Digitized Counterdiabatic Quantum Sampling

Narendra N. Hegade, Nachiket L. Kortikar, Balaganchi A. Bhargava, Juan F. R. Hernández, Alejandro Gomez Cadavid, Pranav Chandarana, Sebastián V. Romero, Shubham Kumar, Anton Simen, Anne-Maria Visuri, Enrique Solano, Paolo A. Erdman · 2025

Read Digitized Counterdiabatic Quantum Sampling by Narendra N. Hegade, Nachiket L. Kortikar, Balaganchi A. Bhargava, Juan F. R. Hernández, Alejandro Gomez Cadavid, Pranav Chandarana, Sebastián V. Romero, Shubham Kumar, Anton Simen, Anne-Maria Visuri, Enrique Solano, Paolo A. Erdman (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-2510-26735.txt.

Introduction

Narendra N Hegade, with eleven co-authors, proposes a hybrid quantum-classical method for sampling from energy-based models. Sampling from a Boltzmann distribution, which favours low-energy states, is central to physics and machine learning, but classical methods slow down badly at low temperatures. Their method, digitized counterdiabatic quantum sampling, uses shortcuts that suppress unwanted transitions during a quantum evolution.

Hegade and the team steer sampling toward low-energy regions step by step with a bias field, then reweight the gathered samples classically to rebuild the target distribution. They test it on one-dimensional Ising models with up to one hundred twenty-four qubits, where exact answers exist. A harder spin-glass case with one hundred fifty-six qubits ran on IBM quantum processors. Classical rivals needed up to a thousand times more samples to match its quality.

Catalog record

CREATORNarendra N. Hegade, Nachiket L. Kortikar, Balaganchi A. Bhargava, Juan F. R. Hernández, Alejandro Gomez Cadavid, Pranav Chandarana, Sebastián V. Romero, Shubham Kumar, Anton Simen, Anne-Maria Visuri, Enrique Solano, Paolo A. Erdman
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE86 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-26735

Browse the full library · About the Tower of Babel

Full text

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

Digitized Counterdiabatic Quantum Sampling Narendra N. Hegade, Nachiket L. Kortikar, Balaganchi A. Bhargava, Juan F. R. Hernández, Alejandro Gomez Cadavid, Pranav Chandarana, Sebastián V. Romero, Shubham Kumar, Anton Simen, Anne-Maria Visuri, Enrique Solano, Paolo A. Erdman Abstract We propose digitized counterdiabatic quantum sampling (DCQS), a hybrid quantum-classical algorithm for efficient sampling from energy-based models, such as low-temperature Boltzmann distributions. The method utilizes counterdiabatic protocols, which suppress non-adiabatic transitions, with an iterative bias-field procedure that progressively steers the sampling toward low-energy regions. We observe that the samples obtained at each iteration correspond to approximate Boltzmann distributions at effective temperatures. By aggregating these samples and applying classical reweighting, the method reconstructs the Boltzmann distribution at a desired temperature. We define a scalable performance metric, based on the Kullback–Leibler divergence and the total variation distance, to quantify convergence toward the exact Boltzmann distribution. DCQS is validated on one-dimensional Ising models with random couplings up to 124 124 qubits, where exact results are available through transfer-matrix methods. We then apply it to a higher-order spin-glass Hamiltonian with 156 156 qubits executed on IBM quantum processors. We show that classical sampling algorithms, including Metropolis-Hastings and the state-of-the-art low-temperature technique parallel tempering, require up to three orders of magnitude more samples to match the quality of DCQS, corresponding to an approximately 2 × 2\times runtime advantage. Boltzmann sampling underlies applications ranging from statistical physics to machine learning, yet classical algorithms exhibit exponentially slow convergence at low temperatures. Our results thus demonstrate a robust route toward scalable and efficient Boltzmann sampling on current quantum processors. I Introduction Sampling from complex probability distributions such as Boltzmann distributions is a fundamental challenge across various disciplines, including statistical physics Huang 1988 , molecular dynamics Frenkel and Smit 2002 ; Noé et al. 2019 , machine learning Ackley et al. 1985 and finance Glasserman 2003 . Traditional sampling methods based on Markov chain Monte Carlo (MCMC) Dongarra and Sullivan 2000 are widely used but often struggle with high-dimensional distributions and in the low-temperature regime, due to slow convergence and poor mixing. As illustrated in Fig. 1 (a), as the temperature tends to zero, the autocorrelation (mixing) time of the Metropolis-Hastings (MH) MCMC method with local updates diverges Newman and Barkema 1999 , making it impractical in this regime. Furthermore, in the strict zero-temperature limit, sampling from the Boltzmann distribution reduces to finding the ground state, which is, in the worst-case scenario, an NP-hard problem, as for spin glasses Barahona 1982 . The classical state-of-the-art technique for low-temperature sampling, sketched in Fig. 1 (b), is parallel tempering (PT) Hukushima and Nemoto 1996 ; Earl and Deem 2005 , also known as replica exchange. It mitigates slow mixing by running multiple MCMC simulations in parallel at different temperatures and periodically attempting exchanges between them. While this can prevent the diverging autocorrelation time and allows escaping local minima, it comes at the expense of running multiple auxiliary simulations. Recent years have seen growing interest in employing quantum computers to perform Boltzmann sampling, taking advantage of quantum rather than thermal fluctuations. Given their open nature, quantum annealers have been proposed as noisy Boltzmann samplers Vuffray et al. 2022 . To improve their accuracy, various adjustments have been proposed, such as introducing spurious hardware-dependent couplings Vuffray et al. 2022 , limiting the class of simulable systems Nelson et al. 2022 , focusing on the low-temperature regime Sandt and Spatschek 2023 , restricting the validity to the high-temperature regime Gyhm et al. , and using embedding schemes Teza et al. . Figure 1: Sketch of (a,b) classical methods for Boltzmann sampling and (c) digitized counterdiabatic quantum sampling (DCQS). (a) The classical Metropolis-Hasting algorithm suffers from a diverging mixing time at low temperatures, since it can get trapped in local minima. The black line pictorially represents the energy landscape of the classical Hamiltonian. Transitions to the local minima on the right are exponentially suppressed in the low-temperature limit. (b) Parallel tempering (PT) is a classical method that overcomes this limitation by running multiple “replicas” in parallel (one per row), each one at a different inverse temperature \beta_{i} . The left column represents the allowed transitions within the energy landscape (horizontal arrows), and between adjacent replicas (vertical arrows), while the right column shows the corresponding energy distributions. This leads to a better low-temperature sampling at the expense of an increased simulation cost. (c) DCQS relies on running quantum circuits with an iterative bias field strategy on a digital quantum computer (left), producing lower and lower energy states (center), which are then reweighed in order to reproduce low-temperature observables (right). Research on employing digital quantum computers as Boltzmann samplers remains limited and requires further conceptual advances, as the unitary evolution of closed quantum systems is, at first glance, not directly connected to the Boltzmann distribution that characterizes open-system statistics. Variational quantum imaginary time evolution has been proposed to train Boltzmann machines, and has been tested on an IBM quantum processing unit (QPU) IBM Quantum with up to 20 qubits Zoufal et al. . Specifically designed Hamiltonians, whose ground-state measurement outcomes mimic the Boltzmann distribution, have been proposed Wild et al. 2021 , as well as sampling states using quantum alternating operator ansatz Pelofske . However, the implementation of these approaches on QPUs has not been investigated. Alternatively, quantum sub-routines have been proposed to improve the quality of classical MCMC methods. For instance, the autocorrelation time of MCMC has been decreased using either quantum annealers Scriva et al. 2023 or digital quantum computers Layden et al. 2023 during the proposal step. We introduce digitized counterdiabatic quantum sampling (DCQS) to overcome the difficulty of classical methods in sampling Boltzmann distributions at low temperatures, the lack of scalable quantum algorithms suitable for digital noisy intermediate-scale quantum (NISQ) computers, and the scarcity of large-scale (100+ qubits) experimental validations. It is a NISQ-compatible quantum-classical algorithm based on counterdiabatic protocols Demirplak and Rice ; Berry 2009 ; del Campo 2013 , designed to efficiently sample low-temperature Boltzmann distributions and compute thermodynamic observables of classical Hamiltonians using digital quantum computers. The proposed method, illustrated in Fig. 1 (c), relies on the observation that, in the limit of infinitely slow driving, adiabatic quantum evolution preserves the system in its instantaneous ground state, thus yielding the zero-temperature limit of the Boltzmann distribution of the final Hamiltonian Gyhm et al. . In practice, however, coherent transitions due to non-adiabatic driving Landau 1932 ; Zener 1932 , noise in NISQ devices, discretization, and Trotter errors Lloyd 1996 lead to excitations in the final state Visuri et al. . We utilize these apparently undesired effects to efficiently obtain not only the ground state but also low-energy states that predominantly contribute to the low-temperature Boltzmann distribution. To turn this idea into a quantum algorithm compatible with NISQ devices, we employ counterdiabatic protocols with an adaptive bias field Cadavid et al. 2025 ; Romero et al. 2025 ; IBM Quantum 2025a to mimic the adiabatic behavior in shorter times, leading to low-depth quantum circuits that run within the device’s coherence time. Finally, we employ post-processing techniques and a reweighting scheme Sandt and Spatschek 2023 to compute low-temperature expectation values. We first validate our approach by computing thermal observables for the 1D Ising model with

Read the complete text