Scaling of Si and GaAs trench etch rates with aspect ratio, feature width, and substrate temperature

Abstract

The scaling of etch rates with feature dimensions is an important issue in the fabrication of microelectronic and photonic devices. Because etch rates depend on circuit layouts and design rules, considerable effort is spent to modify processes each time changes in design are made. Knowing how etch rates scale with design parameters should accelerate the introduction of new designs into manufacturing while minimizing the cost of doing so. Recently it has been shown that etch rates for a variety of conditions scale with the depth/width or aspect ratio and not on width or depth alone. While various mechanisms might be responsible for such scaling, isolating one mechanism from another is not straight forward. Nonetheless, it is important to understand the underlying mechanisms so that differences in etch chemistry and etch reactor design can be accounted for when scaling plasma processes from one design or reactor to another. To assess the effects of etch chemistry alone, the trench etch rates of Si and GaAs are compared under constant plasma conditions. Substrate temperature is varied to further assess the relative importance of surface vs transport phenomena during the etching process. At higher temperatures, both Si and GaAs trench etch rates scale only with aspect ratio in an Ar/Cl2 electron cyclotron resonance plasma. The results are consistent with an ion-neutral synergy model based on Langmuir adsorption kinetics where the scaling can be explained by the aspect ratio dependence of the neutral reactant transport into the trench: charging effects, ion shadowing, and Knudsen transport are not consistent with the data. At −45 °C, the etch rate no longer scales with aspect ratio alone, but now also depends on trench width. The data at this lower temperature are well described by a model incorporating the deposition of an etch inhibiting layer. While the trench etch rates for GaAs and Si scale in similar ways with aspect ratio and trench width, the dependence on feature dimensions of the Si etch rate is much stronger than that for GaAs at all substrate temperatures because the steady-state surface coverage of Cl is smaller for Si. This results in a greater sensitivity to the incoming, aspect ratio dependent, neutral fluxes. The implications of the models on the goal of aspect ratio independent etching are also discussed.

Gas-Surface Dynamics and Profile Evolution during Etching of Silicon

“…"Microtrenching" describes the appearance of a sharp groove near the foot of the etched feature [4,5] and has been attributed in models to reflection of ions from the sidewalls [4,5] and surface diffusion [8]. "Aspect ratio dependent etching" (ARDE) refers to the etch rate dependence on the relative spacing between features [1,6] and has been linked [6] to deposition of etch inhibitors, ion deflection, Knudsen transport of neutrals, ion and neutral shadowing effects, and surface diffusion. Finally, "undercutting," or loss of material beneath the mask that defines the circuit pattern, has been ascribed to reactant species which desorb from the bottom of the trench [7].…”

“…Attempts to model etch profile evolution [3][4][5][6][7][8] have had limited success in identifying the origins of commonly observed profile phenomena. "Microtrenching" describes the appearance of a sharp groove near the foot of the etched feature [4,5] and has been attributed in models to reflection of ions from the sidewalls [4,5] and surface diffusion [8].…”

Silicon photonic crystal cavities at near band-edge wavelengths

“…We optimised fabrication process steps to transfer ebeam lithography (EBL, at 30 kV) profiles into the thin (220 nm) Si device layer of the SOI chip with minimum distortions. Effective realisation of Si PhCs with small lattice constants (< 300 nm) depends on the availability of a lithographic mask that can withstand plasma etching long enough to transfer the patterns efficiently to the 220 nm Si layer 44 . We adopted a bilayer resist of polymethyl methacrylate (250 nm) and PECVD grown oxide layer (300 nm), which is less affected by proximity effects and can provide sufficient etch selectivity and anisotropy for the plasma etch steps under consideration.…”

Effect of Mask Erosion on Patterning of FePt for Heat-Assisted Magnetic Recording Media Using Embedded Mask Patterning

“…In part III we present the simulation result and discussion showing that how ion sputtering yield, etch selectivity and control of hard mask erosion play important role to the formation of isolated columnar FePt grains. The basic ion-neutral synergy etch model based on Langmuir isotherm adsorption kinetics is widely used to describe the plasma etch mechanism [9,11,12]. Bailey et al also proposed a synergy model with etch inhibition [11] using both neutral coverage and inhibition coverage as independent variables.…”

“…The basic ion-neutral synergy etch model based on Langmuir isotherm adsorption kinetics is widely used to describe the plasma etch mechanism [9,11,12]. Bailey et al also proposed a synergy model with etch inhibition [11] using both neutral coverage and inhibition coverage as independent variables. In order to account for ion sputtering induced re-deposition inside the narrow grain spacing in EMP patterns, we consider the neutral coverage is going to compete with re-deposition flux to occupy the un-covered surface site.…”