Computational Study of Amorphous Boron Nitride Compounds

Abstract

The performance limitations imposed by interconnects in scaled microelectronics necessitate new materials that offer both ultralow dielectric constants and robust diffusion-barrier capabilities at nanometre thickness. Amorphous boron nitride (α-BN) shows significant promise experimentally, yet its integration is hindered by an incomplete understanding of how growth conditions govern atomic-scale structure and consequently affect performance. This thesis addresses these questions through large-scale atomistic simulations, primarily employing machine learning interatomic potentials using the Gaussian Approximation Potential approach. These potentials were developed for α-BN compounds, incorporating interactions relevant to common contaminants (carbon, hydrogen and oxygen) and copper metallisation, and were trained on extensive density functional theory dataset. Using molecular dynamics, particularly melt–quench simulations to capture varying growth kinetics, this work establishes that α-BN morphology exists across a spectrum of structural disorder that is systematically controlled by growth conditions such as cooling rates, stoichiometry and contamination incorporation. Key structural descriptors, including mass density, homonuclear bonding and coordination defects, were systematically identified and quantified. The study demonstrates that these morphological features determine key functional properties. Structures with higher density, fewer coordination defects and reduced homonuclear bonding consistently exhibit superior thermal stability, enhanced mechanical properties, improved thermal conductivity and lower dielectric constants. Furthermore, the thesis reveals the atomistic mechanisms governing degradation under conditions relevant to device operation. Oxidative degradation proceeds through structure-dependent pathways, where denser, more relaxed amorphous films exhibit surface-limited reactions, whereas less relaxed, defect-rich films undergo bulk degradation initiated at homonuclear bonds and facilitated by nitrogen loss. These findings align with experimental trends observed at Université Claude Bernard Lyon 1 in France and The Australian National University in Australia. In parallel, simulations of copper diffusion identified failure mechanisms involving the thermal softening of defect-rich regions, subsequent void formation and eventual copper penetration via interconnected pathways, coupled with B and N migration into the metal layer. Barrier experiments conducted by collaborators at UNIST, SKKU and Samsung (South Korea) confirmed the predicted dependence of barrier effectiveness on film structure and thickness. This thesis provides a unified, morphology-driven understanding of α-BN behaviour, establishing clear links between growth conditions, structure and performance. It delivers mechanistic insights into degradation pathways and identifies key structural targets for the rational optimisation of α-BN as a multifunctional material in advanced interconnect and other nanoelectronic applications.

Date of Award	22 Jan 2026
Original language	English
Awarding Institution	Universitat Autònoma de Barcelona (UAB)
Supervisor	Stephan Roche (Director), Tu Le (Director) & Ivan Cole (Director)