Mitigating cosmic ray-like correlated events with a modular quantum processor

TL;DR

CRL events cause transient $T_1$ suppression in superconducting qubits, challenging conventional error correction. The authors implement a modular quantum processor with two flip-chip daughterboards connected to a common router to localize phonon/QP poisoning and measure correlations. They find strong intra-module correlations (~85%) and weak inter-module correlations (~2%), with router-related detections linking to CRL drive paths; qubit recovery times are qubit-specific ($\overline{\tau}$ on the order of a few milliseconds). The results support a mitigation path where modular architectures, combined with higher-level QEC codes, can protect distributed quantum information from chip-scale CRL events, advancing scalable quantum computation.

Abstract

Quantum processors based on superconducting qubits are being scaled to larger qubit numbers, enabling the implementation of small-scale quantum error correction codes. However, catastrophic chip-scale correlated errors have been observed in these processors, attributed to e.g. cosmic ray impacts, which challenge conventional error-correction codes such as the surface code. These events are characterized by a temporary but pronounced suppression of the qubit energy relaxation times. Here, we explore the potential for modular quantum computing architectures to mitigate such correlated energy decay events. We measure cosmic ray-like events in a quantum processor comprising a motherboard and two flip-chip bonded daughterboard modules, each module containing two superconducting qubits. We monitor the appearance of correlated qubit decay events within a single module and across the physically separated modules. We find that while decay events within one module are strongly correlated (over $85\%$), events in separate modules only display $\sim 2\%$ correlations. We also report coincident decay events in the motherboard and in either of the two daughterboard modules, providing further insight into the nature of these decay events. These results suggest that modular architectures, combined with bespoke error correction codes, offer a promising approach for protecting future quantum processors from chip-scale correlated errors.

Figures (11)

• Figure 1: Layout and measurement pulse sequence. (a) Schematic for the modular quantum processor. Two daughterboards (DB), labeled A (B), each supporting two qubits $Q_{1,3}$ ($Q_{2,4}$), are flip-chip bonded with soft polymeric materials to the motherboard, which supports the router $R$ with its four switches $S_j$ as well as control and readout circuitry. Qubits on each daughterboard are galvanically separate from each other, as indicated by the solid red lines in the middle of both red shaded areas. (b) Schematic cross section of assembly (not to scale), with superconducting Al films (grey), polymer standoffs (dark blue), and sapphire substrates (light blue). A high-energy particle (orange) creates phonons (purple) in the sapphire substrates, which in turn generate quasiparticles (red) in the superconducting films. (c) Pulse sequence to detect CRL events in a qubit: We repeatedly apply a 50 ns $\pi$ pulse to the qubit, then following a delay $t_{\rm idle} = 1~\mu$s, measure the qubit during $t_{\rm meas} = 800$ ns. After a wait time $t_{\rm wait} = 8.15~\mu$s, the sequence is repeated, with a total cycle time of $10~\mu$s. (d) Pulse sequence to detect CRL events in the motherboard, by swapping an excitation from $Q_2$ into and out of the router via switch $S_2$ during $t_{\mathrm{swap}}=3.6~\mathrm{ns}$ with an intermediate wait time $t_{\mathrm{idle}}=1~\mu\mathrm{s}$, then measure $Q_2$. This is done in parallel with a CRL-detecting sequence on other qubits (here $Q_4$).
• Figure 2: Short sequences of binned data for qubits $Q_2$ and $Q_4$. (a), (b) Decay probabilities for $Q_2$ and $Q_4$ averaged in $N_b = 100$ cycles. (c), (d) Detailed time dependence for coincident events at $t = 3~\mathrm{min}$ (shaded in (a) and (b)), with an exponential fit yielding the event recovery time $\tau$. (e) Cumulative distribution of event recovery times for qubits $Q_{1, 2, 4}$; shaded regions represent the fit uncertainty of one standard deviation. Vertical dashed lines mark the mean values, with $\overline{\tau_{Q_1}} = 3.7 \pm 0.1$ ms, $\overline{\tau_{Q_2}} = 2.2 \pm 0.1$ ms, and $\overline{\tau_{Q_4}} = 6.2 \pm 0.1$ ms.
• Figure 3: Scatter plots of two-qubit joint decay probabilities averaged over $N_b=100$ measurement cycles. (a), (b) and (c) Data for qubits $Q_2-Q_4$, $Q_1-Q_2$, and $Q_1-Q_4$ respectively, where $Q_2$ and $Q_4$ are located on the same daughterboard module while $Q_1$ is on the other daughterboard. (d) and (e) Scatter plots including decay probabilities for $Q_2$ serving as an ancilla for the motherboard router $R$, together with $Q_4$ and $Q_1$, respectively. (f) $Q_1-Q_4$ coincident data taken simultaneously with data in panels (d) and (e). In all panels, points close to the diagonal indicate correlated increases in decay probabilities in two qubits while points near the axes indicate decay probability increases for a single qubit or the router.
• Figure 4: Event statistics. (a), (b): Short sequences of bit string data, post-processed to identify a decay error $ggg$ as $1$, otherwise as $0$, taken for a coincident CRL event in $Q_2$ and $Q_4$. The convolution of a template contrast function with these data is shown in red (see main text and Appendix \ref{['sec:start_times']}), allowing identification of the event start time, indicated by a green (purple) point in (a) and (b). By using the threshold technique described in Appendix \ref{['sec:pp-filter']}, we identify all peaks in the bit string data for each qubit and for the qubit-router measurements. We tabulate in (c) and (d) the total number of events in each qubit on the table diagonals and the number of coincident events between different qubits in the off-diagonal elements, where events are classified as coincident if they occur within $1$ ms of each other. (e), (f): Ratio of the number of coincident events in $Q_{\mathrm{row}}$ given an event in $Q_{\mathrm{col}}$, extracted from panels (c) and (d), respectively.
• Figure 5: Router $T_1$ measurement. We measure the lifetime of the router by performing qubit-coupler swaps and varying the delay between two swap operations, as shown inset. A fit to an exponential decay yields the router $T_1=10.1~\mu\mathrm{s}$.