Etching kinetics and surface roughening of polysilicon and dielectric materials in inductively coupled plasma beams

Abstract

Plasma etching processes often roughen the feature sidewalls forming anisotropic striations. A clear understanding of the origin and control of sidewall roughening is extremely desirable, particularly at the gate level where variations in line width can adversely impact the electrical performance of the device. In addition, at the back end, feature sidewall roughness of the dielectric materials might degrade the resolution of contacts, interfere with the deposition of conformal liner materials, and make the process integration challengeable. In an inductively coupled plasma apparatus, the etching behavior on real feature sidewalls was simulated by etching blank films at grazing ion bombardment angles. The angular etching yields of polysilicon and dielectric materials in Ar, C12/Ar, and C4F8/Ar plasma beams were studied as a function of ion bombardment energy, ion bombardment angle, etching time, plasma pressure, and plasma composition. Interestingly, the effective neutral-to-ion flux ratio was the primary factor influencing the etching yield. A typical sputtering angular yield curve, with a peak around 600 off-normal angle, was formed at non-saturated etching regime, while an ion-enhanced-etching angular yield curve peaked around 650 was observed in the saturated etching regime.

(cont.) In Ar plasma, various films remained smooth after etching at normal angle but became rougher at grazing angles. Specifically, the striation structure formed at grazing angles could be either parallel or transverse to the beam impingement direction. Encouragingly, the sputtering caused roughening at different off-normal angles could be qualitatively explained by the corresponding angular dependent etching yield curve. In fluorocarbon plasmas, the roughening of thermal silicon dioxide and low-k coral films at grazing ion bombardment angles depended on both the etching kinetics and the etching chemistry. In particular, the surface roughened when the etching process was physical-sputtering like (at low neutral-to-ion flux ratios), even though the polymer deposition effect was trivial; when the etching kinetics was dominated by ion-enhanced etching (at high neutral-to-ion flux ratios), the roughening was mainly caused by the local polymer deposition effects. Moreover, surfaces could be etched without roughening at intermediate neutral-to-ion flux ratios and/or with the addition of oxygen to the discharge. The oxygen addition broadened the region over which etching without roughening can be performed.

(cont.) Additionally, the local-polymer-deposition effect can be used to explain the surface roughening of porous low-k films in fluorocarbon plasmas. Last, it was shown that RMS roughness is not adequate to represent the surface roughness on etched surfaces, especially when anisotropic striations exist. Instead, statistical methods such as the power spectral density and geostatistical analysis are capable of measuring the surface roughening in both vertical and lateral dimensions. In this way, the spatial variation of the streaks formed during plasma etching can be characterized quantitatively.