Which Superconducting Qubit Model is Good Enough? CCS: Computer systems organization Quantum computing

Frej Larssen, Ivy Peng, Stefano Markidis · 2026

Read Which Superconducting Qubit Model is Good Enough? CCS: Computer systems organization Quantum computing by Frej Larssen, Ivy Peng, Stefano Markidis (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-23034.txt.

Introduction

Larssen, Peng, and Markidis ask which superconducting qubit model is good enough for pulse-level simulation of a flux-tunable two-qubit device with a fixed bus coupler. They compare an effective two-level model, one three-mode Duffing model, plus a circuit-based transmon design in the charge basis across spectra, interaction terms, driven dynamics, CZ gates, leakage, and runtime. The Duffing model tracks the circuit reference best for static quantities, while multilevel models reveal driven effects the effective model misses. They endorse layered use of abstraction.

Catalog record

CREATORFrej Larssen, Ivy Peng, Stefano Markidis
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41412
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.23034

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Which Superconducting Qubit Model is Good Enough? CCS: Computer systems organization Quantum computing Frej Larssen, Ivy Peng, Stefano Markidis Abstract . Pulse-level simulators are the lowest-level, most widely used abstraction layer for studying how quantum hardware responds to control signals, but they can be built on Hamiltonian models with very different fidelity and cost. This raises the question: which level of physical abstraction is sufficient for a given simulation objective? We study this question for a flux-tunable two-qubit superconducting device with a fixed bus coupler by comparing three Hamiltonian descriptions of the same hardware: an effective two-level model, a three-mode Duffing model, and a circuit-based transmon model in the charge basis. Using a realistic parameter set, we evaluate these models on a common benchmark suite spanning flux-dependent spectra, extracted two-qubit interaction terms, driven single-qubit dynamics, CZ gate dynamics, leakage outside the computational subspace, and runtime. Across the tested flux range, the Duffing model follows the circuit-based reference more closely than the effective model for static spectra and reduced two-qubit quantities, while in driven benchmarks, the multilevel models reveal effects absent in the effective description. Overall, the results support a layered use of abstraction in pulse-level simulation: effective models for reduced analyses, Duffing models as a practical multilevel default, and circuit-based models for high-fidelity reference simulation or detailed leakage analysis. From Effective Two-Level to Circuit-Based Hamiltonians for Pulse-Level Simulation † † cc-license: by 1. Introduction Pulse-level simulation is the lowest abstraction level commonly used in quantum hardware simulation before one reaches full device-specific electromagnetic or circuit-design workflows. Instead of representing a quantum program as a sequence of ideal gates, it represents the hardware as a controlled dynamical system and simulates how externally applied signals drive its quantum state over time ( 10 ; 7 ; 13 ; 14 ) . For this reason, pulse-level simulation is the natural setting for high-fidelity studies of quantum devices, as it retains the hardware-level dynamics from which gate operations arise. In a pulse-level simulation, the dynamics are governed by a time-dependent Hamiltonian, determined by both the internal structure of the device and the applied controls. A pulse-level simulator does two main kinds of computations. For fixed control settings, it computes eigenvalues and eigenstates of the Hamiltonian in order to obtain energy levels, transition frequencies (the frequencies associated with transitions between energy levels), and dressed states of the interacting system (eigenstates of the coupled qubit-coupler system rather than of the uncoupled subsystems). For time-dependent controls, it propagates the quantum state under the Hamiltonian to predict the dynamical response of the device during a pulse. In this work, we focus on superconducting qubits, where pulse-level simulation is especially important because control is implemented directly through analog signals. In superconducting circuits, the Hamiltonian is modified through two main control channels. The first is a drive pulse , typically a microwave control signal, which acts directly on a qubit and is used mainly for single-qubit gates. The second is a flux pulse , or time-dependent flux-bias control, which changes qubit frequencies and therefore changes the interaction pattern of the device. In this work, the single-qubit benchmark uses a drive pulse and the two-qubit benchmark uses a flux pulse. The basic components of the device are qubits, couplers, and control channels. The qubits store the computational states, the coupler mediates interaction between them, and the control signals modify the Hamiltonian over time. Superconducting circuits are therefore a natural setting for pulse-level simulation. The same logical operation may arise from underlying dynamics that can be modeled at several levels of physical detail. The central question is then one of abstraction: which Hamiltonian description is sufficient for the simulation objective? A reduced model may be fast enough for repeated sweeps, fitting, or control optimization, but it may omit dynamical effects that become important under realistic control pulses ( 1 ; 11 ) . A more detailed model may capture these effects, but at a substantially higher numerical cost. We evaluate the models using a benchmark suite that includes both static and time-domain tasks. The static benchmarks test how well each model reproduces dressed spectra and reduced two-qubit quantities as the controls vary across operating points. The time-domain benchmarks solve the driven dynamics under explicit pulses and test single-qubit behavior, CZ gate dynamics, and leakage outside the computational subspace. We also measure runtime. The main contributions of this paper are: • We compare three Hamiltonian descriptions for pulse-level simulation of a superconducting two-qubit device with a fixed harmonic bus: an effective two-level model, a three-mode Duffing model, and a circuit-based transmon model in the charge basis. • We evaluate these models on a common set of benchmarks covering both static and dynamical behavior, including dressed spectra, extracted effective quantities, driven single-qubit dynamics, CZ gate dynamics, leakage outside the computational subspace, and computational cost. • We show that the ranking of the models depends on the observable: the Duffing model is more faithful for flux-dependent static structure, the effective model can still be competitive for reduced observables such as the CZ conditional phase, and the circuit-based model is needed when detailed leakage pathways must be resolved. • We provide practical guidance on when each model is sufficient, ranging from rapid reduced analysis to leakage-sensitive pulse-level simulation. 2. Preliminaries In gate-level simulation, a quantum program is represented as a sequence of ideal unitary gates acting on a finite-dimensional computational state space. Pulse-level simulation uses a lower-level description. Instead of assuming ideal gates, it models the physical device directly through a Hamiltonian that depends on externally applied controls such as microwave drives and flux biases ( 7 ; 13 ; 12 ) . In pulse-level simulations, the Hamiltonian is the central quantity. It specifies both the internal structure of the device and how control signals act on it. The quantum state \ket{\psi(t)} represents the instantaneous state of the full physical system. Depending on the model, that state may include only the computational states of the qubits, or it may also include higher excited qubit levels and excitations of the coupler. The time evolution is governed by the time-dependent Schrödinger equation (1) i\frac{d}{dt}\ket{\psi(t)}=H(t)\ket{\psi(t)}. Solving this equation leads to the state trajectory induced by the controls. In this sense, pulse-level simulation treats the device as a controlled continuous-time dynamical system. A gate can be viewed as a compressed description of what the system does over a finite time window. If the control pulse from time t_{0} to t_{1} induces the unitary (2) U(t_{1},t_{0})=\mathcal{T}\exp\!\left(-i\int_{t_{0}}^{t_{1}}H(t)\,dt\right), where \mathcal{T} denotes time ordering, then the gate-level description keeps only U(t_{1},t_{0}) , not the intermediate state trajectory during that interval. In this sense, the gate-level view is a stroboscopic view of the underlying pulse-level dynamics ( 10 ) . In this work, we focus on transmon qubits. For a thorough review on superconducting qubits such as transmons, see Refs. ( 6 ; 7 ) . An important issue in pulse-level simulations is the representation of the state space (the set of states that the simulator allows the system to occupy). The physical models considered in this work are not naturally finite-dimensional. Oscillator-based models are infinite-dimensional because an oscillator can, in principle, occupy arbitrarily many excitation levels. Likewise, circuit-based transmon models written in the charge

Read the complete text