Sampling Noise and Optimized Measurement Distribution in Imaginary-Time Quantum Dynamics Simulations

Feng Zhang, Niladri Gomes, Joshua Aftergood, Thomas Iadecola, Yongxin Yao, Peter P. Orth · 2026

Read Sampling Noise and Optimized Measurement Distribution in Imaginary-Time Quantum Dynamics Simulations by Feng Zhang, Niladri Gomes, Joshua Aftergood, Thomas Iadecola, Yongxin Yao, Peter P. Orth (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-2605-20378.txt.

Introduction

Zhang and five co-workers study sampling noise in variational quantum dynamics on near-term devices. Fixed-depth circuits promise real- and imaginary-time simulation, but finite measurements inject noise that limits ground-state preparation in one-dimensional Ising models. They compare regularization strategies and tune how measurement shots are shared across circuits. Optimized sharing more than halves total measurement cost versus uniform sharing, while keeping enough shots on every circuit.

Catalog record

CREATORFeng Zhang, Niladri Gomes, Joshua Aftergood, Thomas Iadecola, Yongxin Yao, Peter P. Orth
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE47921
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.20378

Browse the full library · About the Tower of Babel

Full text

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

Sampling Noise and Optimized Measurement Distribution in Imaginary-Time Quantum Dynamics Simulations Feng Zhang, Niladri Gomes, Joshua Aftergood, Thomas Iadecola, Yongxin Yao, Peter P. Orth Abstract Variational quantum dynamics simulations (VQDS) provide a promising route to simulate real- and imaginary-time quantum dynamics on noisy intermediate-scale quantum devices using fixed-depth circuits. However, their practical performance is strongly limited by sampling noise arising from a finite number of circuit measurements. In this work, we systematically investigate the impact of sampling noise on VQDS, with a focus on ground-state preparation in one-dimensional Ising spin models using imaginary time evolution. We compare different regularization strategies for stabilizing the equations of motion and show that Tikhonov regularization provides robust performance in noisy imaginary-time evolution. We then benchmark measurement-distribution strategies that allocate shots by minimizing a cost function that characterizes the error in solving the equation of motion. Using noisy circuit simulations, we demonstrate that such optimized shot allocation can significantly improve state fidelity and reduce the total measurement cost by more than a factor of two compared to uniform shot distributions. We observe that the best results are found if a sufficiently large number of measurements is guaranteed for all circuits, suggesting that a finite fraction of shots should be distributed evenly. Our results provide practical guidelines for implementing measurement-efficient variational quantum dynamics and ground-state preparation on near-term quantum hardware. I Introduction Simulating the dynamics of quantum many-body systems is one of the most promising applications of quantum computing. While classical implementations of this problem generically require exponential resources to store and manipulate a volume-law entangled quantum state Sandvik 2010 ; White 1992 ; Prosen and Žnidarič 2007 , there exist efficient implementations on a quantum computer Lloyd 1996 ; Childs and Wiebe 2012 ; Low and Chuang 2017 ; Low and Chuang 2019 ; Haah et al. 2021 . For Hamiltonians containing only local interactions, the required resources for Hamiltonian simulation scale polynomially in the final simulation time T and the number of qubits N Low and Chuang 2019 ; Childs et al. 2018 . Methods that exhibit the best asymptotic scaling such as qubitization Low and Chuang 2019 or quantum walk methods Berry et al. 2015 , however, suffer from significant overhead costs in the required number of entangling gates and ancilla qubits. In the current era of noisy intermediate-scale quantum (NISQ) computers Preskill 2018 , the number of qubits is still limited, and the maximal circuit depth is bounded by a finite device coherence time. This severely limits the direct application of Trotter product formulas and other even more resource-intensive algorithms. Trotter dynamics simulations can thus currently only reach final simulation times T that are of order one to ten in units of the natural energy scale of the Hamiltonian Martinez et al. 2016 ; Lamm and Lawrence 2018 ; Smith et al. 2019 ; Frey and Rachel 2022 ; Kim et al. 2023a ; Chen et al. 2022 ; Kim et al. 2023b ; Chen et al. 2024a ; Chertkov et al. 2026 . This has motivated research on variational quantum dynamics simulations with fixed-depth parameterized quantum circuits, applicable along both the real-time Li and Benjamin 2017 ; Yuan et al. 2019 ; Yao et al. 2021 ; Lin et al. 2021 ; Barratt et al. 2021 ; Barison et al. 2021 ; Benedetti et al. 2021 ; Mansuroglu et al. 2021 ; Berthusen et al. 2022 , and the imaginary-time axis McArdle et al. 2019 ; Yuan et al. 2019 , the latter alternatively referred to as variational quantum imaginary-time evolution (VQITE). One example is VQDS based on McLachlan’s variational principle Li and Benjamin 2017 ; Yuan et al. 2019 , where one propagates the parameters \bm{\theta} of a variational wavefunction \ket{\Psi(\bm{\theta}(t))} in real (or imaginary) time according to classical equations of motion M\dot{\bm{\theta}}=V . The coefficients M_{\mu\nu} and V_{\mu} are calculated on a quantum computer. Recent work has aimed to reduce the measurement overhead of the quantum geometric tensor (QGT) M , which exhibits unfavorable quadratic scaling with respect to the number of variational parameters Gacon et al. 2023 ; Gacon et al. 2024 ; Anuar et al. 2024 . For example, operator-projected VQITE relaxes the requirement that the full variational state follows the imaginary-time evolution path, instead enforcing that a selected set of operator expectation values satisfy the projected imaginary-time equations. This approach circumvents explicit QGT estimation and correspondingly reduces the measurement cost scaling with the number of variational parameters Anuar et al. 2024 . VQDS with an adaptive ansatz Yao et al. 2021 (AVQDS) was shown to exhibit a favorable linear scaling of the number of parameters with early simulation times N_{\theta}\propto T when describing generic nonequilibrium real-time dynamics of nonintegrable spin models such as the mixed-field Ising model (MFIM). The number of parameters N_{\theta} was found to saturate at long times to a value N^{\text{sat}}_{\theta} . Numerical benchmarks for nonintegrable models indicate that N^{\mathrm{sat}}_{\theta} grows exponentially with system size N . In contrast, for the integrable transverse-field Ising model (TFIM), a polynomial scaling N_{\theta}\propto N^{2} is observed, consistent with fast-forwarding algorithms Atia and Aharonov 2017 ; Cirstoiu et al. 2020 ; Gibbs et al. 2021 ; Kökcü et al. 2022 ; Bassman et al. 2022 ; Camps et al. 2021 ; Gulania et al. 2021 . This makes VQDS most attractive for simulating early-time dynamics of general models and long-time dynamics for integrable systems, where one expects a polynomial scaling of N_{\theta} with T and N .The imaginary-time version of AVQDS, termed Adaptive VQITE (AVQITE), was shown to produce compact ground-state ansätze for molecules and spin models Gomes et al. 2021a , with performance comparable to other adaptive ansatz construction methods Grimsley et al. 2019 ; Tang et al. 2021 ; Yordanov et al. 2021 . Additional applications include finite-temperature simulations Getelina et al. 2023 ; Chen et al. 2025 , calculations of Green’s functions and nonlinear response functions Mootz et al. 2024a ; Gomes et al. 2023 ; Mootz et al. 2024b , evaluations of topological properties Mootz and Yao 2026 , and simulations of systems governed by non-Hermitian Hamiltonians, including open quantum systems Chen et al. 2024b and certain formulations of partial differential equations Alipanah et al. 2025 . Previous studies of VQDS have mostly relied on statevector quantum simulations, which correspond to a perfect quantum computer using an infinite number of measurements to obtain its results. Here, we explore in detail the impact of shot noise arising from a finite number \mathcal{M} of measurements in VQDS. We consider state propagation of one-dimensional quantum spin models, focusing on imaginary time, with a particular emphasis on ground-state preparation. This is a well-suited task for VQDS due to the expectation that ground states with area-law entanglement entropy in one dimension can be prepared by circuits of modest depth Hastings 2007 ; Eisert et al. 2010 ; Schollwöck 2011 . It was previously found that sampling noise due to finite \mathcal{M} can have a substantial detrimental effect on the performance of VQDS, since the required inversion of the often ill-conditioned matrix M leads to an unfavorable scaling of \mathcal{M} with the condition number Yuan et al. 2019 ; McArdle et al. 2019 ; Stokes et al. 2020 ;

Read the complete text