Silicon Nitride Process

The silicon nitride process deposits Si3N4 (or more precisely, hydrogen-containing SiNx:H — stoichiometric Si3N4 is the idealized limit, not what most fabs actually grow), a film dense enough to block diffusion, strong enough to etch-stop reliably, and — in one application that has nothing to do with either property — mechanically stressed enough to physically stretch or compress the silicon lattice underneath it and make transistors switch faster.

SiH4 + NH3 (or N2)
Precursor choice sets film properties
Deposition method
LPCVD furnace or PECVD chamber
Tune for the jobstress, density, and H content are all dials
LPCVD nitride
750-800°C, dense, inherently tensile
PECVD nitride
300-400°C, stress tunable by RF recipe
Barrier / etch stop / spacer
Diffusion block, SAC, contact etch stop
Strain liner
Tensile over NMOS, compressive over PMOS

LPCVD and PECVD nitride aren't interchangeable, because the deposition method itself sets the film's intrinsic mechanical stress. LPCVD nitride, grown hot from silane and ammonia in a low-pressure furnace, comes out dense and almost always tensile-stressed by its nature — useful as a hard, reliable barrier or etch-stop layer, but not a film a process engineer can dial to either sign of stress at will. PECVD nitride, grown cooler with RF plasma energy added to the reaction, gives real control: adjusting ion bombardment, deposition rate, and hydrogen content lets the same basic chemistry produce either a tensile or a compressive film, with stress magnitudes in the multiple-GPa range achievable in both directions. That tunability is what turned silicon nitride from a passive barrier material into an active strain-engineering tool.

Dual Stress Liner — Tensile Over NMOS, Compressive Over PMOS the same material, deposited twice with opposite intrinsic stress, strains each channel in the direction that actually helps it Silicon substrate NMOS gate tensile SiN liner pulls outward — stretches the channel tensile strain raises electron mobility PMOS gate compressive SiN liner pushes inward — compresses the channel compressive strain raises hole mobility a blanket deposit, a block mask, an etch-away, then the opposite-stress film — two full nitride depositions per strain-engineered node

That strain genuinely changes carrier mobility, which is exactly why dual stress liner (DSL) technology became a mainstream node-to-node performance lever rather than a laboratory curiosity. Mechanically stretching silicon's crystal lattice along the direction current flows measurably raises electron mobility in an NMOS channel; compressing it along the same direction raises hole mobility in a PMOS channel — the two carrier types respond to strain in opposite directions, which is exactly why a single stress sign can't help both transistor types at once. Building the stack requires a full nitride deposition over the whole wafer, a block mask protecting one transistor type, an etch to strip the nitride back off the unprotected type, then a second deposition of the opposite-stress film — real additional mask and etch cost, justified because strain engineering delivered a meaningful mobility (and therefore drive-current) gain at nodes where further pure geometric scaling was running out of straightforward options.

Beyond strain engineering, nitride's more familiar jobs — barrier, etch stop, and passivation — still account for most of the nitride actually deposited on a chip. A thin nitride liner under contacts and vias gives a reliable, high-selectivity etch stop the same way the SiC-barrier film does in copper levels; nitride's density makes it an effective diffusion barrier against mobile ions and moisture, which is why the final passivation layer on top of a finished die is typically nitride over oxide rather than oxide alone; and the bulk SiN layer in a gate sidewall spacer — the dominant mass of the spacer process's two-layer liner-plus-bulk stack — is this same material, chosen there for mechanical and etch-selectivity reasons unrelated to either diffusion blocking or strain.

Read the silicon nitride process as a material whose usefulness comes from how many independently tunable properties — density, hydrogen content, and above all intrinsic stress — live inside the same basic SiH4/NH3 chemistry. Most of those properties get used for the ordinary jobs of blocking, stopping, and sealing; one of them, intrinsic film stress, got repurposed into an entirely different kind of lever — reaching through the gate stack to physically deform the silicon channel underneath and extract more performance from a transistor whose geometry alone had nothing left to give.

Take silicon nitride process further

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