Amorphous Boron Nitride as an Ultrathin Copper Diffusion Barrier for Advanced Interconnects
Onurcan Kaya, Hyeongjoon Kim, Byeongkyu Kim, Thomas Galvani, Luigi Colombo, Mario Lanza, Hyeon-Jin Shin, Ivan Cole, Hyeon Suk Shin, Stephan Roche
TL;DR
This work investigates amorphous α-BN as an ultrathin Cu diffusion barrier for BEOL interconnects by integrating machine-learned MD with a Gaussian Approximation Potential and a simple Tight-Binding electronic model, complemented by PECVD growth and diffusion tests. It shows that barrier performance is highly sensitive to film morphology—dense, slowly cooled networks resist Cu diffusion up to high temperatures, while rapid quenching introduces defects and porosity that enable diffusion; increasing thickness further enhances barrier integrity but morphology remains the key driver. The study provides atomic-scale insights into how B–N bonding, sp coordination, and vacancy-like features influence diffusion pathways and dielectric properties, revealing mid-gap states that modestly affect epsilon and localization. Experimentally, 3–7 nm α-BN films demonstrate barrier behavior consistent with simulations, though crystallisation tendencies in thicker films pose integration challenges. Overall, amorphous α-BN offers a promising, low-dielectric-constant diffusion barrier for next-generation interconnects, with opportunities to optimize growth to maintain amorphousness while leveraging morphology-controlled barrier performance.
Abstract
This study focuses on amorphous boron nitride ($\rm α$-BN) as a novel diffusion barrier for advanced semiconductor technology, particularly addressing the critical challenge of copper diffusion in back-end-of-line (BEOL) interconnects. Owing to its ultralow dielectric constant and robust barrier properties, $\rm α$-BN is examined as an alternative to conventional low-k dielectrics. The investigation primarily employs theoretical modelling, using a Gaussian Approximation Potential, to simulate and understand the atomic-level interactions. This machine learning-based approach allows the performance of realistic simulations of amorphous structure of $\rm α$-BN, enabling the exploration of the impact of different film morphologies on barrier efficacy. Furthermore, we studied the electronic and optical properties of the films using a simple Tight-Binding model. In addition to the theoretical studies, we performed diffusion studies of copper through PECVD $\rm α$-BN on Si. The results from both the theoretical and experimental investigations highlight the potential of $\rm α$-BN as a highly effective diffusion barrier, suitable for integration in nanoelectronics. This research shows that $\rm α$-BN is a promising candidate for BEOL interconnects but also demonstrates the synergy of advanced computational models and experimental methods in material innovation for semiconductor applications.
