ALD/ALE 2026 Session AF2-WeA: Modeling for ALD Processes II

Atomic layer deposition (ALD) processes span a wide range of length scales and operating pressures, from nano-scale semiconductor features to micrometer-scale structures in MEMS. This means that diffusion processes can range from free molecular flow to the transitional Knudsen regime, where conventional fluid models may require empirical corrections to accurately predict film conformality.

Direct Simulation Monte Carlo (DSMC) offers a reliable approach to the simulation of rarefied gas dynamics on a molecular level. The method has been extensively validated against experimental measurements in different applications including turbo-molecular pumps, spacecraft re-entry aerothermodynamics, and vacuum systems in general. Its ability to accurately resolve gas flow across a wide range of Knudsen numbers, makes it ideally suited for ALD applications spanning multiple flow regimes.

We apply the DSMC method using the open-source plasma simulation framework PICLas to simulate Al₂O₃ ALD conformality from TMA/H₂O in lateral high-aspect-ratio channels. The method is validated against experimental thickness profiles from Arts et al. (2019), then employed to evaluate the extended slope method by Gonsalves et al. (2024) in the transitional regime. DSMC enables seamless multi-scale coupling between reactor-level transport and feature-scale deposition, providing a framework for determining sticking coefficients across industrially relevant conditions. Additionally, PICLas offers the capability to investigate plasma-enhanced ALD processes in the future.

The heterogeneous integration of diverse material components is crucial for developing advanced, multi-functional microelectronic circuits. A significant challenge in this approach is the thermal boundary resistance at material interfaces, which can impede heat dissipation and compromise the performance and reliability of devices. As device sizes shrink and power densities rise, enhancing the interfacial thermal conductance (TBC) has become a critical aspect of thermal management. Recent research on GaN/SiC and AlGaN/Diamond interfaces suggests that introducing an interlayer material can effectively improve TBC.

However, the impact of different interlayer materials and their thickness on TBC is not well understood, and conventional theoretical models like the Acoustic Mismatch Model (AMM) and Diffusion Mismatch Model (DMM) are insufficient as they do not account for the critical role of interfacial states. To address this, our study focuses on the TBC of GaAs/Diamond and GaN/Diamond interfaces, both with and without interlayer materials. We employed advanced transport approaches, including the Non-Equilibrium Green's Function (NEGF) method and Reverse Non-Equilibrium Molecular Dynamics (RNEMD), to accurately calculate TBC incorporating the chemical reconstructions that happen at the interface.

Our calculations find that TBC through GaN/AlN/Diamond is significantly higher than GaN/Diamond. This can be explained in terms of much higher TBC associated with GaN/AlN and AlN/Diamond interfaces compared to the direct GaN/Diamond interface. This motivated a broader investigation using a series of interlayer materials (AlN, TiN, and TaN). We found that all interlayer materials improved the TBC for the GaN/Diamond interface, with TiN yielding the maximum improvement. Additionally, a nearly linear increase in TBC was observed when the TiN layer's thickness was increased from 3nm to 6nm, resulting in a 10% enhancement. Conversely, for the GaAs/Diamond interface, we did not observe a similar improvement in TBC, but the introduction of an AlN interlayer was found to improve the structural stability of the interface. This work provides guidance for future ALD experiments with carefully selected interlayer materials to optimize thermal management in next-generation electronic devices.