Sequential Implant Device Structure Emitter Base Collector

A real transistor is never built from a single implant. A modern MOSFET source/drain alone stacks a shallow, low-energy extension implant, an angled halo implant beneath it, and a deep, high-energy source/drain implant — each at a different species, energy, and dose, each placed to do one specific electrical job, and all of them interacting with each other once annealing starts.

The sequencing decision — one shared anneal or several separate ones — is a real trade-off, not a formality. Annealing once after all implants are in place minimizes the total thermal budget the device sees, which keeps every profile as sharp as it was designed, since every implant diffuses by the same (minimal) amount. But some process flows genuinely require an intermediate anneal — for example, activating a base implant before a shallower emitter implant goes in, so the two don't interdiffuse into each other during a single combined bake. The decision is made profile by profile, based on which diffusion interactions the device can tolerate and which it can't.

Implants interact even when nobody intends them to. A halo implant is specifically placed to counter-dope the region just outside the source/drain junction, suppressing short-channel leakage — but during anneal, the halo's dopant and the source/drain's dopant interdiffuse at their shared boundary, partially compensating each other right where the halo is supposed to be doing its job. Getting the final, annealed profile to match the design intent means simulating that interaction up front, not just specifying each implant's as-implanted profile in isolation.

This is why TCAD process simulation exists as its own discipline. Predicting where five or six interacting, diffusing dopant profiles actually end up after a shared anneal — rather than where each one starts — is a coupled diffusion problem that's effectively impossible to get right by hand once more than two implants are stacked in the same small volume.