Skip to content
Quantum.world

Research

Implementing rapid reset of superconducting qubits with on-chip dissipation

On-chip dissipative reset reaches a 48 ns time constant and sub-1% residual excitation, giving superconducting QEC architects a scalable, Purcell-protected reset without warmer external circuitry.

Implementing rapid reset of superconducting qubits with on-chip dissipation
arXiv
2611.14391
Published
2026-11-10
Authors
Feiyu Li, Yang Liu, Zhen Yang, Zhenxing Liu, Jiayu Ding, Yang Gao, Xuehui Liang, Weijie Sun, Yanjun Ma, Chuhong Yang, Guangming Xue, He Wang, Fei Yan

AI summary

Overview

Research area: Superconducting quantum computing hardware — specifically qubit initialization (reset) and leakage removal via engineered dissipation in circuit QED.

Technical level: Intermediate. The paper assumes familiarity with transmon qubits, relaxation time T1, the Purcell effect, and resonator-mediated decay, but the device concept itself is simple enough to be grasped without deep theory.

Scope (1 sentence): The paper experimentally demonstrates a scalable on-chip reset architecture in which each transmon is coupled to a dedicated dissipative resonator that dumps excitation into shared on-chip palladium–gold resistive loads, achieving a median reset time constant of 48.1 ns across 20 qubits on a 72-qubit processor.

What This Paper Is About

Quantum algorithms and error-correction cycles need qubits returned to a known ground state quickly and reliably, and they also need leakage out of the computational subspace (the |f⟩ state) removed. Conventional superconducting reset schemes route dissipation to off-chip, higher-temperature electronics, which exposes qubits to thermal noise and adds wiring and cryogenic-port complexity. This paper shows that the dissipative element can instead be placed directly on the chip as a thin-film resistor, giving fast reset while keeping qubits isolated from warmer circuitry.

Key Contributions

  1. A scalable on-chip dissipation architecture. Each qubit is capacitively coupled to a dedicated λ/4 tantalum dissipative resonator; multiple dissipators couple to a single shared reset transmission line terminated at both ends by impedance-matched palladium–gold (PdAu) thin-film resistors. No reset-specific cryogenic ports or control lines are required — the dissipator operates passively.

  2. A validated lumped-element design model. The authors derive an expression for the dissipator decay rate, κr = 2ωr³M²/(πZrR0), and cross-check it against finite-element electromagnetic simulation: at the design point (ωr/2π = 3.7 GHz, M = 140 pH, R0 = 50 Ω) the analytical model gives κr/2π = 10.0 MHz versus 9.7 MHz from simulation.

  3. Experimental reset across a multi-qubit processor. On 20 operational qubits of a 72-qubit test chip, they measure a median reset time constant of 48.1 ns, reduced from an off-resonant T1 of 24.6 μs, and extract median g/2π = 5.5 MHz and κr/2π = 11.0 MHz, placing the system in the strong-coupling regime (|g| > κr/4).

  4. Single-shot reset and leakage-removal benchmarks. A 400-ns reset pulse yields median residual excited-state populations of 0.29% and 0.67% for qubits initialized in |g⟩ and |e⟩ respectively (median reset efficiency 99.28%), and a 400-ns flux pulse reduces the residual |f⟩ population on Q0 to 0.51%.

Main Findings

  • Fast reset: The median relaxation time drops from T1^off = 24.6 μs off resonance to T1^reset = 48.1 ns during reset, measured across 20 qubits.
  • Residual populations after a 400-ns pulse: Median P(e|g) falls from 0.70% (25th–75th percentile 0.52–0.89%) to 0.29% (0.17–0.46%), and median P(e|e) falls from 94.74% (92.88–95.57%) to 0.67% (0.47–1.75%).
  • Representative qubit (Q0): Single-shot readout with 8192 shots per prepared state shows P(e|g) dropping from 1.39% to 0.18% and P(e|e) from 95.78% to 0.37%, using a discrimination threshold fixed without reset.
  • Effective bath temperature of 30 mK: Interpreting the assigned fractions as thermal populations, P_reset(e|g) = 0.29% evaluated at ωr/2π = 3.724 GHz gives Teff = 30.6 mK, versus 46.2 mK inferred without reset at ωidle/2π = 4.768 GHz. The refrigerator base temperature is approximately 10 mK.
  • Purcell protection works: Beyond detunings typically greater than 500 MHz, relaxation time becomes nearly frequency independent, indicating suppressed dissipator-induced Purcell decay during off-resonant operation. Idle (4.768 GHz) and maximum (4.799 GHz) qubit frequencies sit well above the dissipator median of 3.724 GHz.
  • Leakage removal from |f⟩: On Q0, a 400-ns flux pulse bringing the |f⟩↔|e⟩ transition into resonance with the dissipator reduces P(f|f) from 77.54% to 0.51%; the accompanying increase in P(g|f) from 3.52% to 18.36% may reflect additional relaxation during the pulse sequence.
  • Design tolerance: The lumped model predicts κr/2π stays within 8.3–12.5 MHz for termination resistances within ±20% of the nominal 50 Ω.
  • Measured coupling below design: The median extracted g/2π = 5.5 MHz is lower than the 8.5 MHz design value (corresponding to Cqr = 1.2 fF), possibly due to fabrication variation in the small coupling capacitance or to systematic extraction uncertainties.
  • A route to better frequency selectivity: Within the model, reducing g/2π from 8.5 to 2.5 MHz would reach critical damping (4g = κr) and narrow the detuning range over which 1/κq falls below 1 ms from |Δ|/2π ≲ 2.1 GHz to ≲ 0.63 GHz.

Methodology in Plain English

The team built a "dissipation unit" next to each qubit: a quarter-wavelength microwave resonator that is deliberately made lossy. Instead of letting that resonator leak into external coax cables, they inductively couple it to a short on-chip transmission line whose two ends are capped with matched 50 Ω resistors — 45-nm-thick PdAu alloy films (1:1 Pd:Au by mass), deposited by thermal evaporation after ion milling away the native oxide on the tantalum surface. Because the line is impedance matched, the resonator sees a well-defined resistive load and decays at a designed rate, while the lossy metal stays spatially separated from the qubits. The chip's reset line is roughly 19 mm long and terminates six dissipators, so the architecture shares hardware across qubits.

Operation is passive until needed: the dissipator frequency is placed below the qubit's operating frequency, and a flux pulse tunes the qubit down into resonance to switch reset on. To characterize the scheme, they prepare |e⟩ at the idle frequency, apply a flat-top flux pulse with cosine edges (1 ns each, 2 ns total) that parks the qubit at a variable frequency for a variable hold time, and fit exponential decays to extract T1 as a function of qubit frequency. Fitting the resulting T1(ωq) curve with a resonator-induced decay model yields g and κr for each qubit. They spread measurements over four non-adjacent groups of qubits to suppress crosstalk, and use single-shot readout with fixed discrimination thresholds (and, for the three-state leakage measurement, a spherical Gaussian mixture model trained without leakage removal and reused unchanged) so that before/after comparisons are not biased.

Why This Matters

Reset speed is a direct limit on error-correction cycle time and on how fast experiments can be repeated; a 48.1 ns time constant is fast relative to typical qubit coherence and removes a bottleneck that usually requires extra room-temperature electronics. Just as importantly, the dissipation source lives on the chip, so scaling to many qubits does not mean scaling up cryogenic ports and wiring — a practical constraint in real processors.

Applications implied by the work:

  • Quantum error correction: Fast, high-fidelity initialization and leakage removal are prerequisites for repeated syndrome-extraction cycles; the paper cites surface-code experiments where such protocols helped suppress logical error rates with increasing code distance.
  • Leakage reduction: Transmon leakage into |f⟩ is a known error source that standard error correction does not handle; the demonstrated 0.51% residual |f⟩ population is a direct mitigation path.
  • Scalable cryogenic hardware: On-chip resistive terminations eliminate dedicated reset lines and reduce thermal noise ingress from higher-temperature stages.
  • Engineered dissipation as a resource: The authors point to autonomous quantum error correction and dissipative preparation of quantum-correlated many-body states as longer-term uses of on-chip resistive loads.

Industry relevance: The scheme uses standard thin-film deposition, tantalum resonators, and transmon fabrication, so it is compatible with existing superconducting foundry processes. Its main selling point to hardware builders is that it adds reset capability without adding control channels, cryogenic ports, or active electronics.

Future Directions

  • Improving frequency selectivity: Jointly optimizing g and κr to approach critical damping (4g = κr), or inserting a second resonator between qubit and dissipator, to shrink the detuning window over which residual relaxation is significant.
  • Cryogenic characterization and optimization of the terminations: The paper attributes part of the excess effective temperature (30.6 mK versus a roughly 10 mK base) to imperfect thermalization or residual radiation, and suggests improved thermal anchoring and shielding.
  • Closing the gap between measured and predicted reset times: Fabrication variation in the small coupling capacitance, qubit dephasing, and residual flux-pulse distortion after IIR correction are all unmodeled effects that may bias the fitted T1^reset; the paper leaves these unresolved.
  • Suppressing spurious-mode re-excitation: The broadened spread in P(e|e) and the relaxation feature near ωq/2π ≈ 4 GHz are tentatively linked to two-level-system defects that can temporarily store and return excitation during flux tuning.

Target Audience

Experimental superconducting-qubit researchers and hardware engineers, particularly those working on qubit initialization, leakage reduction, and cryogenic packaging for multi-qubit processors. It is also useful for quantum error-correction experimentalists who need to know what reset and leakage-removal performance is currently achievable on chip, and for graduate students in circuit QED who want a clear worked example of dissipation engineering with a validated lumped-element design model.

Authors’ abstract

Rapid and high-fidelity qubit reset is essential for quantum state initialization and quantum error correction. Conventional reset protocols for superconducting qubits often rely on dissipative elements connected to external circuitry, exposing qubits to thermal noise from higher-temperature environments. Here, we experimentally demonstrate a scalable approach to qubit reset and leakage removal over multiple transmon qubits using dedicated dissipation units comprising…

The full abstract on arXiv
Read the original paper