Demonstrating advantages of dynamic quantum circuits on a hybrid superconducting qubit-cavity processor

AI summary

Overview

Research area: Experimental quantum computing — dynamic quantum circuits (mid-circuit measurement, qubit reset, and classical feed-forward) implemented on a hybrid superconducting qubit–cavity (qubit–qudit) processor.

Technical level: Advanced. The paper assumes familiarity with circuit QED, Fock-space encoding, phase estimation, and order finding, though the central argument (reuse one ancilla instead of many) is easy to grasp.

Scope in one sentence: The paper benchmarks a single 3D circuit-QED device — a transmon ancilla dispersively coupled to a long-lived microwave cavity — across three algorithms of increasing complexity (Bernstein–Vazirani, quantum phase estimation, and Shor's algorithm), showing that dynamic compilation beats prior static and dynamic implementations in scale and accuracy.

What This Paper Is About

Conventional ("static") quantum circuits spend one physical qubit per ancilla and only measure at the end, so wide circuits need many qubits and deep SWAP-heavy compilation. Dynamic circuits instead measure an ancilla in the middle of the run, reset it, and reuse it, with classical electronics feeding results forward. The catch is that this demands fast, high-fidelity, quantum-non-demolition mid-circuit readout, rapid reset, and low-latency classical control — hard to do at once. The authors' goal is to demonstrate that one hybrid qubit–cavity processor can execute a ladder of three algorithms dynamically and outperform published static and dynamic implementations of each.

Key Contributions

1. A single programmable hybrid platform used for all three algorithms. A high-dimensional cavity qudit (Fock states) serves as the computational register, while one transmon ancilla is repeatedly measured, reset, and reused — the same device runs BV, QPE, and Shor.
2. A 10-bit dynamic Bernstein–Vazirani demonstration across all 2^10 = 1024 oracle configurations, with an average success probability (squared statistical overlap, SSO) of 82%, exceeding the 5-bit dynamic implementation of Ref. [14] (maximum success probability ∼64%) and the 10-bit static implementation of Ref. [30] (average ∼78%).
3. An 8-bit dynamic quantum phase estimation protocol with estimation errors below 10^-3 for all target phases, described as nearly an order of magnitude better than the previous result in Ref. [9].
4. The first dynamic-circuit implementation of Shor's algorithm on a superconducting platform, factoring N = 15 over all six coprime bases a ∈ {2, 4, 7, 8, 11, 13} with SSO values above 99.8%, versus the 0.90–0.97 range reported for the prior ion-trap dynamic implementation [22].

Main Findings

• Device performance underpins the results. Cavity coherence: T1^C = 0.64 ms, T2^C = 1.01 ms. Transmon: T1^Q = 91 μs, T2^Q = 124 μs. Dispersive coupling χ_QC/2π = 2.59 MHz. Average readout fidelity F = 99.65% (P_g = 99.94%, P_e = 99.36%), QND fidelity F_QND = 99.09%, active reset fidelity F_reset = 97.48%, with a 700 ns feed-forward latency.
• BV scales to 10 bits with 82.0% average success. Each of the 1024 oracle instances was run 45,000 times. Success ranged from 70.9% (s = 1111111111, Hamming weight 10) to 99.6% (s = 0000000000). Every instance exceeded the standard 2/3 BQP success threshold. Post-selecting the cavity in |1⟩ improved the Hamming-weight-10 case by about 6%. Simulations identify qubit and cavity decay as the dominant error sources.
• QPE achieves sub-10^-3 phase error at 8 bits. Target phases spanned θ ∈ {0.1, 0.15, …, 1.05}, each with 45,000 circuit repetitions. Using the resource convention of Ref. [9] (one qubit measurement = one unit), R = 200 resources per estimate gives ⌊R/m⌋ = 25 repetitions of the m = 8 round circuit, and the full dataset was split into 1800 independent estimation trials. Conditional phase gates were compiled into a single virtual phase rotation.
• Shor's algorithm is faithful rather than cherry-picked. All six valid coprime bases for N = 15 were implemented (a = 14 excluded because a = N − 1 has order 2 and needs no order finding). SSO values were 0.9979(5), 0.9988(3), 0.9977(4), 0.9985(2), 0.9982(4), and 0.9983(3). Because U_a^4 = I, the first measurement round was omitted; the protocol runs three adaptive rounds with GRAPE-optimized controlled modular exponentiation.
• Residual Shor errors are traced to two sources. Numerical simulations point to imperfections in the numerically optimized pulses at higher Fock states, and to qubit decay during the feed-forward step.
• The hybrid qudit encoding is what makes this work. The cavity's long lifetime preserves algorithmic states across the measurement-and-reset sequences, while the cavity's many levels replace what would otherwise require many ancilla qubits.

Methodology in Plain English

The team built a 3D superconducting cavity with a transmon qubit attached to it, plus a separate readout resonator protected by a Purcell filter. The cavity stores quantum information in its photon-number (Fock) states — one physical object holding many levels — and the transmon acts as a single reusable helper qubit.

For each algorithm, the transmon and cavity are prepared in a simple starting state, then a cycle repeats: apply a conditional operation, measure the transmon mid-circuit, reset it based on that measurement result, and use the classical outcome to decide what to do next. Because everything is decided on the fly, gates that would normally be conditioned on many prior bits are folded into a single "virtual" phase rotation on the transmon.

For the modular-exponentiation steps needed by Shor's algorithm, the authors did not hand-build gate sequences; they used numerical optimal control (GRAPE) to find microwave pulses that map each physically reachable input state to its required output. Crucially, they restricted the optimization to the states that the feed-forward protocol can actually produce, which shrinks the search space and raises achievable fidelity. They then validated each of the six controlled operations by preparing relevant states, applying the pulse, and reading out with a photon-number-selective π pulse. Finally, they modeled the whole experiment with Lindblad master-equation simulations in QuTiP to attribute the remaining error.

Why This Matters

Impact on research. The paper argues that the standard benchmark for dynamic circuits — one impressive single task — is too weak, and demonstrates a consistent advantage across three algorithms on one chip. It also supplies concrete numbers that future dynamic-circuit experiments can be measured against, and shows that a cavity qudit plus one reusable ancilla is a genuinely hardware-efficient alternative to adding more qubits.

Real-world applications (as motivated or implied by the paper):

• Cryptographic risk assessment. Shor's algorithm on realistic scales would break RSA-style public-key encryption; faithful small-scale implementations such as this inform timelines for post-quantum migration.
• Quantum chemistry and materials simulation. The authors cite quantum chemistry and gauge-theory simulation as target applications of scalable hybrid qubit–oscillator hardware.
• Quantum metrology and sensing. Precise phase estimation, demonstrated here at 8 bits with errors below 10^-3, is the primitive behind precision measurement protocols.
• Error-corrected computing. Dynamic circuits with mid-circuit measurement and reset are a structural requirement of many quantum error-correction architectures, which the authors name as a natural next application.

Industry relevance. Bosonic (cavity) encodings are an active industrial and academic route toward hardware-efficient, error-correctable quantum processors. A demonstrated reusable-ancilla workflow with 700 ns feed-forward latency and roughly 99% QND readout is directly relevant to anyone building control electronics and real-time feedback stacks for superconducting hardware.

Future Directions

• Scaling to multiple interconnected cavities and qubits. The authors propose this as the path from a single-register demonstration to a genuinely programmable multi-mode processor, alongside further coherence-time improvements.
• Improving pulse control at high photon number. Simulations attribute part of the Shor error budget to GRAPE imperfections at higher Fock states; better calibration, drift mitigation, and pulse-distortion correction are implied targets.
• Reducing feed-forward latency and qubit decay during classical control. Qubit decay during the ~700 ns feed-forward is named as an error source in both reset fidelity and Shor performance.
• Extending the paradigm to other qubit–oscillator platforms. The authors suggest trapped ions or neutral atoms coupled to motional modes, and superconducting qubits coupled to acoustic phonons or ferromagnetic magnons.

Target Audience

Experimental quantum-computing researchers and hardware engineers working on superconducting circuits, bosonic/cavity encodings, or real-time classical control. It is also useful for theorists designing dynamic-circuit algorithms who need realistic hardware benchmarks, and for graduate students in quantum information who already know the basics of circuit QED and want a concrete, numbers-rich case study. Readers without a quantum-computing background will find the algorithmic sections demanding, though the device-reuse argument and the comparative benchmark tables remain accessible.

Authors’ abstract

Dynamic quantum circuits (DQCs) provide a hardware-efficient route to quantum computing by reducing physical-qubit overhead and compressing circuit topology through mid-circuit measurements, qubit reset and reuse, and classical feed-forward control. Here, we demonstrate the advantages of DQCs on a single hybrid superconducting qubit-cavity processor by implementing a hierarchy of algorithms with increasing complexity. This hybrid architecture consists of a high-dimensional cavity qudit serving as the computational register and a dispersively coupled superconducting transmon ancilla that is repeatedly measured, reset, and reused to enable dynamic control. Using this device, we implement a 10-bit Bernstein-Vazirani algorithm with an average success probability of 82%, surpassing state-of-the-art dynamic and static implementations in both scale and performance; an 8-bit quantum phase-estimation protocol with estimation errors below 10-3; and the first dynamic-circuit implementation of Shor's algorithm on a superconducting platform, factoring 15 over all coprime bases with squared statistical overlap values above 99.8%. These results provide concrete benchmarks for future DQC implementations and highlight the versatile advantages of DQCs with the hybrid qubit-qudit architecture, establishing it as a promising route toward scalable, programmable quantum computation.