Machine learning for phase ordering dynamics of charge density waves - arXiv preprint

Aug 5, 2026

Upload by Patrick

A machine learning framework enables large-scale dynamical simulations of charge density wave (CDW) states by replacing costly electronic-force calculations with a neural-network force-field. The model leverages locality, predicting efficient local electronic forces from neighborhood configurations and structural inputs. This supports linear-complexity dynamical simulations of CDWs. Applied to the Holstein model, simulations reveal CDW domain growth that deviates from the Allen-Cahn law for Ising-type order-parameter coarsening, attributed to complex domain-wall structure.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Introduction

Background on charge density waves (CDWs)

Ultrafast dynamics and open questions

Challenges of full quantum dynamical simulations

Locality-based ML framework for adiabatic CDW dynamics

Relation to ML interatomic potentials

arXiv :2303 .03493v1 [ cond-mat .str-el ] 6 Mar 2023 Machine learning for phase ordering dynamics of charge density waves Chen Cheng, 1, 􀀃 Sheng Zhang, 1 and Gia-Wei Chern 1, y 1 Department of Physics, University of Virginia, Charlottesville, VA 22904, USA (Dated: March 8, 2023) We present a machine learning (ML) framework for large-scale dynamical simulations of charge density wave (CDW) states. The charge modulation in a CDW state is often accompanied by a concomitant structural distortion, and the adiabatic evolution of a CDW order is governed by the dynamics of the lattice distortion. Calculation of the electronic contribution to the driving forces, however, is computationally very expensive for large systems. Assuming the principle of locality for electron systems, a neural-network model is developed to accurately and e􀀎ciently predict local electronic forces with input from neighborhood con􀀌gurations. Importantly, the ML model makes possible a linear complexity algorithm for dynamical simulations of CDWs. As a demonstration, we apply our approach to investigate the phase ordering dynamics of the Holstein model, a canonical system of CDW order. Our large-scale simulations uncover an intriguing growth of the CDW domains that deviates signi􀀌cantly from the expected Allen-Cahn law for phase ordering of Isingtype order parameter 􀀌eld. This anomalous domain-growth could be attributed to the complex structure of domain-walls in this system. Our work highlights the promising potential of ML-based force-􀀌eld models for dynamical simulations of functional electronic materials. I. INTRODUCTIN A charge density wave (CDW) is a periodic modulation of electron charge density which breaks translational symmetries of the system [1{5] . In many materials, the charge density modulation is accompanied by a structural distortion due to electron-lattice couplings [1, 2] . Indeed, while CDW order could originate from pure electronic mechanisms [6, 7], electron-lattice couplings playa crucial role in the emergence of the charge modulation in most CDW materials. For example, the well-known Peierls transition, which describes the instability of a one-dimensional (1D) metal towards the formation of agapped CDW state, is caused by a periodic lattice distortion that opens a gap at the Fermi points of a partially 􀀌lled band [8] . The collective nature of charge and lattice dynamics in CDW states leads to several interesting phenomena such as nonlinear electrical conduction, giant dielectric response, and multi-stable conducting states, to name a few [4, 9 , 10] . A reinvigorated interest in the CDW physics was recently sparked by the synthesis of quasi-two-dimensional (2D) materials such as transition metal dichalcogenides [11{16] . The research on CDW phases has largely focused on their thermodynamic, electronic, and transport properties, as well as their interplay with other symmetrybreaking phases, such as superconductivity. A new frontier of research, enabled by the advent of ultrafast technology, is the formation and coherent dynamics of CDW orders. Such experiments aim to study the electronic and structural dynamics of CDW materials under shortpulse excitations with femtoseconds or picoseconds resolution [17{21] . Despite enormous interest in the ultrafast dynamics of CDW systems, their dynamical behaviors in the adiabatic limit, such as the phase-ordering kinetics, have yet to be systematically investigated, especially on the theoretical side. Important issues such as whether the coarsening of CDW domains follows well-established power laws or exhibits dynamical scaling remain open. Understanding the adiabatic dynamics will not only serve as important references for the ultrafast dynamics, but also shed light on the morphology of CDW states, especially in the presence of impurities and strains [22{24] . A full quantum dynamical simulation of the coarsening dynamics of CDW is, however, a di􀀎cult multi-scale task. On the one

FAQs

What is the main goal of the machine learning framework in this work?

To enable large-scale dynamical simulations of charge density wave states by accurately predicting local electronic driving forces without expensive quantum calculations at every time step.

How does the method achieve computational efficiency for large CDW systems?

It uses the locality (nearsightedness) principle, training a neural network to infer local electronic forces from neighborhood structural configurations, leading to scalable (linear-complexity) dynamics.

What key result is observed when applying the approach to the Holstein model?

The simulations show anomalous CDW domain growth that significantly deviates from the expected Allen-Cahn law for phase ordering in Ising-type order parameters.

Aug 5, 2026

Upload by Patrick