journal article·2026-04-09Open access

Unveiling thermal transport properties of defective β − G a 2 O 3 through machine learning potentials

Yang Su, Jin Yan, Meiyang Yu, Chen Shen, Yuhao Fu, Tianhang Zhou, Lijun Zhang

Physical Review Materials· Volume 10· Issue 4

Abstract

The intrinsically low thermal conductivity of β−G{a}2{O}3 poses a major challenge for high-power and high-frequency electronic applications. This issue becomes more severe in the presence of defects, which further suppress heat dissipation, exacerbate self-heating, and degrade device performance. In this work, we develop an efficient machine learning potential (MLP) based on a deep neural network model for accurately describing pristine and point-defective β−G{a}2{O}3 . Using equilibrium molecular dynamics (EMD) simulations, we quantify the reduction in thermal conductivity induced by intrinsic point defects. Our results demonstrate that Ga interstitials exert the strongest suppression, decreasing the thermal conductivity by 79.5%. Moreover, Ga-related defects generally have a more pronounced impact than O-related defects. This behavior originates from the enhanced vibrations of weakly bonded Ga atoms, increased phonon anharmonicity, and a substantial reduction in the group velocity of low-frequency phonons. These results provide atomic-level insight into thermal transport in defective β−G{a}2{O}3 , offering guidance for thermal-management strategies and establishing a general workflow for investigating thermal physics in complex semiconductor materials.
Source: crossref · Quality: candidate unverified · Source record ↗

Research context

OpenAlex topics: Ga2O3 and related materials · Machine Learning in Materials Science · GaN-based semiconductor devices and materials

SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.

Affiliations

Jilin University

Technical University of Darmstadt

China University of Petroleum

Funding

National Natural Science Foundation of China · Jilin University

References (59)

Show source-provided references
1. 10.1063/1.506284110.1063/1.5062841 ↗
2. 10.1038/s41467-022-31664-y10.1038/s41467-022-31664-y ↗
3. 10.1021/acs.jpclett.1c0139310.1021/acs.jpclett.1c01393 ↗
4. 10.1039/C7TC02221B10.1039/C7TC02221B ↗
5. 10.1002/aelm.20160050110.1002/aelm.201600501 ↗
6. 10.1063/1.500694110.1063/1.5006941 ↗
7. 10.1063/1.505457310.1063/1.5054573 ↗
8. 10.1063/1.491607810.1063/1.4916078 ↗
9. 10.1016/j.jcrysgro.2014.07.02110.1016/j.jcrysgro.2014.07.021 ↗
10. 10.1088/0268-1242/30/2/02400610.1088/0268-1242/30/2/024006 ↗
11. 10.1063/1.492774210.1063/1.4927742 ↗
12. 10.1116/6.000187010.1116/6.0001870 ↗
13. 10.1021/acsomega.1c0636710.1021/acsomega.1c06367 ↗
14. 10.1016/j.mtphys.2024.10163210.1016/j.mtphys.2024.101632 ↗
15. 10.1103/PhysRevMaterials.3.02380410.1103/PhysRevMaterials.3.023804 ↗
16. 10.1016/j.mtphys.2020.10018110.1016/j.mtphys.2020.100181 ↗
17. 10.1039/D4CP02584A10.1039/D4CP02584A ↗
18. 10.1039/D3TA00845B10.1039/D3TA00845B ↗
19. 10.1103/PhysRevApplied.21.02404310.1103/PhysRevApplied.21.024043 ↗
20. 10.1038/s41467-024-46806-710.1038/s41467-024-46806-7 ↗
21. 10.1103/ndlz-hxwq10.1103/ndlz-hxwq ↗
22. 10.1103/PhysRevB.108.18420310.1103/PhysRevB.108.184203 ↗
23. 10.1103/PhysRevMaterials.9.02380110.1103/PhysRevMaterials.9.023801 ↗
24. 10.1038/s41467-025-62450-110.1038/s41467-025-62450-1 ↗
25. 10.1038/s41524-023-00974-010.1038/s41524-023-00974-0 ↗
26. 10.1016/j.commatsci.2024.11340910.1016/j.commatsci.2024.113409 ↗
27. 10.1016/j.cossms.2025.10121410.1016/j.cossms.2025.101214 ↗
28. 10.1038/s43246-023-00390-310.1038/s43246-023-00390-3 ↗
29. 10.1088/1361-648X/abcf6110.1088/1361-648X/abcf61 ↗
30. 10.1063/5.002505110.1063/5.0025051 ↗
31. 10.1063/5.018585410.1063/5.0185854 ↗
32. 10.1103/PhysRevX.9.04102710.1103/PhysRevX.9.041027 ↗
33. 10.1016/j.jallcom.2021.16066510.1016/j.jallcom.2021.160665 ↗
34. 10.1103/PhysRevLett.77.386510.1103/PhysRevLett.77.3865 ↗
35. 10.1103/PhysRevB.54.1116910.1103/PhysRevB.54.11169 ↗
36. 10.1103/PhysRevB.50.1795310.1103/PhysRevB.50.17953 ↗
37. 10.1038/s41524-023-01117-110.1038/s41524-023-01117-1 ↗
38. 10.1107/S010827019501640410.1107/S0108270195016404 ↗
39. 10.1016/j.scriptamat.2015.07.02110.1016/j.scriptamat.2015.07.021 ↗
40. 10.1016/j.cpc.2014.02.01510.1016/j.cpc.2014.02.015 ↗
41. 10.1103/PhysRevB.90.09411710.1103/PhysRevB.90.094117 ↗
42. 10.35848/1882-0786/ad0ba810.35848/1882-0786/ad0ba8 ↗
43. 10.1039/C8CP05139A10.1039/C8CP05139A ↗
44. See Supplemental Material at http://link.aps.org/supplemental/10.1103/fdrt-2chf for parameter testing of thermal conductivity calculated by BTE methods and EMD simulations; comparison of prediction results of GAP potential and DFT on defect structure test datasets; thermal conductivity of β−Ga2O3 along different crystal orientations at 300 K as a function of correlation time using the Green-Kubo methods with the MLP; the trend of thermal conductivity of different defect β−Ga2O3 structure with defect concentration; the cumulative thermal conductivity of β−Ga2O3 and its different defect structure; electron localization function (ELF) and projected crystal orbital Hamiltonian population (pCOHP) and of (a) VGa and (b) Oi; comparison of thermal conductivity between randomly distributed vacancy structures and supercell periodic amplification structures at 300 K; comparison of coherent thermal conductivity contribution and particle-like channel thermal conductivity contribution at 300 K; calculated phonon mode heat capacity Cλ of β−Ga2O3 and its different defect structure; calculated integration of the pCOHP (ICOHP) of β−Ga2O3 and its different defect structure.
45. 10.1103/PhysRevB.111.09430810.1103/PhysRevB.111.094308 ↗
46. 10.1103/2n3p-hz6210.1103/2n3p-hz62 ↗
47. 10.1016/j.cpc.2018.03.01610.1016/j.cpc.2018.03.016 ↗
48. 10.1006/jcph.1995.103910.1006/jcph.1995.1039 ↗
49. 10.1103/PhysRevB.65.14430610.1103/PhysRevB.65.144306 ↗
50. https://gitlab.abinit.org/xuhe/unfolding.git.
51. 10.1063/5.017709310.1063/5.0177093 ↗
52. 10.1021/acs.jpcc.3c0286610.1021/acs.jpcc.3c02866 ↗
53. 10.1063/5.016532010.1063/5.0165320 ↗
54. 10.1039/D1TC04531H10.1039/D1TC04531H ↗
55. J. M. Ziman · 2007
56. 10.1103/PhysRevB.89.18430410.1103/PhysRevB.89.184304 ↗
57. 10.1103/PhysRevB.91.14430410.1103/PhysRevB.91.144304 ↗
58. 10.1002/jcc.2430010.1002/jcc.24300 ↗
59. 10.1103/PhysRevB.109.23520210.1103/PhysRevB.109.235202 ↗