PE1-1.4 Etching, Ion Implantation & Diffusion
Standard semiconductor process technology — written September 2026
What this is and why it exists
Lithography decides where things happen. This topic covers what actually happens there: material removed, and dopant atoms put in.
Both halves have the same shape of argument. There is an older method that is simple and imprecise. There is a newer one that costs more and gives you control over what matters.
The vocabulary
- Isotropic etching — removal at the same rate in every direction.
- Anisotropic etching — removal much faster in one direction than the others.
- Undercut — material removed sideways beneath the mask edge.
- Selectivity — how much faster the etch attacks the target than the mask or the layer beneath.
- Plasma etching — etching with an energised gas rather than a liquid.
- Ion implantation — firing dopant ions into the wafer at high energy.
- Range — the average depth an implanted ion reaches.
- Dose — how many ions are delivered per unit area.
- Annealing — heating the wafer to repair implantation damage and activate the dopant.
- Diffusion — doping by heating so that atoms spread in from the surface.
The mental model
Etching first, and the whole distinction turns on direction.
Liquid chemistry attacks in every direction at the same rate. So while it cuts down through the layer, it also cuts sideways beneath the edge of the mask. That undercut is roughly as wide as the layer is thick. For a thick layer and a large feature, nobody minds. For a feature no wider than the layer is deep, the undercut destroys it.
Plasma etching escapes that. Ions in the plasma are accelerated towards the wafer, so they arrive travelling downwards. The etching is then much faster downwards than sideways, and the sidewalls come out close to vertical. Vertical sidewalls are what let features be packed closely together, so this is the enabling step for small geometries.
Reactive plasma etching combines both mechanisms. The gas is chosen so that it reacts chemically with the target material as well as arriving with momentum. That gives both directionality and selectivity, which is the combination a process needs.
Electrochemical etching sits beside these as a specialised tool, using a current to drive the reaction and gaining more control over rate and selectivity.
Now doping, and again there are two methods.
Diffusion is the older one. Heat the wafer with a source of dopant at its surface and the atoms wander inward. Two limiting cases give the standard results. If the surface concentration is held constant throughout, the profile is one standard function. If instead a fixed quantity is deposited first and then driven in, the profile is a different one. It spreads and flattens as it goes. Which case applies depends entirely on whether the surface is being replenished.
The problem with diffusion is that one knob controls two things. Temperature and time set both how many atoms go in and how deep they go, and you cannot choose the two independently.
Ion implantation separates them. Dopant ions are accelerated and fired into the wafer. The energy sets the depth, because a faster ion travels further before stopping. The beam current and the time set the dose, because you are literally counting the charge that arrives. Two knobs, two results, and that independence is why implantation replaced diffusion for most doping.
The depth is statistical rather than exact. Each ion follows its own path through the crystal and stops where it happens to stop. The result is a distribution about an average range. That is a different way of thinking from diffusion profiles, and it is worth noticing the change.
Implantation also breaks things. Ions arrive fast enough to knock silicon atoms out of their lattice sites, leaving a damaged crystal. A dopant atom sitting between lattice sites does not act as a dopant at all. Annealing fixes both. Heating lets the crystal reorder itself and lets the dopant atoms settle onto proper sites, where they become electrically active. The heat also moves the dopant a little further, which has to be accounted for in the design.
The machine itself explains where the control and the cost come from. Ions are produced, selected by mass so only the wanted species arrives, accelerated, and scanned across the wafer. Beam current and scanning uniformity are the parameters that decide whether the dose is the same everywhere.
What you should now be able to explain or do
Explain undercut and why it limits wet etching to large features. Say how a plasma etch achieves directionality and why vertical sidewalls matter. Compare the two limiting diffusion cases and say what distinguishes them. Explain why implantation controls depth and dose independently while diffusion cannot. Say what annealing repairs and what side effect it has.
Check yourself
Why can a wet etch not make small features?
It removes material sideways as fast as downwards. The undercut beneath the mask is about as wide as the layer is thick, which destroys narrow features.
How does a plasma etch become directional?
Ions are accelerated towards the wafer, so they arrive travelling downwards. They remove material far faster downwards than sideways.
Why did implantation replace diffusion for most doping?
It sets depth and dose independently, through energy and beam current. In diffusion, temperature and time control both together.
What does annealing do after an implant?
It repairs the crystal damage the ions caused, and it moves dopant atoms onto lattice sites so they become electrically active.
What distinguishes the two limiting diffusion cases?
Whether the surface concentration is held constant. A replenished surface gives one profile; a fixed quantity driven in gives another that spreads and flattens.
Go deeper
We haven't checked most of these for screen reader use yet.
Back to Etching, Ion Implantation & Diffusion: work through the checklist