Silicon Oxynitride Process

The silicon oxynitride process incorporates nitrogen into a thermally grown gate oxide — producing SiON rather than pure SiO2 — most often by annealing the oxide in nitric oxide (NO) or nitrous oxide (N2O) rather than by depositing a separate film. It exists to solve a problem that pure oxide scaling couldn't: as gate oxides thinned into the low-nanometer range, boron from heavily doped PMOS gates started punching straight through them, and plain SiO2 had no way to stop it.

Thermal oxide grown
Standard dry O2 oxidation first
NO or N2O anneal
800-1050°C nitridation step
Nitrogen incorporationconcentrated near the Si/SiO2 interface, not uniform
Interfacial N pile-up
Blocks boron diffusion through the film
Raised dielectric constant
k rises from ~3.9 toward ~7 (pure Si3N4 limit)
Gate dielectric ready
Thicker physical film, same electrical thinness
Poly or metal gate deposit
Continues the standard gate flow

Boron penetration is a specific failure mode that pure oxide gate dielectrics ran straight into as PMOS gates thinned. A heavily boron-doped polysilicon PMOS gate, annealed at the high temperatures needed to activate the dopant, lets some of that boron diffuse straight through a sufficiently thin SiO2 gate oxide and into the channel underneath — shifting threshold voltage unpredictably and degrading the oxide's reliability. Nitrogen incorporated near the Si/SiO2 interface forms Si-N bonds that are far more resistant to boron diffusion than the Si-O network alone, creating an effective diffusion barrier without adding a separate film or a separate masking step — the nitrogen goes into the oxide that's already there.

Why Pure Oxide Fails and SiON Doesn't boron diffusion through a thin PMOS gate oxide, with and without interfacial nitrogen Pure SiO2 gate oxide Si channel SiO2 B-doped poly gate boron diffuses straight through channel Vt shifts, reliability degrades SiON gate oxide Si channel SiO2 bulk N-rich interface layer B-doped poly gate Si-N bonds block the path channel protected, Vt stable the nitrogen doesn't have to be uniform through the film — concentrated right at the interface is enough to stop the diffusion path

The same nitrogen incorporation also raises the film's dielectric constant, which bought gate-oxide scaling a few extra node generations before high-k became unavoidable. Pure SiO2 has a dielectric constant near 3.9; pure Si3N4 sits near 7. Partial nitrogen incorporation moves SiON somewhere between those two values, which means a SiON film can be grown physically thicker than an equivalent pure-oxide film while presenting the same electrical thinness (the same capacitance) to the transistor — and a physically thicker film leaks less gate current through direct tunneling than an ultra-thin pure oxide would. That leakage reduction, not boron blocking alone, is why SiON became the default gate dielectric industry-wide for roughly a decade of nodes before high-k/metal-gate finally took over below 45 nm, when even SiON's modest k-value boost could no longer keep leakage within budget.

Nitridation method and nitrogen profile are tuned as carefully as the oxide growth itself. A straight N2O anneal tends to pile nitrogen near the gate electrode interface as well as the channel interface, which is less effective at stopping boron than nitrogen concentrated specifically at the bottom interface; NO anneals, or plasma-based nitridation in more advanced flows, give tighter control over exactly where the nitrogen ends up. Too much nitrogen, or nitrogen in the wrong location, degrades channel carrier mobility by scattering carriers at the interface — so the nitridation process is a genuine optimization between boron blocking, dielectric constant gain, and mobility loss, not a simple "more nitrogen is better" knob.

SiON shows up as an anti-reflective coating too, but that's a related material doing an unrelated job, worth keeping separate. A silicon oxynitride film tuned for optical absorption at a lithography wavelength, deposited as a bottom anti-reflective coating (BARC) or dielectric ARC (DARC) under photoresist, shares the SiON name and composition family with the gate-dielectric version but is engineered entirely for refractive index and absorption coefficient at 193 nm or 248 nm, not for boron blocking or channel capacitance — a separate process optimized against a separate set of targets, even though both are, strictly, silicon oxynitride.

Read the silicon oxynitride process as a small compositional change — swapping some oxygen-silicon bonds for nitrogen-silicon ones — that solved two unrelated problems at once: it gave a thin gate oxide a diffusion barrier it didn't have before, and it raised the film's dielectric constant just enough to keep gate leakage in check for another several node generations, buying the industry time before a genuinely new class of high-k dielectric became unavoidable.

Take silicon oxynitride process further

Ask the copilot about this term, or have our engineers assess it against your process.