Understanding the physics of D-Wave annealers: From Schrödinger to Lindblad to Markovian Dynamics

Vrinda Mehta, Hans De Raedt, Kristel Michielsen, Fengping Jin · 2025

Read Understanding the physics of D-Wave annealers: From Schrödinger to Lindblad to Markovian Dynamics by Vrinda Mehta, Hans De Raedt, Kristel Michielsen, Fengping Jin (2025) 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-2503-21565.txt.

Introduction

Vrinda Mehta, Hans De Raedt, Kristel Michielsen, and Fengping Jin test whether D-Wave quantum annealers truly sample as quantum physics predicts, in work from twenty twenty-five. These machines hunt low energy states by slow cooling like quantum annealing. What physics governs their output is still debated.

Under standard and fast annealing runs, the machines' output frequencies matched the classic Gibbs spread when anneals ran long enough. The team compared data with quantum and stepwise random models at one and two qubit scale. Simulations of the Bloch and Lindblad equations framed the checks. Results shed light on how much quantum behavior drives the samplers. The work feeds the wider debate on annealer physics.

Catalog record

CREATORVrinda Mehta, Hans De Raedt, Kristel Michielsen, Fengping Jin
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE52 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-21565

Browse the full library · About the Tower of Babel

Full text

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

Understanding the physics of D-Wave annealers: From Schrödinger to Lindblad to Markovian Dynamics Vrinda Mehta, Hans De Raedt, Kristel Michielsen, Fengping Jin Abstract Understanding the physical nature of the D-Wave annealers remains a subject of active investigation. In this study, we analyze the sampling behavior of these systems and explore whether their results can be replicated using quantum and Markovian models. Employing the standard and the fast annealing protocols, we observe that the D-Wave annealers sample states with frequencies matching the Gibbs distribution for sufficiently long annealing times. Using Bloch equation simulations for single-qubit problems and Lindblad and Markovian master equations for two-qubit systems, we compare experimental data with theoretical predictions. Our results provide insights into the role of quantum mechanics in these devices. I Introduction How crucial is quantum theory in describing the results of D-Wave annealers? Put differently, can the data produced by these systems be explained or replicated using non-quantum models? If so, to what extent can these machines truly be considered “quantum”? These questions have been lying around since the introduction of the D-Wave annealers [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 7 , 11 , 12 , 13 , 14 , 15 , 16 ] . The role of quantum phenomena in determining the performance of D-Wave quantum annealers has been a subject of intense debate since their introduction [ 3 , 5 , 6 ] . A fundamental question is whether the observed behavior of these devices can be explained using non-quantum models or whether quantum effects, such as entanglement and tunneling, have to be incorporated to reproduce experimental results. Addressing this question is crucial for understanding the extent to which these machines leverage quantum mechanics for computational advantage [ 7 , 10 , 17 , 14 ] . Given the complexity of these large-scale annealers, an effective approach to investigate their behavior is to focus on minimal problems that retain the essential characteristics of these systems while being amenable to analytical and numerical methods. As demonstrated in our previous work [ 18 ] , in theory, even 1- and 2-qubit problems can capture key features of the underlying dynamics, including the role of quantum coherence and thermal effects. Indeed, these small problem instances have been an excellent testbed for evaluating the quantum nature of these devices [ 8 , 9 ] . In this work, we analyze the statistics of the states sampled by D-Wave quantum annealers for specific problem instances. We study the frequency distributions of these sampled states, construct theoretical models, investigate the dynamical evolution of these systems by employing numerical simulations based on the Schrödinger equation, the Lindblad master equation [ 19 ] , and the Markovian master equation, and scrutinize the extent to which the experimental data aligns with the different simulation results. In the same vein as with the reverse annealing protocol in our previous work [ 18 ] , the present work focuses on understanding the behavior of the D-Wave quantum annealers using the standard and the recently introduced fast annealing protocols. Our results contribute to the broader discussion on the quantum nature of D-Wave devices and the extent to which their behavior can be emulated by classical processes. By systematically comparing theoretical predictions with experimental data, we provide insights into the role of quantum effects in quantum annealing and their implications for practical applications. It is important to emphasize that the present work explores regimes that are complementary to those studied in Refs. [ 12 , 13 , 14 ] , particularly in terms of problem complexity and scale. First, our study focuses on a fundamental aspect—the frequencies with which the D-Wave annealer samples different states—and compares them to theoretical expectations from Schrödinger, Lindblad, and Markovian dynamics. This approach offers a raw and direct perspective on the system’s behavior. Second, our analysis is centered on 1- and 2-spin systems, which allow for a more controlled and detailed examination of the underlying physical processes. These small-scale problems provide the best opportunity to observe signatures of quantum coherence; failure to detect such effects at this level would significantly diminish the prospects of observing them in larger, more complex systems. The paper is structured as follows: Section II outlines the methods and problems considered in this study. Section III presents the results obtained from the D-Wave annealer, while Section IV discusses the corresponding simulation results. Finally, Section V summarizes the key findings of the paper. II Methods and problems The Hamiltonian for the quantum annealing process, implemented on the D-Wave annealers, is given by [ 20 ] \displaystyle\frac{H(t)}{\hbar} \displaystyle=\frac{\pi A(s=t/t_{a})}{h}H_{D}+\frac{\pi B(s=t/t_{a})}{h}H_{P}, \displaystyle H_{D} \displaystyle=-\sum_{i}\sigma_{i}^{x}, \displaystyle H_{P} \displaystyle=\sum_{i}h_{i}\sigma_{i}^{z}+\sum_{i>j}J_{ij}\sigma_{i}^{z}\sigma_{j}^{z}, (1) where t_{a} is the annealing time, \sigma_{i}^{x} , \sigma_{i}^{z} are the Pauli matrices, h_{i} are effective local fields, and J_{ij} are the effective coupling terms between qubits i and j . Functions A(s) and B(s) (expressed in units GHz) define the annealing schedules and are vastly different for the standard and the fast annealing protocols (see appendix A ). Under the conditions that A(0)\gg B(0) , A(1)\ll B(1) , and that t_{a} is sufficiently large, the adiabatic theorem [ 21 , 22 ] guarantees that the system will find the ground state of the problem Hamiltonian H_{P} , if the system is initialized in the ground state \ket{++\cdots+}_{N}=\frac{1}{\sqrt{2}^{N}}(\ket{0}+\ket{1})^{\otimes N}\;, (2) of the driver Hamiltonian H_{D} , where N is the number of qubits. The fast annealing protocol requires that the fields in Eq. ( 1 ) have to be zero [ 20 ] . A workaround for implementing problems with non-zero h_{i} ’s is to use an additional qubit for every variable that has a non-zero h_{i} [ 20 ] . The state of the additional qubit is set by applying a fixed flux bias offset to it. Then, the coupling J_{ij}=h_{i} between the original qubit i and the additional qubit j effectively implements the original problem. For a 1-spin problem with a magnetic field h_{1}^{z} , the resulting annealing Hamiltonian reads \displaystyle\frac{H^{\prime}(t)}{\hbar}=-\frac{\pi A(s)}{h}~(\sigma_{1}^{x}+\sigma_{2}^{x})+\frac{\pi B(s)}{h}~J_{12}\sigma_{1}^{z}\sigma_{2}^{z}+h_{FB}\sigma_{2}^{z}\;, (3) where J_{12}=h_{1}^{z} and h_{FB}\gg\pi\max(A(s),B(s)h_{1})/h is the constant flux bias offset applied to the second qubit. Similarly, for a given 2-spin problem with fields h_{1} and h_{2} , and coupling J , the corresponding annealing Hamiltonian for the fast annealing protocol reads \displaystyle\frac{H^{\prime\prime}(t)}{\hbar}= \displaystyle-\frac{\pi A(s)}{h}~(\sigma_{1}^{x}+\sigma_{2}^{x}+\sigma_{3}^{x}+\sigma_{4}^{x})+h_{FB}(\sigma_{3}^{z}+\sigma_{4}^{z}) \displaystyle+\frac{\pi B(s)}{h}~(J_{12}\sigma_{1}^{z}\sigma_{2}^{z}+J_{13}\sigma_{1}^{z}\sigma_{3}^{z}+J_{24}\sigma_{2}^{z}\sigma_{4}^{z})\;, (4) where J_{13}=h_{1} and J_{24}=h_{2} . We study the annealing time dependence of the probabilities of all the relevant energy levels, assigned by using their sampling frequencies. All the results presented in this paper are obtained by averaging over multiple runs across different qubits available in the D-Wave systems. Additionally, these multiple runs are performed either sequentially—where only one instance of the problem is executed at a time—or simultaneously—where multiple instances are submitted in parallel. To numerically study the annealing process of the D-Wave systems, we use the Bloch equations, the Lindblad master equation, and the Markovian master equation. We use the functions obtained by fitting to the D-Wave annealing schedule data for the standard and, in addition, the newly introduced fast annealing protocols (see appendix A ). We study the same simple yet diverse problem set as the one in Ref. [ 18 ]

Read the complete text