Session M52: Quantum Error Mitigation at Scale

Dynamic in-situ mitigation of time-varying noise in magic state factories

Future fault-tolerant quantum computers must maintain stability over long periods of time to carry out complex calculations. However, existing quantum devices exhibit high levels of instability, caused by phenomena including slow drift over time, two level system (TLS) defect resonances, and disruptive cosmic ray impacts. If unaddressed, these issues could dramatically limit the runtime of future quantum algorithms. Our key insight is that physical error rates can be monitored by observing auxiliary data produced by error correction circuits. We propose a series of dynamic mitigation strategies, focusing on magic state distillation protocols, which can continually stabilize the device while allowing the computation to continue uninterrupted.   

Noise-Resilient Quantum Simulation with the Quantum Error Detection Code

Quantum computing promises substantial speed-up in simulating physical systems, but noise in near-term quantum processors prevents us from fully realizing its power. In this work, we propose a procedure to reduce errors in quantum simulations on current and near-term quantum computers using the [[n,n-2,2]] quantum error detection code. We employ a weakly fault-tolerant construction and develop systematic methods to construct a logical exponential map for a general Pauli operator. Used together with mid-circuit syndrome measurements, this method can detect errors in deep circuits consisting of a sequence of exponential maps, as generally occur in dynamical or molecular system simulations using Trotter–Suzuki formulas. Our work suggests that one can use elements of fault-tolerance to reduce noise in practical near-term quantum computers.   

Virtual projection onto the code space for rotation-symmetric bosonic codes

Bosonic quantum error correction codes offer hardware-friendly strategies for error correction because they can use the infinite dimension of Hilbert space of harmonic oscillators. In this talk, we give a framework for virtually projecting the noisy state quantum state for rotation symmetric bosonic codes, e.g., cat codes and binomial codes in an error mitigation manner. Our work allows us to compute the same expectation value in a much more hardware-friendly way than the case in which we apply quantum error detection for rotation symmetric codes. Our framework paves the way towards high-accuracy bosonic quantum computing.   

Virtual entanglement distillation with quantum error mitigation

A promising approach for scalable quantum computing involves distributed quantum computation, where small-scale remote devices are connected through entanglement typically generated using a photon interference effect. Since entanglement is susceptible to noise, entanglement distillation is often used to mitigate the impact of noise. However, this process frequently demands substantial resources and computational time. An alternative solution is a quasi-probability simulation using Schmidt decomposition with local operations and classical communication (LOCC), although this approach incurs significant sampling overheads.

In this work, we propose a complementary method that leverages quantum error mitigation through LOCC and classical post-processing to suppress errors in noisy entanglement. We introduce two methods: one is error-agnostic, based on virtual quantum error detection, and the other is error-aware, relying on probabilistic error cancellation.

Our findings demonstrate that the sampling overhead of our proposed methods is lower than that of Schmidt decomposition within a practical error range of noisy entanglement states.   

One-Way Hashing Method with Finite Resources

The one-way hashing method refers to a well-known entanglement distillation protocol that can be used to transform multiple copies of a bipartite state into high fidelity Bell pairs, using only local operations and classical communication. While it is known that this protocol is very effective at distilling entanglement from large-scale systems, previous results indicated that the one-way hashing method may not be useful when the number of initial copies is small. For this reason, the protocol has not yet been realized by any experimental setup. By leveraging properties of the Hartley entropy, we provide significantly improved analytical lower bounds on this protocol's distillation rate, as well as numerical simulations, which demonstrate that entanglement can be distilled via the one-way hashing method at a higher rate and using fewer initial copies than previously expected. These results show that the one-way hashing method is not only of interest for future large-scale quantum networks; it is also a viable option for distilling entanglement on state-of-the-art quantum technologies.   

Zero noise extrapolation on logical qubits by scaling the error correction code distance

In this work, we migrate the quantum error mitigation technique of Zero-Noise Extrapolation (ZNE) to fault-tolerant quantum computing. We employ ZNE on logically encoded qubits rather than physical qubits. This approach will be useful in a regime where quantum error correction (QEC) is implementable but the number of qubits available for QEC is limited. Apart from illustrating the utility of a traditional ZNE approach (circuit-level unitary folding) for the QEC regime, we propose a novel noise scaling ZNE method specifically tailored to QEC: distance scaled ZNE (DS-ZNE). DS-ZNE scales the distance of the error correction code, and thereby the resulting logical error rate, and utilizes this code distance as the scaling `knob' for ZNE. Logical qubit error rates are scaled until the maximum achievable code distance for a fixed number of physical qubits, and lower error rates (i.e., effectively higher code distances) are achieved via extrapolation techniques migrated from traditional ZNE. Furthermore, to maximize physical qubit utilization over the ZNE experiments, logical executions at code distances lower than the maximum allowed by the physical qubits on the quantum device are parallelized to optimize device utilization. We validate our proposal with numerical simulation and confirm that ZNE lowers the logical error rates and increases the effective code distance beyond the physical capability of the quantum device. For instance, at a physical code distance of 11, the DS-ZNE effective code distance is 17, and at a physical code distance of 13, the DS-ZNE effective code distance is 21. When the proposed technique is compared against unitary folding ZNE under the constraint of a fixed number of executions of the quantum device, DS-ZNE outperforms unitary folding by up to 92% in terms of the post-ZNE logical error rate.   

Reducing Quasiparticle-Induced Dissipation Using Gap-Engineering

Superconducting qubits based on Josephson junctions are a leading platform to implement universal quantum computing. However, non-equilibrium quasiparticles present challenges to scaling superconducting qubits because they can both degrade qubit lifetimes and generate correlated errors that compromise quantum error correction. To date, most approaches have focused on mitigating the generation of non-equilibrium quasiparticles, via shielding and filtering. In this talk, I will discuss a complementary approach that uses gap-engineering to mitigate the impact of non-equilibrium quasiparticle on chip. In particular, I will discuss progress toward the fabrication and measurement of transmon qubits with Josephson junctions that have spatially varying superconducting gap profiles in the leads.   

Quasiparticle poisoning in superconducting qubit arrays from controlled gamma irradiation

When high-energy radiation, such as a gamma ray or muon, impacts a superconducting qubit chip, large numbers of electron-hole pairs and phonons are created. The ensuing dynamics of the electrons and holes changes the local offset charge environment for qubits near the impact site. The phonons that are produced have energy above the superconducting gap in the device layer, leading to excitations above the superconducting ground state known as quasiparticles. An elevated density of quasiparticles degrades qubit coherence, leading to errors in qubit arrays. Because these pair-breaking phonons spread throughout much of the chip, these errors can be correlated across a large portion of the array, posing a significant challenge for error-correction schemes. In order to study the dynamics of gamma ray impacts on superconducting qubit arrays, we use a gamma-ray source outside the dilution refrigerator to controllably dose our chip within. By using charge-sensitive transmon qubits, we can measure quasiparticle poisoning and offset charge jumps due to the gamma irradiation at different doses.   

Characterization of thermalization timescales in superconducting qubits using quasiparticle poisoning rates

Quasiparticles, excitations above the superconducting ground state, can lead to errors on superconducting qubits. Various measurements of quasiparticle poisoning of superconducting circuits report slow decays of the poisoning rates following the initial cooling of the low-temperature cryostat. We use weakly charge-sensitive transmons to measure quasiparticle poisoning in qubit arrays. We track these poisoning rates as a function of time after the initial cooldown to characterize the thermalization timescales of our chips. We compare different methods of mounting the devices in our sample boxes, such as adhesive (GE varnish), wire bonds, or machined aluminum clamps.   

Multi-qubit correlated quasiparticle transitions in large array of tantalum transmon qubits: Part I

Quantum error correction is a crucial aspect in realizing a universal quantum computer. Ensuring errors are uncorrelated in both space and time is a significant requirement. However, recent studies have revealed correlated errors, such as charge noise, two-level-system drift, and energy relaxation in superconducting qubits, induced by cosmic rays. This work focuses on observing correlated errors related to quasiparticle transitions in a large array of tantalum transmon qubits. The results demonstrate that multi-qubit correlated quasiparticle transition events are more frequent than multi-qubit correlated energy relaxation events, affirming their association with quasiparticle bursts. Interestingly, the average rate of multi-qubit correlated energy relaxation events in tantalum qubits was found to be smaller than that in aluminum qubits, possibly attributed to the stronger electron-phonon interaction in tantalum. Part I of the presentation emphasizes the experimental design, setup, and calibration.   

Multi-qubit correlated quasiparticle transitions in large array of tantalum transmon qubits: Part II

Quantum error correction is a crucial aspect in realizing a universal quantum computer. Ensuring errors are uncorrelated in both space and time is a significant requirement. However, recent studies have revealed correlated errors, such as charge noise, two-level-system drift, and energy relaxation in superconducting qubits, induced by cosmic rays. This work focuses on observing correlated errors related to quasiparticle transitions in a large array of tantalum transmon qubits. The results demonstrate that multi-qubit correlated quasiparticle transition events are more frequent than multi-qubit correlated energy relaxation events, affirming their association with quasiparticle bursts. Interestingly, the average rate of multi-qubit correlated energy relaxation events in tantalum qubits was found to be smaller than that in aluminum qubits, possibly attributed to the stronger electron-phonon interaction in tantalum. Basing ourselves on Part I, Part II we peresent the efficience of the quasiparticle transitions as detectors of burst correlated errors, show the result of errors mitigation in tantalum film and signals of burst correlated quasiparticle transitions.   

Protecting Superconducting Qubits from Quasiparticle Bursts with Phononic Filters

The investigation and mitigation of decoherence mechanisms in superconducting qubits is critical for the development of superconducting quantum computers. One known source of decoherence is quasiparticle tunneling in Josephson junctions [1]. Experiments have demonstrated the generation of quasiparticles can in part be attributed to ionizing radiation interacting with the sample substrate and producing high energy phonons capable of breaking cooper pairs [2,3]. The isolation of qubits from such phonons could enable the suppression of this decoherence mechanism. To investigate this we design, fabricate, and measure transmon qubits on nanomechanical resonators acting as acoustic low-pass filters. This allows us to study phonon-induced quasiparticle generation rates of on-chip acoustically shielded qubits. We anticipate this technique to be beneficial towards the development of highly coherent superconducting qubits and large scale qubit arrays.

[1]: G. Catelani, S. E. Nigg, S. M. Girvin, R. J. Schoelkopf, and L. I. Glazman. Phys. Rev. B 86, 184514 (2012)

[2]: McEwen, M., Faoro, L., Arya, K. et al. Resolving catastrophic error bursts from cosmic rays in large arrays of superconducting qubits. Nat. Phys. 18, 107–111 (2022).

[3]: Vepsäläinen, A.P., Karamlou, A.H., Orrell, J.L. et al. Impact of ionizing radiation on superconducting qubit coherence. Nature 584, 551–556 (2020).