AxBAxB… pulsed atomic layer deposition: Numerical growth model and experiments

Abstract

Atomic layer deposition (ALD) is widely used for the fabrication of advanced semiconductor devices and related nanoscale structures. During ALD, large precursor doses (>1000 L per pulse) are often required to achieve surface saturation, of which only a small fraction is utilized in film growth while the rest is pumped from the system. Since the metal precursor constitutes a significant cost of ALD, strategies to enhance precursor utilization are essential for the scaling of ALD processes. In the precursor reaction step, precursor physisorption is restricted by steric hindrance (mA1) from ligands on the precursor molecules. On reaction, some of these ligands are removed as by-products resulting in chemisorbed species with reduced steric hindrance (mA1 → mA2, where mA2 < mA1) and some of the initially hindered surface reaction sites becoming accessible for further precursor physisorption. To utilize these additional reaction sites, we propose a generalized AxBAxB… pulsed deposition where the total precursor dose (ΦA) is introduced as multiple x (x > 1, x ∈ I) short-pulses rather than a single pulse. A numerical first-order surface reaction kinetics growth model is presented and applied to study the effect of AxBAxB… pulsed ALD on the growth per cycle (GPC). The model calculations predict higher GPC for AxBAxB… pulsing than with ABAB… deposition. In agreement with the model predictions, with AxBAxB… pulsed deposition, the GPC was found to increase by ∼46% for ZrN plasma enhanced ALD (PEALD), ∼49% for HfO2 PEALD, and ∼8% for thermal Al2O3 ALD with respect to conventional ABAB… pulsed growth.