Transient behavior in quasi-atomic layer etching of silicon dioxide and silicon nitride in fluorocarbon plasmas

Abstract

The mechanism for atomic layer etching (ALE) typically consists of two sequential self-limited half-reactions—passivation and ion bombardment—which provide unique control over the process. Some of the possible benefits of this control include increased selectivity, reduced plasma induced damage, improved uniformity and aspect ratio independence. To achieve the greatest benefit from ALE, both half-reactions should be fully self-limited. In the experimental demonstration of ALE of SiO2 using fluorocarbon plasmas, the passivation phase typically consists of deposition of fluoropolymer on the SiO2 surface. This passivation step is not a self-limited reaction as the final polymer thickness depends on the passivation time. In this paper, results are presented from a computational investigation of the ALE of SiO2 and Si3N4 focusing on the implications of this nonself-limited passivation phase. The polymer overlayer was found to be critically important to the ALE performance, providing the main mechanism for selectivity between SiO2 and Si3N4. The polymer overlayer acts as a fuel for etching SiO2, which couples the etch depth per ALE cycle to the passivation time. Due to the inherently pulsed nature of the ALE mechanism, the polymer overlayer requires a finite number of cycles to reach a pulsed periodic steady-state thickness. Since the thickness of the polymer overlayer largely determines selectivity between SiO2 and Si3N4, the initial formation of an overlayer results in a transient period at the beginning of etching where high selectivity may not be achieved. For the etching of thin films, or applications which require very high selectivity, this transient etching period may be a limiting factor. Results are also presented using ALE to etch high aspect ratio self-aligned contacts which could not be cleared using continuous plasma etching with similar ion energies and flux ratios.