Gate Spacer RIE Etch Profile Control

Spacer formation lays down a conformal nitride film and etches it back, but it is the shape that etch-back leaves behind, not just the spacer's bare presence, that the rest of the transistor flow depends on, and that shape is set entirely by how anisotropic the reactive-ion etch (RIE) is. A conformal liner deposit, typically 5-15nm of LPCVD or PECVD silicon nitride, or a thin oxide liner under a thicker nitride in a bilayer stack, coats the gate top, the vertical gate sidewall, and the flat source/drain silicon with roughly equal thickness. An anisotropic RIE step then strips that film back, but only by making the etch strongly directional: ion bombardment accelerated by the plasma sheath field drives reactive species almost straight down onto the wafer, so the etch rate along the vertical direction runs many times higher than the etch rate along any lateral direction, typically a 10:1 to 30:1 vertical-to-lateral ratio for a well-tuned nitride spacer etch. On a flat surface, that straight-down ion flux clears film at the full vertical rate; on the near-vertical gate sidewall, the same ion flux arrives at a grazing angle and barely etches at all, so material builds up there in relative terms even as the flat regions clear, and the film left clinging to the sidewall becomes the self-aligned spacer, with its final width set mostly by the original deposited film thickness rather than by the etch itself. Typical finished spacer widths run 10-40nm depending on the node and the offset the downstream source/drain implant or recess etch needs, and getting that number right the first time, without a separate trim step, is the entire economic case for using RIE instead of an isotropic wet or dry etch here.

The etch itself runs in two distinct phases, a main etch and a over-etch, because clearing the flat regions completely without wrecking the sidewall or the silicon underneath takes more care than a single fixed-time step can deliver. The main etch runs at a higher ion energy, commonly in the 100-300eV range for a nitride spacer RIE, and is timed or optical-emission-endpointed to stop just as the bulk of the nitride film clears off the horizontal gate-top and source/drain surfaces, tracked by watching the plasma's emission spectrum for the characteristic drop in nitrogen or fluorine byproduct signal as the nitride-to-oxide (or nitride-to-silicon) interface is approached. Stopping exactly there is deliberately conservative, because etch rate and film thickness both vary by a few percent across a wafer and from lot to lot, so some nitride residue is expected to remain on the flat surfaces right at main-etch endpoint. A short over-etch, typically running 20-40% of the main-etch time, follows to clear that residue completely; it runs at a gentler condition, often lower ion energy or a more passivating chemistry mix, specifically so it finishes the clearing job on the flat surfaces without aggressively eating into the vertical spacer or punching into the exposed source/drain silicon once the nitride there is gone. Getting the main-etch/over-etch split right is a tuning exercise against two failure directions at once: too little over-etch leaves nitride stringers behind, too much over-etch erodes the spacer and recesses the silicon, and the acceptable window narrows as gate pitch shrinks and the etch has less margin to work with.

Conformal spacer liner deposit
5-15nm LPCVD/PECVD nitride, or oxide/nitride bilayer, blanket on all surfaces
Anisotropic RIE main etchion-driven, vertical rate 10-30x lateral rate
Directional ion bombardment
Clears gate-top and flat S/D silicon fastest; sidewall film survives at grazing incidence
Endpoint detection
OES signal drop near nitride/oxide or nitride/Si interface
Over-etch20-40% of main-etch time, gentler condition
Clear residual flat-surface film
Without eroding sidewall spacer or recessing underlying Si
Final vertical spacer profile
Sets up S/D recess etch boundary for epitaxy and gate-to-silicide spacing

The four ways this etch can land, ideal versus three distinct failure modes, are the real process-control story, because each one hands the downstream modules a different starting geometry. An ideal profile leaves a clean, near-vertical sidewall with the flat surfaces fully cleared and no silicon loss, giving the source/drain recess etch a sharply defined boundary to etch against. A "foot" profile leaves a thin wedge of residual nitride or oxide clinging to the base of the spacer where it meets the flat silicon, usually from under-etching or non-uniform film thickness across the wafer; that residual foot physically blocks part of the intended source/drain recess etch footprint, deforming the sigma-cavity shape the strained source/drain epitaxy process depends on and weakening the strain transfer into the channel. An undercut or sidewall-erosion profile goes the other way, over-etching or over-trimming the spacer until it is thinner or shorter than intended, exposing more of the gate edge than the design calls for; that exposed edge risks silicide bridging during the later salicide step, where excess nickel or cobalt silicide creeping too close to the gate can short gate to source/drain, and it also lets the source/drain epi recess etch encroach closer to the channel than intended. A silicon-recess or substrate-loss profile is the most damaging outcome, where an overly aggressive over-etch keeps punching into the source/drain silicon itself once the nitride there has cleared, gouging the surface before the epitaxial recess and refill process ever begins and seeding crystal defects that the later clean step cannot fully repair.

Profile outcomeVisual signatureLikely root causeDownstream consequence
Ideal vertical profileClean sidewall, flat surfaces fully cleared, no Si lossCorrectly tuned main-etch/over-etch split and endpointSharp, well-defined boundary for source/drain recess etch and epitaxy
FootResidual nitride/oxide wedge at spacer baseUnder-etch, non-uniform film thickness, insufficient over-etchDeforms sigma-cavity recess shape, weakens strain transfer into channel
Undercut / sidewall erosionSpacer thinner or shorter than designExcessive over-etch time or energy, aggressive chemistryExposes gate edge, risks silicide bridging, lets epi recess encroach on channel
Si recess / substrate lossGouged or roughened source/drain silicon surfaceOver-etch continues too long after nitride clearsSeeds crystal defects before epitaxy, degrades strain and junction quality
Gate Spacer RIE: Anisotropic Mechanism and the Four Ways It Can Land Directional ion flux clears the flat surfaces while the sidewall film survives at grazing incidence Anisotropic RIE (used here) Vertical etch rate 10-30x the lateral rate: flat regions clear, sidewall film survives Isotropic etch (not used here) Etches all directions equally: sidewall spacer film is removed too Four spacer etch outcomes Same anisotropic etch, four different results depending on main-etch/over-etch tuning 1 · Ideal Clean vertical sidewall, flat Si fully cleared 2 · Foot Residual nitride wedge clings to spacer base 3 · Undercut Spacer thin/short, gate edge exposed 4 · Si recess Gouged source/drain silicon surface Gate Correct spacer nitride Eroded / undercut spacer Silicon loss / damage Spacer profile sets the boundary the source/drain recess etch and silicide proximity both inherit Main etch to endpoint, then a 20-40% over-etch, is the standard control window

This spacer profile is not a cosmetic detail; it is the geometric boundary condition that the source/drain recess etch for strained-channel epitaxy and the later silicide formation step both inherit without any further say in the matter. The sigma-cavity recess etch that opens the way for strained SiGe or Si:C epitaxial source/drains is a timed and chemistry-tuned etch run directly against the spacer's outer edge; if that edge sits where the main-etch/over-etch recipe intended, the recess lands where the strain-engineering process was designed to put it, but a foot residue or an undercut spacer shifts that boundary and reshapes the cavity the epitaxial growth has to fill. Gate-to-contact spacing and the parasitic Miller capacitance between the gate and the source/drain contact are set by the same spacer width, since silicide formation grows its ohmic contact right up to wherever the spacer's outer face happens to be; an undercut spacer that exposes extra gate edge is exactly the condition that risks silicide creeping into bridging distance of the gate, while a correctly held spacer width keeps that proximity, and the parasitic capacitance that comes with it, within the value the circuit design assumed.

Read against the rest of this batch, spacer etch profile control is the quiet precursor step that both the strained source/drain epitaxy module and silicide formation depend on before either can do its own job. Spacer deposition and etch-back as a general process already established that a self-aligned sidewall film can exist with no separate lithography mask; this narrower look at the anisotropic RIE physics explains why that film ends up with a specific vertical shape rather than just existing, and why the main-etch/over-etch split, not merely "etching the spacer," is the real control knob. Get that profile right, and source/drain epitaxy recesses the silicon exactly where the sigma-cavity was designed to form, and silicide formation lands its contact at the gate-to-contact spacing the circuit was designed around; get it wrong in any of the three failure directions, and both of those downstream modules inherit a geometry problem they have no way to correct on their own.

Take gate spacer RIE etch profile control further

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