Quasiatomic layer etching of silicon oxide selective to silicon nitride in topographic structures using fluorocarbon plasmas

Abstract

The precision etch of dielectric materials (SiO2, Si3N4) in self-aligned contacts and self-aligned multiple patterning at the 10 nm technology node and beyond is required to solve/mitigate the trade-offs of selectivity, profile, and aspect ratio independent etching. Atomic layer etching (ALE) has significant potential to solve the trade-offs challenge. Unlike the halogenation step of atomic layer etching of silicon using chlorine (e.g., chlorine adsorption on c-Si), the fluorocarbon deposition step of ALE of nitride and oxide is not a self-limiting process. The argon ion bombardment step used to remove CF polymer layer and activated layer of oxide and nitride can be self-limiting if ion energy is kept below the sputter threshold. In this paper, the authors will discuss concurrent engineering approaches including both modeling and experimentation that can provide visibility to the parameters characterizing a viable process. The core of the approach is a new integrated chamber hybrid plasma equipment model-feature scale Monte Carlo feature profile model for silicon dioxide etch experiments intended to be selective to organic planarization layer masked silicon nitride structures. The experiments were conducted on a dual frequency capacitively coupled plasma source using a benchmark Ar/C4F6 chemistry for adsorption and an argon plasma step for desorption in the cyclic etch process. The concurrent engineering approach comprises stages of simulation development and prediction tests using both blanket wafer and patterned wafer data, and finally, process parameter optimization. Plasma parameters for each of the fluorocarbon layer adsorption and desorption etch steps are presented. The authors show how nitride/oxide etch selectivity can be optimized using saturation, “S”-curves, and detail pattern top clogging and its correlation with mask topography. S-curves describe etch/deposition rate trends as a function of desorption time in cyclic etch processes. The results show that clogging margin is a sensitive function of desorption time and feature topography.