Nickel Silicide Process
The nickel silicide process is also the clearest example of "salicide" — self-aligned silicide — a generic module that works with titanium, cobalt, or nickel largely the same way, with the material swapped for the phase-stability reasons the titanium, cobalt, and nickel process articles each cover. What makes it "self-aligned" is worth examining on its own: it's a contact-forming step that needs zero additional lithography masks, which is unusual enough to be worth understanding mechanically rather than taking for granted.
"Self-aligned" means the silicide forms exactly where it's needed without a single additional masking or lithography step, and the mechanism behind that is pure materials chemistry, not geometry. A blanket metal film deposited after the gate and spacer are already in place covers everything indiscriminately — the exposed silicon at the gate top and source/drain, and the dielectric spacer and field oxide right alongside it. During the silicide anneal, the metal reacts with silicon wherever silicon is directly exposed, forming the silicide compound there; over the spacer and field oxide, where the metal sits on a dielectric rather than silicon, no reaction occurs at all, because there's no silicon for the metal to combine with. The pattern that results — silicide precisely on the gate and source/drain, nothing anywhere else — falls directly out of where silicon happens to be exposed, which is exactly the pattern the gate and spacer process already defined several steps earlier.
The selective wet strip that removes unreacted metal afterward is the other half of what makes the whole scheme reliable, and it depends on a chemistry that can tell silicide apart from the metal it was made from. After the first anneal, unreacted nickel still sits untouched on top of the spacer and field oxide — it didn't react there, but it's still physically present and would short every gate to its neighbors if left in place. A wet etch chemistry chosen specifically to attack unreacted nickel metal while leaving the already-formed nickel silicide essentially untouched strips that excess metal away cleanly, which only works because the silicide compound's chemical behavior toward that etchant is different enough from the pure metal's to give real etch selectivity, not just a difference in degree.
Running the anneal in two separate steps, rather than one, is what gives nickel silicide a second chance to land on its best phase before the unreacted metal is removed. A first rapid thermal anneal (RTA1) forms the silicide and lets the wet strip clear away whatever metal didn't react; a second anneal (RTA2), run afterward with the excess metal already gone, can then push the already-formed silicide the rest of the way toward its lowest-resistance phase without the complication of unreacted metal still present on the wafer, and without over-driving the phase transformation that the nickel process's NiSi-to-NiSi2 problem describes — the split-anneal structure is itself part of how that phase risk gets managed in practice, not just a redundant extra step.
Read the nickel silicide process as a reminder that "self-aligned" is a chemistry trick wearing a geometry-sounding name: the alignment comes entirely from silicon being present in some places and absent in others, a pattern the gate and spacer formation steps already committed to several process steps earlier, and the silicide module simply reads that existing pattern back out — first by letting the metal react only where it finds silicon, then by stripping away everywhere it didn't.