Enhancing Circuit Fidelity in Transmon Qubit Rings via Operation Duration Tuning under Strong Connectivity Noise

Superconducting transmon qubits are a promising platform for quantum computation, yet they face significant fidelity degradation due to connectivity noise, particularly in the intermediate coupling regime where noise levels are substantial. While prior works suggest that high fidelity requires operating in regimes with strongly suppressed noise, maintaining such conditions under practical experimental constraints remains challenging. To address this, we investigate quantum gate operations in fully connected transmon rings, examining both SWAP and general circuits. Our study reveals that fidelity can be significantly enhanced by tuning gate operation durations, with local maxima emerging even under strong noise conditions. These fidelity enhancements occur consistently across different qubit numbers and operation types, and for specific initial states – particularly those with favorable symmetry or entanglement properties – the achieved fidelities approach quantum error correction thresholds. Furthermore, we develop a supervised machine learning model that accurately predicts the optimal operation durations for new devices, enabling efficient optimization without extensive experimental simulations. These results provide a pathway toward robust quantum circuit design in noisy experimental environments.

💡 Research Summary

This paper tackles the persistent problem of connectivity‑induced noise in superconducting transmon qubit processors by showing that, contrary to the common belief that “shorter gates are always better,” there exist optimal gate durations even in regimes where noise is strong. The authors focus on fully‑connected transmon rings where each qubit interacts with its nearest neighbours via cavity‑mediated exchange couplings (J) and with all other qubits via a coplanar waveguide (CPW) bus (K). Two dominant noise channels are modeled: parasitic capacitance fluctuations that scale with J (λ_J) and photon‑loss in the CPW that scales with K (λ_K). Realistic ratios R = λ_K/λ_J between 3 and 20 are used, matching experimental observations.

Using a quasi‑static Gaussian noise model, the authors perform extensive Monte‑Carlo simulations (1,500 noise realizations per parameter set) to evaluate the average state fidelity F = |⟨Ψ₀|R†U†R|Ψ₀⟩|² for a variety of circuits. First, they study sequential SWAP operations on rings of size L = 4, 6, 8, starting from a simple product state |↑↓↓…⟩. By sweeping the dimensionless coupling‑to‑noise ratio J/λ₀ from 1 to 1,000 (corresponding to gate times τ from ≈8 µs down to ≈8 ns), they discover a pronounced non‑monotonic behavior: fidelity peaks in the intermediate‑coupling regime (10 ≲ J/λ₀ ≲ 100), where τ is on the order of 80–800 ns. For L = 4, a fidelity of ≈90 % is achieved at J/λ₀ ≈ 10, despite a nominal 10 % noise level—far higher than the >99 % fidelity previously thought to require J/λ₀ ≫ 100. Larger rings still exhibit clear optimal points, although the absolute maximum fidelity modestly declines with system size.

Next, the analysis is extended to generic multi‑qubit unitaries. Random unitaries R are generated with a prescribed distance to the identity D = |⟨Ψ₀|R|Ψ₀⟩|², ensuring a broad spectrum of circuit complexities. The same noise‑sweep reveals that the optimal‑duration phenomenon is not specific to SWAP gates; any sufficiently complex operation displays a fidelity maximum in the same intermediate‑coupling window.

State‑dependence is examined by comparing three representative initial states: (i) a product state, (ii) a single‑pair singlet embedded in a product background, and (iii) a GHZ‑type maximally entangled state. The GHZ state consistently yields the highest fidelities at the optimal points, reaching ≈99.9 %—the typical threshold for surface‑code error correction. This highlights the role of symmetry and global entanglement in mitigating the impact of connectivity noise.

A scaling analysis shows that if all noise amplitudes are multiplied by a factor α, the optimal gate time simply rescales as τ_opt ∝ 1/α, leaving the fidelity curve versus J/λ₀ unchanged. This scale‑invariance suggests that the identified optimal points are robust against modest variations in device fabrication.

Finally, to make the approach practical for experimentalists, the authors train a supervised machine‑learning regressor (e.g., random‑forest or shallow neural network) on a modest dataset of simulated optimal durations. Input features include the coupling‑to‑noise ratio J/λ₀, ring size L, initial‑state class, and the noise‑ratio R. The model predicts τ_opt with mean‑square error below 1 % across a test set of unseen device parameters, effectively eliminating the need for exhaustive Monte‑Carlo sweeps for each new chip.

In summary, the paper establishes a new design principle for transmon‑based quantum processors: by deliberately selecting gate durations in the intermediate‑coupling regime, one can achieve fidelity peaks that rival or surpass those obtained only in ultra‑strong‑coupling, low‑noise regimes. This principle holds for both elementary SWAP primitives and arbitrary multi‑qubit circuits, is enhanced for highly symmetric entangled states, and can be efficiently exploited using a lightweight machine‑learning predictor. The results open a realistic pathway toward fault‑tolerant quantum computation on near‑term superconducting hardware without demanding extreme suppression of connectivity noise.