Microscopic and data-driven modeling and operation of thermal atomic layer etching of aluminum oxide thin films

Sparse identification modeling and predictive control of wafer temperature in an atomic layer etching reactor

“…The convective heat transfer originated from gas flow is also considered as a boundary condition on wafer. As demonstrated in a prior work (Yun et al, 2022a), the gas flow in the thermal ALE process is laminar and has small surface flow velocities for small reactor volumes (Ponraj et al, 2013). As a result, the convective heat transfer coefficient on the wafer is approximated as a fixed value where forced convection is negligible.…”

“…However, to ensure that complete byproduct removal is observed, the specific optimal operating temperature depends on the material of the substrate. For instance, aluminum oxide, Al 2 O 3 requires high operating temperatures of 573 K, which has been proposed experimentally (Lee et al, 2016) and studied in prior in silico modeling work (Yun et al, 2022a). In addition to reaching the chosen operating temperatures, it is important that the temperature of the substrate surface is uniform.…”

“…However, realistic thermal ALE processes often experience disturbances that disrupt the standard operating conditions of the process and introduce surface defects that result in product nonconformance, performance degradation, and lower productivity, depending on the magnitude of the disturbance. For example, a temperature fluctuation of 5 K will substantially change the half-cycle time, as observed in Yun et al (2022a). Therefore, precise temperature https://doi.org/10.1016/j.cherd.2023.12.024 Received 30 November 2023; Received in revised form 12 December 2023; Accepted 14 December 2023 2 F. Ou et al control is necessary to maintain the desired operating temperatures uniformly across the substrate surface for thermal ALE processes.…”

The 2022 Plasma Roadmap: low temperature plasma science and technology

“…Atomistic modeling of ALE processes can leverage significant advances that have been achieved in the application of computational modeling methods to ALD processes. For an example of this to thermal ALE of aluminum oxide thin films, the reader is referred to the work by Yun et al [72]. They describe for this problem the application of density functional theory (DFT) -based electronic structure calculations and evaluate possible reaction pathways and kinetic parameters using kinetic Monte Carlo methods.…”

Science-based, data-driven developments in plasma processing for material synthesis and device-integration technologies

“…A neural network was used to reduce computational resources and optimize the model-based feedback control using proportion integral (PI) loops. 277,278) The feed-forward ANN model was used to determine the optimal operating conditions for the thermal ALE process of aluminum oxide, as reported by Yu et al 279) The kMC algorithm was based on the calculated activation energies of half cycles of kinetic mechanisms for hydrofluoric acid (HF) and tri-methyl aluminum (TMA) using the DFT.…”