UC Berkeley

Vacuum Beam Studies of Ruthenium Etching

Ru is known to have two volatile oxidation products, RuO₃ and RuO₄, although the etch rate is negligible when Ru is exposed to an O₂ plasma discharge. The introduction of a small amount of additive gas, such as Cl₂, has been shown to increase the Ru etch rate sixfold. The reason for this dramatic shift in etching is poorly understood, primarily because it is difficult if not impossible to study plasma-surface interactions in a plasma environment. The unique capabilities of the beam system have made it possible to explore the mechanism of Ru etching. It has been shown that under 500 eV Ar+ ion bombardment, the addition of O radicals lowered the etch rate by a factor of 2.5. This process was relatively insensitive to temperature over the range studied (room temperature to ~175°C). It was also shown that O radicals alone spontaneously etched Ru at a very slow rate over the entire temperature range.

Statistical Analysis of Polysilicon Etching and Gate Profile Evolution in Dual-Doped Polysilicon Gates

Polysilicon gate etching for the 90nm lithography node and below requires extremely precise control of the gate CD and profile. Generally speaking, the current requirement for Gate CD control is that the 3 sigma should less than ~5 nm for all gates, including across the chip, across the wafer, wafer-to-wafer, lot-to-lot, and tool-to-tool variations. Similarly, for gate sidewall angle control, the 3 sigma angle variation should be less than ~1 degree, inclusive of all sources of variation. This is particularly challenging for technologies which employ dual-doped gates, since the chemistry and physics of the etching process induces a different profile evolution between gates with different doping.

The goal of this project was to identify a parameter space where the differences in gate profile evolution across different polysilicon dopant types were minimized. Blanket etch rates and patterned wafers were used to determine the effect of different gate etch process variables on the gate profile. The materials studied were undoped polysilicon and polysilicon that had been doped with P, As, Sb, and B. Prediction models were created for the blanket etch rate studies that were used to optimize the processing conditions and to propose some simple mechanisms that identify which species are adsorbed on the surface.