Machine-learning potential for phonon transport in AlN with defects in multiple charge states - Research highlights

Aug 4, 2026

Upload by Riley

Understanding phonon transport in defect-laden AlN is crucial for improving device performance in solid-state lighting and power electronics. The work develops a Behler–Parrinello-type machine-learning potential with ab initio accuracy, extended to treat multiple charge states of defects. Model validation uses phonon bands, three-phonon anharmonicity, isotope and phonon-defect scattering rates, and thermal conductivity. Results show defect- and charge-state-dependent phonon-property changes, with V3N+ producing the strongest scattering and V1N+ the weakest. Structural distortions from defects are found to be significant for elastic scattering rates.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages11

Reading Time28 min

AccessFree to download

Introduction

Motivation: AlN thermal and electronic properties

Role of phonon-defect scattering

Limits of mass-only perturbation models

First-principles BTE+DFT and defect-charge-state gap

Study context: prior DFT on charged vacancies

Machine-learning potential for phonon transport in AlN with defects in multiple charge states Ying Dou , 1 Koji Shimizu ,2, 3 Jesús Carrete ,4 Hiroshi Fujioka, 1 and Satoshi Watanabe 2 1 Institute of Industrial Science, The University of Tokyo, 4-6-1 Komaba, Meguro, Tokyo 153-8505, Japan 2 Department of Materials Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan 3 Research Center for Computational Design of Advanced Functional Materials (CD-FMat), National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba Central 2, 1-1-1 Umezono, Tsukuba, Ibaraki 305-8568, Japan 4 Instituto de Nanociencia y Materiales deAragón (INMA), CSIC-Universidad de Zaragoza, Zaragoza 50009, Spain (Received 13 August 2024; revised 16 January 2025; accepted 6 February 2025; published 6 March 2025) Understanding phonon transport properties in defect-laden AlN is important for their device applications. Here, we construct a machine-learning potential to describe phonon transport with ab initio accuracy in pristine and defect-laden AlN, following the template of Behler-Parrinello-type neural network potentials (NNPs) but extending them to consider multiple charge states of defects. The high accuracy of our NNP in predicting secondand third-order interatomic force constants is demonstrated through calculations of phonon bands, three-phononanharmonic, phonon-isotope and phonon-defect scattering rates, and thermal conductivities. In particular, our NNP accurately describes the difference in phonon-related properties among various native defects and among different charge states of the defects. They reveal that the phonon-defect scattering rates induced by V3N+ are the largest, followed by V3A−l, and that V1N+ is the least effective scatterer. This is further conﬁrmed by the magnitude of the respective depressions of the thermal conductivity of AlN. Our ﬁndings reveal the signiﬁcance of the contribution from structural distortions induced by defects to the elastic scattering rates. The present work shows the usefulness of our NNP scheme to cost-efﬁciently study phonon transport in partially disordered crystalline phases containing charged defects. DOI: 10.1103/PhysRevMaterials.9.034601 I. INTRODUCTION Aluminum nitride (AlN) belongs to a rare class of materials having both a large electronic band gap (∼ 6. 1eV) and a large thermal conductivity. Based on these properties, AlN plays a key role in solid-state lighting and modern power electronics. Owing to a small lattice mismatch, AlN is also widely used as a buffer during GaN growth in devices such as power high-electron mobility transistors (HEMTs) . Semiconductor devices are sensitive to heat dissipation, and AlN-related semiconductors are no exception: especially under high-power and high-temperature operation conditions, heat dissipation is extremely important for AlNrelated semiconductors to guarantee their performance and reliability. Phonon-defect scattering is one of the main factors reducing the thermal conductivity in semiconductor materials, as defects with low formation energies are inevitably introduced during crystalline growth. Computationally, the strength of phonon scattering by defects is often underestimated by only taking into account small perturbations of the dynamical matrix coming from the defect mass difference [1,2] . Published by the American Physical Society under the terms of the Creative Commons Attribution 4 .0 International license. Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI. Speciﬁcally, in thermal conductivity calculations of AlNrelated alloys and thin ﬁlms, the behavior of acoustic-phonon scattering by defects has been mainly explained on the basis of the mass-difference effect [3,4] . In studies using an analytical model based on a simpliﬁed Boltzmann transport equation (BTE) and the Debye approxi

FAQs

What is the main goal of the study on AlN?

To build and validate a machine-learning interatomic potential that can accurately describe phonon transport in pristine and defect-laden AlN, including defects in multiple charge states.

How is the machine-learning potential validated?

By comparing predictions with phonon-related calculations such as phonon bands, three-phonon anharmonic effects, phonon-isotope and phonon-defect scattering rates, and thermal conductivity.

Which charged defect produces the strongest phonon-defect scattering, and which is the weakest?

V3N+ yields the largest phonon-defect scattering rates, while V1N+ is the least effective scatterer.

Aug 4, 2026

Upload by Riley