Machine learning molecular dynamics study of thermal boundary resistance between barrierless interconnect metals and SiO2 interlayer dielectric

S Hashimoto and Y Nishimura and T Watanabe, JAPANESE JOURNAL OF APPLIED PHYSICS, 64, 04SP03 (2025).

DOI: 10.35848/1347-4065/adb9f2

Employing machine learning molecular dynamics (MD) calculation, we evaluated thermal boundary resistance (TBR) between barrierless interconnect unitary metals and SiO2 interlayer dielectric (SiO2-ILD). Our non-equilibrium MD calculation with machine learning interatomic potential revealed that the TBR of the W/SiO2 interface is lower than that of the Ru/SiO2 and Mo/SiO2 interface. Greater overlap of phonon density of states (DOS) in the W/SiO2 interface than that of the Ru/SiO2 and the Mo/SiO2 interface, resulting in the lower TBR. SiO2-induced lattice strain and disorder in W contribute to the greater overlap of the phonon DOS in the W/SiO2 interface. Furthermore, greater adhesion energy which is influenced by interfacial bond strength at the W/SiO2 interface, also contributes to the lower TBR. These findings highlight the importance of interfacial properties on phonon thermal transport through barrierless interconnect metal/ILD interfaces in deeply-scaled logic nodes.

Return to Publications page