Quantum annealing for materials
Alfredo Fiorentino, Nicola Marzari · 2026
Read Quantum annealing for materials by Alfredo Fiorentino, Nicola Marzari (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-2606-03405.txt.
Introduction
Fiorentino and Marzari present a new way to run quantum annealing for materials problems. The core task is finding the lowest point of a potential energy surface. That task matters for protein folding, clusters, and many other systems. The count of near-stable states can grow far too large to scan by hand. Quantum annealing uses quantum fluctuations to search, unlike heat in classical annealing.
Their method builds annealing into path-integral molecular dynamics. That framework samples the quantum nuclear density without handling full wavefunctions. The protocol keeps the ease of standard dynamics runs. It still performs well across a wide range of atomic systems. Tests use both simple force fields and machine-learned potentials. The tool can act as a global optimizer of the energy surface. It can also fold nuclear quantum effects straight into structure searches. That trait suits high-pressure hydrides and similar materials.
Catalog record
| CREATOR | Alfredo Fiorentino, Nicola Marzari |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 47 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.03405
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Quantum annealing for materials Alfredo Fiorentino, Nicola Marzari Abstract Finding the global minimum of a potential energy surface is a fundamental challenge in materials science, with applications ranging from protein folding to cluster physics and, more broadly, to systems in which the number of (meta)stable configurations grows prohibitively large. In recent decades, quantum annealing (QA) has emerged as a promising global optimization strategy, exploiting quantum fluctuations in contrast to the thermal fluctuations that drive its classical counterpart. Here, we introduce a novel implementation of QA based on path-integral molecular dynamics, an efficient and well-established framework for sampling the quantum nuclear density without the need to manipulate many-body wavefunctions explicitly. While retaining the flexibility and simplicity of molecular dynamics simulations, this quantum-annealing protocol delivers strong performance across a wide range of atomic systems, simulated by either empirical force fields or machine-learning interatomic potentials. The method can be used either as a global optimizer of the potential-energy surface, or as a quantum-informed structure-search strategy in which nuclear quantum effects are included directly in the optimization workflow—a feature particularly relevant for materials such as high-pressure hydrides. I Introduction The search for the global minimum (GM) of a complex potential energy surface (PES) is a central problem in physics, chemistry, and materials science. Many relevant systems—from biomolecules and glasses to atomic and molecular clusters—exhibit rugged energy landscapes characterized by an exponentially large number of metastable states. While the choice of an optimal optimization strategy generally depends on the specific problem, simulated annealing (SA) remains a time-honored and widely used approach 51 ; 36 , valued for its generality and flexibility. Inspired by metallurgical practices, SA mimics a melting phase followed by slow cooling, using thermal fluctuations to explore configurational space. However, thermal fluctuations are not the only possible driving force, and it is well known that SA can struggle in the presence of large energy barriers and multi-funnel PESs. In an effort to overcome these limitations, an alternative paradigm has been introduced: quantum annealing (QA) 33 ; 7 ; 16 ; 1 . Drawing inspiration from quantum mechanics, QA relies on the idea that quantum fluctuations enable the exploration of classically rare or forbidden configurations, and can therefore outperform SA in certain scenarios 55 ; 58 ; 56 ; 59 . QA has become a popular tool for optimizing discrete problems, such as disordered Ising models, with implementations on both classical computers 1 ; 33 ; 55 ; 58 ; 39 and quantum hardware—where it is often referred to as adiabatic quantum computing 16 ; 32 ; 48 ; 2 . By contrast, its application to multidimensional continuous systems, which is the focus of this work, has remained comparatively limited despite early successes 1 ; 19 ; 25 . A major obstacle lies in the need to represent and evolve a many-body wavefunction according to the time-dependent Schrödinger equation, either in real or imaginary time. Several strategies have been proposed to address this challenge. These approaches can be broadly divided into two categories. The first consists of assuming a specific ansatz for the wavefunction and evolving it according to the Schrödinger equation 1 . Alternatively, one may rely on stochastic techniques such as diffusion Monte Carlo (DMC) 19 or path-integral Monte Carlo (PIMC) 25 to sample the quantum nuclear density and thereby capture equilibrium nuclear quantum effects (NQEs), at the expense of abandoning exact quantum dynamics. Both routes require considerable care. In the former case, the ansatz must remain sufficiently flexible throughout the adiabatic evolution. In the latter, the choice of Monte Carlo moves and the associated fictitious dynamics can strongly influence—and even qualitatively alter—the performance of the algorithm 59 ; 31 . In this work, we propose a novel implementation of QA based on path-integral molecular dynamics (PIMD). While conceptually related to earlier QA-PIMC approaches 25 ; 56 , QA-PIMD preserves the simplicity and scalability of molecular dynamics simulations, while sampling the configurational space according to the path-integral approximation to the quantum nuclear density, which becomes exact in the limit of infinitely many beads and sufficient equilibration. When applied to standard optimization benchmarks such as Lennard–Jones (LJ) clusters 28 ; 47 ; 63 , the algorithm efficiently locates the global minimum for cluster sizes at least as large as those considered in previous QA studies 1 ; 19 ; 25 . Moreover, for these systems, QA-PIMD often shows a clear performance advantage over its classical counterpart, SA, reaching the reference minimum with shorter annealing schedules. This advantage becomes even more pronounced when the replica-pinned version of QA (RPQA) 25 is used, which consistently solves the challenging \mathrm{LJ}_{38} benchmark. Beyond its role as a global optimization tool, a quantum algorithm such as QA-PIMD is particularly appealing for materials science applications. For certain materials like high-pressure hydrides 12 ; 14 ; 46 ; 13 ; 50 ; 57 , the inclusion of NQEs—such as zero-point energy contributions—can modify the relative stability of competing structures 8 ; 13 . In such cases, a straightforward minimization of the PES may lead to incorrect structural predictions or phase diagrams. By contrast, an optimization strategy that inherently accounts for quantum nuclear effects can naturally identify physically meaningful structures and stability rankings. The remainder of the article is organized as follows. Section II presents the results and discussion. We first introduce the theoretical framework underlying quantum annealing and its path-integral molecular dynamics implementation. We then assess the ability of QA-PIMD to adiabatically follow the quantum nuclear density using an asymmetric double-well potential. Next, we compare the performance of QA-PIMD, its replica-pinned variant, and (classical) simulated annealing for Lennard-Jones clusters. We also demonstrate the applicability of the method to reconstruction problems in crystalline materials described by machine-learning interatomic potentials (MLIPs), a problem of practical relevance in experimental materials characterization. Then, the QA-PIMD is successfully applied to study structural transitions driven by NQEs, in particular for the superconducting high-pressure \mathrm{LaH}_{10} structure. Finally, section III concludes the paper with a summary and outlook. Technical and methodological details are provided in the Materials and Methods section. II Results and Discussion Quantum annealing framework Quantum annealing is a computational technique aimed at identifying low-energy configurations of complex systems by exploiting quantum fluctuations. The algorithm consists of adiabatically evolving the wavefunction \psi according to the time-dependent Schrödinger equation, written here in the position representation: \displaystyle i\hbar\frac{\partial}{\partial t}\psi(\mathbf{x},t)=\hat{H}(t)\psi(\mathbf{x},t) \displaystyle\hat{H}(t)=-\alpha(t)\sum_{i=1}^{N}\frac{\hbar^{2}}{2m_{i}}\nabla_{i}^{2}+U(\mathbf{x}_{1},\ldots,\mathbf{x}_{N}), (1) where U denotes the interaction potential of an N -body system, \hbar is the reduced Planck constant, m_{i} are the particle masses, and \mathbf{x} is a shorthand notation for the full set of particle coordinates, \mathbf{x}=\mathbf{x}_{1},\ldots,\mathbf{x}_{N} . The operator \nabla_{i}^{2} is the Laplacian with respect to \mathbf{x}_{i} . The positive annealing function \alpha(t) controls the relative weight of the kinetic and potential contributions. Equation ( II ) describes a real-time evolution; however, in practice, QA is often implemented in imaginary time rather than real time 1 ; 55 ; 58 ; 56 (effectively replacing i\partial_{t} with -\partial_{t} ). Numerical studies indicate that this choice can lead to a quantitative improvement in performance, although the asymptotic scaling of the residual energy with annealing time remains unchanged 58 . The annealing procedure is typically initialized in the ground state of a simpler Hamiltonian, after which \alpha(t) is gradually decreased