Air Gap Spacer Removal
Air-gap spacer removal does the opposite of almost every other spacer step in this flow — instead of depositing and shaping a material that stays in the finished chip, it deliberately dissolves out a sacrificial oxide that was only ever meant to be temporary, leaving a literal void of air between the bitline and its neighbor because air itself is a better insulator than any solid dielectric a fab can deposit. Every other spacer in this list exists to protect, isolate, or set up a self-aligned etch; this one exists purely to be removed, and its entire purpose was already finished the moment it successfully held its shape through the processing steps that happened before this removal etch ran.
The reason this works at all is that nothing beats a vacuum, and air is close enough to one for the purpose of lowering capacitance between two closely spaced conductors. Every solid low-k dielectric material still has some dielectric constant above the absolute floor of roughly 1 that air and vacuum both sit near, so no amount of porous or fluorinated low-k material development can ever fully close that gap — replacing the dielectric entirely with air is the only way to approach that floor, which is exactly why this technique shows up specifically in the places on a DRAM array where bitline-to-bitline capacitance most directly limits read speed and power.
Getting this etch right means removing exactly the sacrificial material and nothing else, since the structures on either side of the gap now have to hold their own shape with nothing filling the space between them. The selectivity has to be extremely high against the bitline conductor stack itself and against whatever permanent spacer or support material is adjacent to the sacrificial oxide, because this etch runs after the bitline is already fully formed — there is no later step to fix a bitline damaged by an overly aggressive or insufficiently selective removal, and the resulting unsupported gap also has to be mechanically survivable through every process step that comes after it without collapsing.
This same sacrificial-material-then-removal pattern shows up again later in the DRAM flow at a much larger scale, in the capacitor module's mold dip-out. Both steps share the identical underlying logic — a material is deposited purely to hold geometry in place temporarily, then deliberately etched away once its structural job is done, leaving the real, permanent structure to stand on its own in the space that opens up — the air-gap spacer removal is simply the bitline-scale version of exactly the same trick the storage node module uses at a much larger scale and later in the flow.
Read air-gap spacer removal through a *the-material-was-never-meant-to-stay* lens rather than a *defect-risk* lens — unlike nearly every other etch in this flow, success here isn't measured by how well a feature survives, it's measured by how completely a feature that was never supposed to survive gets taken away, leaving only the air that does the real electrical work behind.