arXiv:2610.38148
Experimental Observation of Error-Resilience Phase Transition on Superconducting Quantum Processor
Emanuel Dallas, Mario Morford-Oberst, Eric Weine
Abstract
Random unitaries scramble quantum information into highly nonlocal degrees of freedom, thereby providing protection from local noise. However, as demonstrated by Turkeshi and Sierant (Phys. Rev. Lett. 132, 140401, 2024 and arXiv:2603.00235), this protection may fail once local noise grows sufficiently strong. These works describe a phase transition in a noisy random encoding-decoding circuit model between an error-protecting phase, in which encoded information is recoverable, and an error-vulnerable phase, in which encoded information is lost, with the transition occurring at a critical coherent noise angle $α_c \approx 1.14$. Here, we present the first experimental observation of this transition, conducted on IBM's ibm_pittsburgh Heron r3 device for system sizes of up to $N = 26$ qubits. We detect this transition via three order parameters: fidelity, Rényi-2 entanglement entropy, and stabilizer Rényi-2 entropy. For the latter two, we employ randomized measurement protocols to construct estimators, and for all three, we rely on zero-noise extrapolation and M3 SPAM error mitigation. We then jointly model the finite-size scaling collapse across all three order parameters with a Gaussian process regression, yielding a critical angle $α_c \approx 1.01$, well below the theoretical value. Inputting noiseless simulated data into the same model yields $α_c \approx 1.05$, implying that the discrepancy between experiment and theory is due to our encoding circuits, which are not deep enough to be truly Haar-random. Finally, we observe that the same model cannot adequately identify critical behavior in unmitigated, raw experimental data. This indicates that error mitigation techniques are crucial for detecting the transition on hardware.
AI Summary
The paper investigates whether the predicted error-resilience transition in random encoding–decoding circuits can be detected on superconducting hardware. The protocol encodes an initial product state, applies identical coherent $Z$ rotations of angle $\alpha$, decodes, and post-selects ancillas in their initial state. For Haar-random encoding at logical rate $k/N=1/2$, theory predicts a transition at $\alpha_c\approx1.14$ from recovery to vanishing logical fidelity, accompanied by extensive entanglement and non-stabilizerness. Experiments on IBM’s Heron r3 processor use depth-$N$ brickwork encoders assembled from native gates, with four encoding instances per size. Fidelity is measured up to $N=26$, subsystem Rényi-2 entropy up to $N=20$, and stabilizer Rényi-2 entropy up to $N=18$. Randomized measurements estimate the entropy observables; processing combines M3 readout mitigation, quadratic zero-noise extrapolation, and normalization against $\alpha=0$. The experiments use approximately $5.07\times10^8$ shots. A joint Gaussian-process finite-size-scaling fit gives an effective $\alpha_c\approx1.01$; noiseless simulations of the implemented ensemble give approximately $1.05$, with overlapping 95% credible intervals. Applying the same fit to unmitigated data does not consistently identify critical behavior. The stated evidence consists of finite-size crossings, scaling collapses, and agreement with simulations, rather than a hypothesis test establishing a transition. The encoders are not Haar-random, and the fitted threshold is not identified with the thermodynamic-limit value. Post-selection probability decreases exponentially with size, limiting precision. Hardware states remain mixed: subsystem entropy is therefore not strictly entanglement entropy, and the global-purity correction to stabilizer entropy is omitted.
Central claim may not hold: The experiment establishes finite-size signatures consistent with an assumed scaling ansatz, rather than independently establishing the thermodynamic transition or its critical angle. The experimental reduced-state entropy is not an entanglement measure for the acknowledged mixed logical states, and the experimental magic estimator omits the unmeasured full-state purity correction required by the theoretical definition. The GP likelihood omits cross-angle correlations arising from reused random encodings and shared estimated normalization baselines, leaving its posterior uncertainty inadequately justified. The stated annealed-average definition gives asymptotically zero purity-corrected stabilizer entropy at alpha=pi, contradicting the extensive value in the paper's quoted formula. Agreement between experimental and noiseless fitted angles supports an encoding-ensemble explanation but does not uniquely establish insufficient circuit depth as the cause of the discrepancy from Haar theory.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.