Teos Process
The TEOS process deposits silicon dioxide from tetraethyl orthosilicate — Si(OC2H5)4, a liquid organosilicon precursor — instead of growing it thermally or depositing it from silane. It exists because thermal oxidation only works on exposed silicon, and silane-based oxide, while cheap and fast, can't conformally coat the increasingly tall, narrow topography that modern chips are built from. TEOS solves both problems by decomposing into oxide on whatever surface it touches, uniformly, regardless of what's underneath:
TEOS wins on step coverage because it decomposes everywhere the gas can reach, not just where line-of-sight allows. Silane-based oxide deposition tends to react close to the gas inlet and on upward-facing surfaces, leaving sidewalls and the bottoms of trenches thinner than the field — a problem that gets worse exactly as aspect ratios get taller. TEOS's organosilicon chemistry decomposes via a surface reaction that proceeds at comparable rates on horizontal and vertical surfaces alike, so a trench sidewall ends up coated nearly as thick as the open field above it. That conformality is the entire reason TEOS displaced silane oxide for interlayer dielectric gap-fill once feature aspect ratios stopped being forgiving.
LPCVD and PECVD TEOS trade thermal budget for throughput, the same way dry and wet oxidation trade quality for speed. LPCVD TEOS runs hot (roughly 650–750°C) in a low-pressure furnace, batch-processing dozens of wafers at once with excellent conformality and film quality, but that temperature rules it out anywhere dopant profiles or metal layers underneath can't tolerate the thermal budget. PECVD TEOS adds RF plasma energy to drive the same decomposition at 350–400°C, trading some conformality and film density for compatibility with back-end-of-line processing where aluminum or copper is already on the wafer and can't see furnace temperatures again.
TEOS oxide's job didn't end when copper took over interconnect — it became the material that makes copper's damascene process possible at all. Every damascene trench and via is etched into a TEOS-based (or similar organosilicate) interlayer dielectric specifically because its conformal, dense, and etch-predictable nature gives a reliable, planar-enough surface to pattern precisely — the metal scheme changed in 1997, but the dielectric doing the actual electrical isolation between those metal layers has stayed TEOS-based oxide (plain or carbon-doped for low-k) across both the aluminum and copper eras.
Hitting a precise TEOS fill thickness matters as much as the conformality itself, so production tools watch the film grow rather than trusting a fixed deposition time. The target sits in a real filmstack — not grown in isolation — and the recipe has to stop exactly where the next process step expects it to:
That target is tracked with the same kind of in-situ optical signal used anywhere a film's growth has to be stopped precisely — except here the signal climbs instead of falling, because the film is being built up, not etched away:
And isolated from the loop diagram, here is the actual trace a deposition tool follows as the film climbs toward target:
Read TEOS as proof that a precursor's chemistry, not just the final film composition, decides whether a deposition can keep up with shrinking geometry: silane and TEOS both ultimately deposit SiO2, but only TEOS's surface-limited decomposition coats a trench evenly enough to still be the industry's default interlayer dielectric decades after aspect ratios made silane's gas-phase shortcuts obsolete. And whether the film is being grown or an etch is clearing it away, the same instrumented discipline — measure the signal, don't trust the clock — is what actually hits the target.