Aluminum Metal Etch

The Al metal etch is the structural opposite of everything Section 8 has covered so far: instead of carving an empty mold into dielectric for metal to fill afterward, it starts with a blanket aluminum film already deposited across the wafer and etches the unwanted metal directly away, leaving the wires themselves as the finished product of the etch step rather than a cavity waiting for one. This subtractive approach survives alongside copper damascene specifically at layers — certain upper global-routing levels, fuse layers, and especially bond pad metal — where aluminum's simpler integration, lower cost, and freedom from copper's diffusion-barrier requirements still outweigh its higher bulk resistivity, which is a real engineering tradeoff this project's M1 trench and via etch rounds deliberately set up as the mirror image of this one.

Chlorine-based chemistry is what makes aluminum etchable at all, and it is also the direct source of the one defect mode unique to this specific step among everything covered in this project: post-etch corrosion. Aluminum etches cleanly in a chlorine or boron-trichloride plasma, but that same chemistry leaves chlorine-containing residue on the newly exposed metal sidewalls, and if that residue sits exposed to atmospheric moisture for more than a short window — often specified in minutes, not hours — it reacts to form hydrochloric acid in situ, which corrodes the aluminum lines from the outside in; this is why Al metal etch process flows build an immediate post-etch rinse or passivation step directly into the tool sequence rather than treating cleaning as a separate, schedulable downstream operation the way most other etches in this project can.

A small, deliberate addition of copper into the aluminum film is what keeps these lines from failing under current stress, which is a genuinely counterintuitive detail given that copper is the metal this entire etch philosophy exists to avoid using directly. Pure aluminum interconnect lines are vulnerable to electromigration — current-driven atomic migration that opens voids and grows hillocks over years of operation — and alloying in roughly 0.5 percent copper measurably slows that grain-boundary diffusion process without meaningfully changing the subtractive etch chemistry needed to pattern the film; the resulting Al-Cu alloy is not a step toward damascene integration at all, it is a narrowly targeted reliability fix that borrows just enough of copper's atomic behavior to survive, while keeping every other aspect of the simpler subtractive process intact.

Sidewall redeposition and fence formation are the profile-control problem specific to subtractive metal etch, with no real equivalent in the damascene trenches this project has already covered. As aluminum etches, some of the removed material can redeposit onto adjacent photoresist sidewalls and survive the main etch step as a thin standing fence once the resist is stripped, which can bridge adjacent lines or interfere with the next layer's deposition if not cleared by a dedicated post-etch treatment; a damascene trench etch has no equivalent failure mode because it never removes bulk conductive metal in the first place, which is a direct structural consequence of the two approaches' opposite starting points — one etches a void, the other etches a solid.

Read through a finished-product-not-a-mold lens, the Al metal etch is where this project's back-end sequence briefly steps outside the damascene philosophy that governs most modern interconnect, trading copper's lower resistivity and better reliability margin for aluminum's simpler, cheaper, barrier-free integration — a tradeoff this specific etch step pays for immediately in the form of a corrosion clock that starts ticking the moment the plasma turns off.