Magic state cultivation on a superconducting quantum processor

Fault-tolerant quantum computing requires a universal gate set, but the necessary non-Clifford gates represent a significant resource cost for most quantum error correction architectures. Magic state cultivation offers an efficient alternative to resource-intensive distillation protocols; however, testing the proposal’s assumptions represents a challenging departure from quantum memory experiments. We present an experimental study of magic state cultivation on a superconducting quantum processor. We implement cultivation, including code-switching into a surface code, and develop a fault-tolerant measurement protocol to bound the magic state fidelity. Cultivation reduces the error by a factor of 40, with a state fidelity of 0.9999(1) (retaining 8% of attempts). Our results experimentally establish magic state cultivation as a viable solution to one of quantum computing’s most significant challenges.

💡 Research Summary

This paper presents the first experimental demonstration of magic‑state cultivation on a superconducting quantum processor, offering a practical alternative to the traditionally resource‑intensive magic‑state distillation protocols required for fault‑tolerant non‑Clifford operations. The authors implement a full cultivation pipeline that begins with the injection of a noisy logical |T⟩ state into a distance‑3 color code, proceeds through two rounds of fault‑tolerant “phase‑kick‑back” measurements (the core of the cultivation step), and interleaves three quantum error‑correction (QEC) cycles to maintain a code distance of three throughout.

Key to the approach is a fault‑tolerant measurement of the logical Hadamard operator ˆH_L, decomposed transversally as ˆT† ˆX ˆT. The non‑conditional T‑gates are applied globally, while the conditional logical X measurement is performed via an ancilla (“helper”) qubit that also serves as a flag detector. Whenever a flag signals an error—either from the measurement itself or from the surrounding stabilizer checks—the entire attempt is aborted and restarted. This post‑selection on “no‑error” runs dramatically filters out faulty preparations.

To quantify the resulting state fidelity without the bias introduced by conventional tomography (which can over‑estimate purity in the presence of coherent noise), the authors devise a “kick‑back tomography” (KT) protocol. KT measures directly along the desired magic‑state axis (the X = Z line on the Bloch sphere), reducing projection noise and allowing a statistically significant estimate of an error rate on the order of 10⁻⁴ with a realistic number of shots (≈10⁸).

Experimental results show that after the two cultivation rounds and three QEC cycles, the logical state’s Bloch vector collapses onto the X = Z line, and the post‑selected data retain a fidelity of 0.9999 ± 0.0001 with respect to the ideal |T⟩ state. This corresponds to a 40‑fold reduction in error compared with the raw injected state. However, the stringent post‑selection discards a large fraction of attempts; only about 8 % of the total runs survive all flag checks, while the injection‑only stage retains roughly 76 % of shots.

Beyond state preparation, the work demonstrates a “code‑switching” operation that grafts the cultivated magic state from the color code into a distance‑5 surface code. This step is essential for integrating the high‑fidelity resource into larger fault‑tolerant architectures, leveraging the surface code’s high threshold, matching‑based decoders, and well‑studied leakage‑suppression techniques. The successful transfer validates the compatibility of cultivated states with leading QEC schemes and opens the path toward using cultivated magic states as direct inputs to quantum algorithms, potentially eliminating several rounds of distillation.

Control optimization plays a crucial supporting role. The authors employ reinforcement‑learning techniques to fine‑tune gate amplitudes, timings, and measurement windows, targeting a minimization of flag events across the entire circuit. This adaptive calibration, combined with a dense network of flag qubits, yields a robust error‑detection infrastructure that scales with circuit depth.

In summary, the paper delivers three major contributions: (1) a concrete, hardware‑level implementation of magic‑state cultivation on a superconducting platform; (2) a fault‑tolerant, flag‑based measurement and post‑selection framework that achieves an experimentally verified logical error rate of ~10⁻⁴ and a state fidelity of 0.9999; and (3) a demonstration of code‑switching that integrates cultivated states into a larger surface‑code logical qubit. These results substantiate the theoretical promise that cultivation can dramatically lower the overhead of non‑Clifford gate synthesis, bringing universal, fault‑tolerant quantum computation closer to practical realization.