Blog

One-stop comprehensive solution provider

Blog | 2026-10-14

Silicon Nitride Deposition: Choosing the Right Precursor for LPCVD, PECVD, and ALD

DeshangChem
1 Views

Silicon Nitride Deposition: Choosing the Right Precursor for LPCVD, PECVD, and ALD

Silicon nitride does more jobs inside a modern device than almost any other film. It works as an etch stop, a passivation barrier, a CMP stop, a moisture block, and the spacer material that quietly determines transistor behavior. Yet the precursor question behind all these layers is often settled by habit. A recipe that worked for one nitride application gets copied to the next, and the film carries problems it never needed to have.

The precursor choices are genuinely different from each other, and each one buys you something and costs you something. The sensible way through them is by process regime, because that is how the decision actually gets made in a fab.

High-temperature LPCVD: where chlorine earns its keep

For batch furnaces running above roughly 700 °C, dichlorosilane with ammonia is the reference recipe. Dichlorosilane (DCS) holds that position for practical reasons: nitride grown from DCS shows better thickness uniformity across a boat than silane-based nitride, and the chlorine by-product chemistry keeps particle counts manageable through long campaign runs. For oxidation masks, hard masks, and any layer where a full furnace load has to come out uniform, DCS is still the default many engineers reach for first.

Monochlorosilane (MCS) is the less-discussed middle option. It carries more silicon per molecule than DCS, which translates into higher growth rate at a given temperature, while keeping the surface benefits of chlorine chemistry. Lines that need nitride throughput without giving up uniformity often qualify MCS alongside DCS and keep both on the gas panel for different load types.

Silane-based nitride still has a place. Silane (SiH4) reacts faster and at lower temperature than chlorosilane chemistry, which matters when the thermal budget cannot stretch even in a furnace. The trade-off is uniformity and a higher hydrogen content in the film, and that trade-off is precisely why chlorosilanes displaced silane in high-temperature nitride work over the years.

PECVD: fast films with a hydrogen bill

Below roughly 450 °C, plasma deposition takes over, and silane plus ammonia is the overwhelming standard here. The film that comes out is hydrogen-rich, with silicon-hydrogen and nitrogen-hydrogen bonds in significant density. For a final passivation layer that never sees high heat again, that is usually fine. Where it bites is when the film later goes through a hot step: hydrogen effuses, stress shifts, and wet etch rates drift. Anyone who has debugged an etch-stop layer that behaves differently after a subsequent anneal knows this failure mode well.

One useful thing to understand before anyone proposes a precursor swap for a PECVD problem: the plasma breaks the silicon source down almost completely, so switching between silicon sources rarely changes what the film looks like. The levers that matter at low temperature are plasma conditions, gas ratios, and post-deposition anneals, not the precursor chemistry.

ALD and the low-thermal-budget regime

When the nitride has to be thin, conformal in a deep structure, and deposited below 400 °C, ALD with aminosilanes is where the industry has settled.

Diisopropylaminosilane (DIPAS) is the reference for ALD silicon nitride. Paired with an ammonia or nitrogen plasma, it offers a stable, repeatable saturation window, and the resulting films hold nitrogen density well enough for gate spacers and etch-stop layers at advanced nodes. Its reactivity is moderate in the good sense: predictable dosing, clean purges, and film properties that respond to recipe changes in ways you can actually model.

BDEAS is best known for ALD oxide with ozone or oxygen plasma, but the same molecule can be driven toward nitride with a nitrogen-based plasma. That dual capability matters more than it sounds: many fabs standardize on one liquid precursor for both the oxide and nitride steps in a stack, because one delivery system and one qualification package beat two. Whether that logic applies to your stack depends on how tight the nitrogen stoichiometry specification is for the nitride layer, since oxide-optimized recipes do not always hit tight N content without rework.

Trimethylsilylamine (TSA) completes the shortlist. It is a highly volatile liquid, which simplifies dosing hardware, and its ammonia-like leaving group reacts readily on treated surfaces at low temperature. It is specified up to 8N purity on the supply side, which makes it attractive where contamination budgets are tightest. Film properties differ enough from DIPAS that a comparison run on the actual tool is worth the week it takes; datasheet numbers for plasma nitride vary more between tools than between precursors.

The chlorine family beyond deposition

Nitride process flows also consume chlorine chemistry outside the deposition step itself. Chlorosilane exhaust management, chamber cleans, and doping steps elsewhere in the line typically share infrastructure, and gases such as boron trichloride (BCl3) often ride on the same trusted supply relationship. This is one reason fabs tend to consolidate chlorosilane sourcing with a single qualified supplier rather than fragmenting the work across several.

Purity, packaging, and total cost of ownership

Nitride film properties are sensitive to moisture and metal impurities, so the purity grade is not an afterthought. Chlorosilanes on the site's grade sheet ship at 4N, silane at 6N, while the liquid aminosilanes run 7N and above. Packaging differs just as sharply: gases arrive in cylinders sized to draw duration, while liquids like BDEAS ship in 5- and 10-gallon containers that feed bubblers or direct liquid injection. When comparing precursors, the honest calculation includes cylinder change frequency, bubbler lifetime, scrubber load, and the cost of the qualification lot itself. A precursor that deposits faster but doubles the changeout cadence is rarely the bargain the datasheet suggests.

A short framework for the decision

  1. Fix the thermal budget first. Above 700 °C the chlorosilane route opens up; below 400 °C you are in aminosilane ALD territory almost by default.
  2. Decide how conformal the film must be. High-aspect-ratio spacers and liners push strongly toward ALD.
  3. Set the hydrogen tolerance. Post-deposition thermal steps argue against hydrogen-rich PECVD nitride.
  4. Check whether one liquid precursor can serve both oxide and nitride steps in the stack.
  5. Match purity grade and packaging to annual volume before signing anything.

Most flows land on a familiar set of answers: DCS or MCS for furnace nitride, silane for plasma nitride, and DIPAS, BDEAS, or TSA for ALD below 400 °C. The product pages linked above list purity grades, capacities, and packaging for each option, and a short conversation with the supplier's process team usually settles the last two candidates faster than another round of datasheets.

Related Articles