Researchers at Chalmers University of Technology have computed the effective qubit couplings, state-dependent interactions and eigenstate locality of a 96-transmon array in under one hour using a single GPU. The team employed continuous unitary transformations to derive effective Hamiltonians, scaling polynomially with qubit number and enabling modeling of processors beyond the reach of exact classical simulation.
Combining these models with semiclassical sampling, they then predicted the quench dynamics of 100 qubits and 180 couplers, a sector of dimension 10^(29), validating the approach against smaller arrays. The results chart which observables lie within classical reach and offer a practical tool for designing future devices.
Modeling 96-Transmon Arrays with Tunable Couplers
Calculating the effective interactions within a quantum processor is becoming increasingly difficult as these systems grow, yet precise models remain essential for both design and verifying claims of quantum advantage. Specifically, the team derived effective Hamiltonians for a 96-transmon array using continuous unitary transformations, a technique that achieves a computational cost scaling polynomially with the number of qubits.
This approach allowed for the extraction of key parameters guiding processor design, including effective qubit couplings, state-dependent interactions such as ZZ couplings, and the spatial locality of eigenstates. The computation, performed on a single GPU, requires at most one hour per design point, a speed that facilitates rapid iteration during the development of new array configurations.
Beyond static properties, the researchers validated their models by predicting the “quench dynamics” of a 100-qubit and 180-coupler system, a computational space with a dimension of , using semiclassical sampling. This prediction, while not challenging existing claims of quantum advantage based on cross-entropy benchmarks, successfully captured key behaviors like relaxation, thermalization and transport when compared against exact dynamics of smaller arrays.
The predicted transport carries a fingerprint of the device: under a staggered detuning, excitations cross over from ballistic to diffusive spreading, slow down at a doublon resonance and diffuse at different rates on each sublattice. The core of this advancement lies in the use of continuous unitary transformations, which approximately eliminate non-excitation-conserving terms and couplers within two-dimensional transmon arrays.
Extracting Effective Qubit Couplings and Eigenstate Locality
The resulting models are not merely theoretical exercises. They directly inform processor design by revealing parameters like state-dependent ZZ couplings and how quantum information localizes across the chip. Researchers achieved this prediction by combining their effective models with semiclassical dissipative time-evolution approximation (DTWA) sampling, allowing them to simulate relaxation, thermalization and transport within the array. This allows for precise determination of low-excitation physics even when higher-order terms are discarded.
The team’s work identifies key hardware metrics for superconducting qubit arrays, including effective qubit couplings, state-dependent energy shifts like ZZ couplings, and spatial eigenstate locality. These parameters are not simply calculated in isolation. The method provides a way to assess how well the model aligns with the physical device, with predictions carrying “fingerprints” that can be experimentally verified through detuning sweeps.
CUT Method Improves Stability in Bosonic Models
This speed stems from a generator, an extension of the sign generator to operator expansions, that achieves faster convergence than alternatives for near-degenerate and ordered bosonic systems. The technique uses continuous unitary transformations, or CUTs, to build renormalization flows towards effective models that scale polynomially in system size due to finite operator expansions. Bosonic systems present unique challenges for CUT implementations, introducing multi-excitation resonances that worsen convergence, particularly in ordered and non-perturbative regimes.
Unlike previous approaches that attempt to perturb trajectories away from unstable fixed points, this method alleviates convergence difficulties by extending the sign generator, offering an advantage near degeneracies. The resulting effective Hamiltonian captures remaining off-diagonal couplings, represented as V(l) = H(l)-H_0(l), and allows for rapid iteration on processor designs.
Predicting Quench Dynamics with Semiclassical Sampling
This computational speed stems from a method that parallelizes over tensor contractions, a technique important for handling the complex interactions within these systems. The resulting effective qubit models, when combined with semiclassical sampling, allow prediction of the behavior of quantum processors exceeding the capabilities of exact classical simulation. This prediction, validated against exact dynamics of smaller arrays, demonstrates the ability to model systems far beyond the reach of traditional simulation methods.
The team’s work focuses on understanding which properties of quantum processors remain predictable on classical computers, a critical step for both building these machines and establishing the limits of quantum advantage. The method’s efficacy extends to scenarios with longer-range interactions, generated by couplers and non-excitation-conserving terms, which dictate the magnitude of hopping and coupling strengths.
Predicting dynamics in these devices requires deriving an effective Hamiltonian from the device model, a process the Chalmers team has streamlined. “At resonance, the two DTWAs agree to within their sampling noise,” the researchers report, indicating a high degree of consistency in their modeling.
100-Qubit Simulation Validates Effective Model Approach
The resulting effective qubit model, combined with a semiclassical sampler, allows for the prediction of complex quantum dynamics with efficiency. This prediction was validated against exact dynamics calculations performed on smaller arrays, confirming the model’s accuracy in simulating complex quantum behavior. The approach successfully captures relaxation, thermalization and transport phenomena, offering a powerful tool for understanding and optimizing quantum processor performance. The team focused on excitation-conserving terms, first eliminating non-conserving terms and then the couplers themselves to arrive at the effective qubit model.
This computation parallelizes over tensor contractions and thus takes an hour at most per design point on a single GPU, enabling rapid design iterations of the full array. While the simulations assume closed systems without decoherence and neglect thermal or leaked coupler excitations, the results demonstrate the potential of effective models to predict and understand the behavior of increasingly complex quantum processors. The work provides a baseline for future investigations incorporating these additional factors, further refining the accuracy and predictive power of the model.
Classical Reach of Observables in Quantum Processors
The ability to rapidly iterate through design points is a significant advantage in optimizing device performance. The approach captures key physical phenomena, including relaxation, thermalization and transport, but does not extend to predicting bitstring statistics, the very data used in cross-entropy benchmarks. The researchers note that the model yields local densities and low-order correlators, but falls short at predicting bitstring probabilities, clarifying the limits of the model’s predictive power. The work suggests that as quantum processors grow in size and complexity, these types of effective models will become increasingly important for guiding their development.
👉 More information
🗞 Modeling couplings and quench dynamics of hundred-qubit quantum processors
✍️ Simon Pettersson Fors, Jorge Fernández-Pendás and Anton Frisk Kockum (Department of Microtechnology and Nanoscience, Chalmers University of Technology, Sweden); Eric Djäknegren (Department of Physics and Astronomy, Chalmers University of Technology, Sweden)
🧠 ArXiv: https://arxiv.org/abs/2611.25053




See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
