Hybrid Quantum Design Using Superconducting Resonators Could Ease Gate Fidelity Demands 1,000-Fold, Theory Shows

• What happened: A research team led by Eleanor Crane of King's College London has published in Physical Review X a new quantum computing framework that holds the quantum states of bosons directly in superconducting resonators.
• Why it matters: The framework reduces the burden of encoding boson states into multiple qubits. The team's theoretical analysis suggests that gates with 99.9% fidelity could achieve the same circuit accuracy that a qubit-only approach would need 99.9999% or higher fidelity to match.
• What to watch next: The results rest on resource estimates and numerical simulations. Demonstrating gate operations on real superconducting hardware, and handling physical errors such as photon loss, remain open challenges.

Particles in nature fall into two broad classes: fermions, such as electrons and quarks, which make up matter, and bosons, such as photons and phonons. Simulating how these particles interact on a quantum computer poses a fundamental challenge for today's mainstream qubit-based machines. Common qubits, such as those built from superconducting circuits or trapped ions, use two states, corresponding to 0 and 1, as their basic unit, which suits the representation of fermions. Bosons, however, can occupy the same quantum state in large numbers, so handling them requires combining many qubits, which greatly increases the computational resources needed.

To address this problem, an international team led by Eleanor Crane, a lecturer at King's College London, has proposed a new quantum computing framework. The results were published on October 7, 2026, in the physics journal Physical Review X (Phys. Rev. X 16, 041008). The proposal is a hybrid scheme that combines superconducting transmon qubits with high-Q three-dimensional microwave cavity resonators, holding the quantum states of bosons directly in the electromagnetic field modes inside the resonators. This could eliminate the step of encoding bosons into many qubits and substantially reduce the number of gates needed for quantum simulation.

The study, however, is not an experiment in which a hybrid quantum processor was built and used to run algorithms. The team evaluated the proposed scheme through mathematical gate decompositions, resource estimates, and numerical simulations based on the Lindblad master equation that account for physical noise. The results suggest that, even before large-scale quantum error correction becomes practical, some quantum systems in which fermions and bosons interact may be computable.

Why two-level qubits struggle to represent bosons

Quantum particles are divided into fermions and bosons according to their statistical properties. Fermions obey the Pauli exclusion principle, so multiple particles cannot occupy the same quantum state. A single fermion mode therefore has only two states, particle present or absent, which a two-level qubit represents naturally.

Bosons face no such restriction: any number of them can occupy the same quantum state. Photons and phonons, the quantized vibrations of a crystal lattice, are typical examples. To handle bosons on a computer built only from qubits, one must first cap the maximum occupation number with a cutoff $S$ and encode the possible states into several two-level qubits. Representing occupation numbers from 0 to $S$ requires at least $\lceil {\log }_2(S+1)\rceil$ qubits. Computing the creation, annihilation, and interaction of particles then requires repeating complex gate operations across these qubits.

Anthony Ciavarella, a computational scientist at Lawrence Berkeley National Laboratory in the US, noted in Physics (Physics 19, 135), the American Physical Society's commentary journal, that this encoding has been a major burden for quantum simulation. The more qubits are needed to represent bosons, the more complex the gate operations become, and the more easily errors accumulate during a computation. Introducing error correction requires additional qubits and control operations, which raises resource demands even further.

In noisy intermediate-scale quantum (NISQ) computers, which do not rely on error correction, decoherence and gate errors become serious as computations grow. Trying to improve simulation accuracy by raising the maximum boson occupation number increases the required gate operations, risking results buried in noise. Many problems in condensed matter physics, quantum chemistry, and nuclear physics hinge on interactions between fermions and bosons, but the large difference in the resources needed to represent each has been a major barrier for conventional approaches.

{
  "type": "bar",
  "title": "Gate error rates required for quantum simulations involving bosons (theoretical estimate)",
  "unit": "%",
  "categories": ["Hybrid approach (using resonators)", "Qubit-only approach (assuming best method)"],
  "series": [
    { "name": "Maximum tolerable gate error rate", "values": [0.1, 0.0001] }
  ],
  "caption": "Per-gate error rate needed to achieve equivalent fidelity in the compared circuits. Smaller values demand higher precision. Estimates based on the research team's theoretical analysis.",
  "source": "Physical Review X 16, 041008 (2026)"
}

According to the team's analysis, achieving equivalent fidelity in the compared circuits with a qubit-only approach would require a per-gate error rate of 0.0001% or lower, meaning fidelity of 99.9999% or higher. The hybrid approach, which uses the resonator directly as a computational register, is estimated to reach the same circuit fidelity with an error rate of 0.1% (99.9% fidelity). That is a 1,000-fold difference in tolerable gate error rate, suggesting the hybrid approach could run computations at a level close to current experimental technology. This is, however, a theoretical comparison based on specific circuits and error models, and does not mean the same performance has been achieved on real hardware.

A new design that uses superconducting resonators as computational registers

Crane's team focused on the microwave cavity resonators used in superconducting circuit QED (quantum electrodynamics). In conventional superconducting quantum computers, high-Q three-dimensional microwave cavities have mainly been used to mediate interactions between qubits or to read out quantum states. But the resonators themselves can also hold quantum information, and their physical properties are well suited to representing boson states.

The electromagnetic field mode in a resonator is described quantum mechanically as a harmonic oscillator. The photon number increases in discrete steps of 0, 1, 2 and so on, and these states can be placed in quantum superposition. The team exploited this property to devise a scheme that maps the quantum states of bosons onto photon states in the resonator. Instead of combining multiple two-level qubits to represent bosons, the approach uses the state space the resonator naturally has for computation.

The paper was submitted on September 20, 2024, accepted on July 23, 2026 after revision, and formally published on October 7 of the same year. The underlying theory is also available as the preprint arXiv:2409.03747. Researchers from King's College London, Pacific Northwest National Laboratory (PNNL) in the US, the Canadian Institute for Advanced Research (CIFAR), France's École Normale Supérieure in Paris (ENS-PSL), the French National Centre for Scientific Research (CNRS), and the French National Institute for Research in Digital Science and Technology (Inria) took part in the collaboration.

The team also organized the instruction sets, abstract machine models, and universal gate sets needed to operate this hybrid approach as a general-purpose quantum computer in a related paper, "Hybrid Oscillator-Qubit Quantum Processors: Instruction Set Architectures, Abstract Machine Models, and Applications" (PRX Quantum 7, 010201, 2026). Published on January 7, 2026, it serves as a design guide for the hardware and software underlying the present work.

In the proposed configuration, fermionic degrees of freedom are assigned mainly to transmon qubits and bosonic degrees of freedom to high-Q microwave resonators. The distinguishing feature is that the resonator, formerly in an auxiliary role, serves as a computational register that stores and processes quantum states. By letting qubits and resonators each handle the quantum states they are best suited to, the scheme aims to compute fermion-boson interactions efficiently.

How gate counts stay low as the number of bosons grows

A key point of the study is that it uses a qubit-boson gate set with "$O(1)$ scaling," in which the number of basic operations needed for certain computations does not grow as the maximum boson occupation number increases, and applies it to concrete interaction calculations. Here, $O(1)$ means the number of operations stays constant regardless of the boson cutoff $S$. In qubit-only schemes, the number of qubits and gate operations grows as the maximum occupation number rises, so this difference has a major effect on computational efficiency.

For example, when computing the gauge-invariant hopping term describing particles moving between lattice sites, gate complexity in the conventional qubit-only approach grows as $O(\log (S{)}^2)$. In contrast, Crane's framework derives exact gate decompositions for the gauge-invariant hopping term and for density-density terms representing interactions between particle number densities. In a configuration where qubits and resonators are arranged on a two-dimensional square lattice, these interactions can reportedly be implemented at a constant circuit depth $O(1)$ independent of the cutoff $S$. The team also estimated that, for the gauge-invariant hopping term of bosonic matter, gate counts could be reduced by more than 10,000 times compared with the qubit-only approach. However, the absolute gate counts of the comparison baseline are not given in the published materials we could confirm.

The team also examined new computational methods for more complex interactions. One is the magnetic (plaquette) term of the (2+1)-dimensional U(1) quantum link model, described with two spatial dimensions and one time dimension. Because this term involves the creation and annihilation operators of several bosons, it is difficult to decompose exactly with simple gate operations. The team therefore developed an approximate decomposition using the Baker-Campbell-Hausdorff (BCH) formula.

In this construction, operation scaling with respect to the Trotter step $t$ and coupling constant $g$ is $O(t^1.25g^{-}2.5)$, while circuit depth with respect to the boson cutoff $S$ stays constant. The team claims this is the first method, on any physical platform, to compile the magnetic term of the U(1) quantum link model at a constant depth independent of the cutoff. In conventional qubit-only approaches, the cost of this term grows as $O(\log (S))$, so the hybrid approach is expected to have an advantage when handling boson systems with large occupation numbers.

The proposed framework goes beyond gate decompositions for individual interactions. It also includes Trotter expansion, which computes time evolution by dividing it into short intervals; a method that uses Gauss's law to detect some errors without ancilla qubits for error detection; and methods for measuring nonlocal physical quantities. The team further organized a variational quantum eigensolver (VQE) combining qubits and resonators, and a method for estimating ground-state energy using quantum signal processing (QSP). Together, these methods systematize the computational procedures for simulating an entire physical model, not just a single operation.

Although the work centers on superconducting circuit QED, the authors note that it may apply to other quantum hardware. The same ideas could be applied to any platform that can use bosonic degrees of freedom, such as the vibrational modes of trapped ions or neutral-atom arrays. However, the detailed hardware error analysis in this study targets superconducting circuit QED, and it has not been confirmed that other platforms would achieve the same performance.

Comparisons using two physical models, plus error simulations

To verify the framework's effectiveness, the team compared computational resources and ran numerical simulations on several physical models. The first comparison target was the simulation of a beam splitter, a device that splits and combines light. For a fair comparison with the qubit-only approach, the team developed a new qubit-only algorithm whose cost scales as \tilde(N_^) in the maximum boson occupation number N_. This algorithm is asymptotically more efficient than previously known methods. Even so, the hybrid approach, which uses resonators directly, was assessed as substantially reducing the number of gate operations needed to simulate the same physical process.

The team then ran detailed resource analyses on two simplified (2+1)-dimensional physical models. One was a \mathbb_2 lattice gauge theory coupled to bosonic matter, and the other was the (2+1)-dimensional U(1) quantum link model. These are "toy models," simplified versions of real physical systems used to evaluate the performance of quantum simulation methods that involve complex interactions.

According to Ciavarella, the \mathbb_2 lattice gauge theory used here shares the structure of the Higgs field sector and basic gauge fields of the Standard Model of particle physics. It is therefore an important test case for validating quantum simulation methods aimed at nuclear and particle physics. Using this model, the team evaluated not only gate counts but also the effects of errors that could occur on real hardware.

Working in (2+1) dimensions, with two spatial dimensions and one time dimension, is also meaningful. Models with one spatial dimension can often be computed relatively efficiently on classical computers, so they may be unsuitable for testing quantum advantage. Reproducing real three-dimensional space directly, on the other hand, is too demanding for current computational resources. Two spatial dimensions sit between the two, and are also important for studying electron correlations in two-dimensional materials such as graphene and in the copper-oxide planes related to high-temperature superconductivity.

To evaluate hardware errors, the team used numerical simulations based on the Lindblad master equation, reflecting realistic performance of superconducting circuit QED. This method can account for photon loss from the resonator as well as energy relaxation ($T_1$) and dephasing ($T_2$) of the transmon qubits. The team evaluated these errors even under conditions of large boson occupation numbers, examining how well the theoretical gate constructions withstand physical noise. According to the authors, this is the first study to evaluate hybrid quantum computation of lattice gauge theories consistently, from the specific algorithm down to hardware errors.

The results showed a large gap between the qubit-only and hybrid approaches. By the team's estimate, matching the overall circuit fidelity that the hybrid approach achieves with 99.9% per-gate fidelity (0.1% error rate) would require at least 99.9999% per-gate fidelity (0.0001% error rate or lower) in a qubit-only approach. In this comparison, tolerable gate error rates differ by a factor of 1,000. The authors note that achieving the extremely high fidelity required by the qubit-only approach would likely require large-scale quantum error correction.

The team also compared the approach with classical computers. Classical approximation methods such as tensor networks save resources by compressing quantum states, but their efficiency drops as quantum entanglement grows. In the \mathbb_2 lattice gauge theory simulations, entanglement between particles grew rapidly with time evolution, making it difficult for classical compressed computation to keep track. This is one basis for considering whether hybrid quantum computers could eventually outperform classical computation on similar problems.

Based on this analysis, the team proposed concrete experiments to test "post-qubit quantum advantage." The idea is that hybrid quantum computers may be able to handle problems that are hard for near-term qubit-only quantum computers and classical computers. At this stage, however, quantum advantage has not been demonstrated on real hardware, and the proposed computational methods and performance estimates must be verified in future experiments.

Can the theoretical advantage be confirmed on real machines?

The study shows that the computational resources needed for quantum simulation could be substantially reduced. In a general-audience summary of the paper, the team also mentions the possibility of cutting the computational burden to one-thousandth. It would not be appropriate, however, to read this as meaning that real quantum computers have become 1,000 times faster. What the study shows is a theoretical improvement in computational complexity for specific operations and physical models, together with an evaluation of error tolerance based on numerical simulation.

In particular, the comparison of per-gate error rates of 0.1% and 0.0001% is not a difference in processing speed measured on real hardware. It is an estimate of how much gate precision each approach needs to obtain equivalent fidelity in the target circuit. Reducing the number of required gates can also shorten computation time, but actual processing speed also depends on gate execution time and the time needed for measurement and control. The figure of 1,000 therefore cannot be taken as an improvement factor in execution speed or power consumption.

The team cites possible future applications including elucidating the mechanisms of photosynthesis in plants, research on room-temperature superconductivity, molecule-target binding in drug discovery, and particle interactions in high-energy physics. All are complex quantum systems involving fermions and bosons, and include problems that are hard to analyze with conventional computational methods. These, however, are motivations for the research and prospects for future application; the paper has not clarified any specific chemical reaction or the properties of a new material. Applications to quantum computers based on trapped ions or neutral atoms likewise remain at the stage of theoretical possibility.

Still, experts have praised the proposal. In his commentary in Physics, Ciavarella cited as the most promising aspect of the proposed approach that the hardware performance it requires is at a relatively realistic level. He assessed that it could be realized with near-term devices without waiting for large-scale quantum error correction to be completed. He also noted that the study sets concrete targets for hardware development, and that the required gate fidelity is close to levels existing experimental devices are approaching.

Of course, this too is an expert's assessment of a theoretical proposal and does not mean actual operation has been confirmed. Even so, it is significant that against a comparison condition in which conventional approaches demand extremely high gate fidelity of 99.9999% or more, the hybrid approach points to 99.9%, a level within sight of current superconducting circuit technology. For experimentalists, being able to estimate the required hardware performance concretely provides important input for planning future development.

The study points to a new direction for reducing the burden of quantum simulation by assigning fermions and bosons to the computational elements best suited to each. What matters is that it backs up, through both gate decomposition and error analysis, the idea of using the resonator not as a mere auxiliary device but as a main computational component. At the same time, achieving the theoretical efficiency on real machines will require further verification of the precision of gate operations combining resonators and transmons, and of the effects of photon loss and other errors.

The next focus is whether a hybrid quantum processor based on the proposed design can actually implement gate operations with $O(1)$ scaling independent of the cutoff. It will also be important to see whether circuits combining multiple qubits and resonators can be run stably while maintaining the target gate fidelity. How far the theoretical advantage carries over to real quantum computation is the next stage of verification.