a time-optimal gate and a smooth-amplitude gate. The smooth-amplitude gate was designed to minimize population in the “bright” dressed state containing the short-lived intermediate state, thereby suppressing scattering errors. The system was improved through the use of a lower-lying Rydberg state (n=53) excited by a higher power laser and A-enhanced grey molasses cooling to improve the atomic temperature. Gate fidelity was characterized using several complementary benchmarking methods: Bell-state fidelity measurements, which showed raw fidelity above 98%; a method involving an odd-numbered sequence of CZ gates interleaved with single-qubit gates to extract the CZ gate fidelity; and global randomized benchmarking with random single-qubit rotations between CZ gates, providing independent estimates of fidelity. The experiments were performed on arrays of up to 60 qubits, leveraging larger Rydberg beams to maintain sufficient intensity across a wider area. To understand scaling limitations, detailed modelling of the system was performed considering Rydberg decay, coupling to other Rydberg states, intermediate-state scattering, and ground-Rydberg T2 times. Furthermore, the researchers extended their methods to realize three-qubit entangling gates by employing time-optimal CCZ gate pulse profiles and characterizing their fidelity.

The researchers achieved a significant breakthrough by experimentally demonstrating two-qubit controlled-Z (CZ) gates with a fidelity of 99.5%, a value exceeding the surface-code threshold for fault-tolerant quantum computation. This fidelity was consistently observed across multiple benchmarking methods, confirming the robustness of the results. The high fidelity was achieved through a combination of optimized gate schemes and improvements in experimental techniques. The use of fast, single-pulse gates with continuous phase modulation proved crucial, along with measures to suppress spontaneous scattering from the intermediate atomic state. Remarkably, this high fidelity was maintained even when scaling the system to 60 qubits in parallel. Analysis of error sources revealed that Rydberg decay, coupling to other Rydberg states, intermediate-state scattering, and ground-Rydberg decoherence (T2) were the dominant contributors. The error analysis also showed that correlated errors were largely absent, indicating the potential for further scaling. The methodology was generalized to create three-qubit entangling gates with a fidelity of 97.9%, significantly advancing capabilities for multi-qubit operations. The time required for the CCZ gate was only 44% longer than the optimal CZ gate, highlighting efficient scaling of multi-qubit gate implementation. The authors demonstrated that the gate fidelity was consistent across multiple gate sites within the array, supporting the scalability of their approach. The analysis of microscopic error sources suggests that maintaining laser power and beam homogeneity are the primary challenges for further scaling to even larger numbers of qubits.

The results presented in this study represent a significant advance in neutral-atom quantum computing. The achievement of two-qubit gates with fidelities exceeding 99.5% on up to 60 qubits in parallel surpasses the threshold required for quantum error correction, paving the way for the construction of fault-tolerant quantum computers. This progress addresses a major bottleneck in the field, where achieving such high fidelities at scale has been a significant hurdle. The detailed analysis of error sources provides valuable insights for future improvements, such as increasing laser power and maintaining beam homogeneity for scaling to even larger systems. The extension to three-qubit gates, with a fidelity approaching 98%, further demonstrates the versatility and scalability of the approach. The consistency of fidelity across multiple gate sites highlights the inherent scalability of the methodology, suggesting that increasing the number of qubits will not necessarily lead to an exponential increase in calibration overhead. This research not only advances the development of fault-tolerant quantum computers but also opens exciting opportunities for various quantum algorithms, digital quantum simulations, and other applications that benefit from high-fidelity multi-qubit entanglement.

This research successfully demonstrates high-fidelity (99.5%) parallel two-qubit entangling gates in a neutral-atom quantum computer, exceeding the surface code threshold for error correction. The methodology extends to three-qubit gates with good fidelity and shows potential for scaling to larger numbers of qubits. Future work should focus on further increasing laser power, improving beam homogeneity, and suppressing specific error mechanisms to achieve even higher fidelities and larger scale.

While the achieved fidelities are remarkably high, the study's limitations include the experimental challenges associated with maintaining laser power and beam homogeneity across a large array of qubits. This is the primary obstacle identified for further scaling. Furthermore, the three-qubit gate fidelity, while good, is slightly lower than the two-qubit fidelity. More sophisticated benchmarking methods for the multi-qubit gates could provide a more precise characterization of their performance.