Contact Etch Stop Layer Cesl Stress Liner Process

Once silicide formation has turned the exposed gate top and source/drain silicon into a low-resistance ohmic contact, a conformal silicon nitride film goes down directly over the entire finished transistor, and that film does two jobs at once rather than one. The contact etch stop layer (CESL) is deposited by plasma-enhanced chemical vapor deposition (PECVD) in a thickness of roughly 20-40nm, blanketing the gate, spacer, and silicided source/drain surfaces before the interlayer dielectric (ILD0) oxide fills the space above it. Its named function is right there in the name: when the contact-hole etch later has to punch straight down through hundreds of nanometers of ILD0 oxide to reach the source/drain, that etch needs a backstop material with very different etch chemistry so the plasma can be timed or endpointed to stop cleanly on the liner instead of over-etching through it and gouging into the silicide or the silicon underneath. But PECVD silicon nitride is not etch-stop-neutral the way an inert spacer layer would be; the same deposition step that builds the film also locks in a specific intrinsic mechanical stress, and tuning that stress on purpose turns the CESL into a second, independent strain-engineering lever layered on top of whatever strain the embedded source/drain epitaxy process already put into the channel.

That intrinsic stress is controlled almost entirely by the plasma deposition conditions rather than by film thickness or composition changes, which is what makes CESL stress tunable without redesigning the film. Raising ion bombardment energy and silane flow during PECVD increases Si-H and N-H bond content and changes the film's bonding network in a way that leaves the nitride highly tensile, typically in the +1 to +1.5 GPa range, and a tensile liner sitting directly over an NMOS channel pulls the silicon lattice underneath it into tension along the current-flow direction, which raises electron mobility. Running the opposite deposition recipe, lower ion energy with a higher-density plasma that favors a different bonding configuration, produces a highly compressive film, typically -1.5 to -3 GPa and higher in aggressive dual-liner schemes, and a compressive liner over a PMOS channel compresses the lattice in a way that raises hole mobility. Because NMOS and PMOS want opposite stress signs for the same mobility benefit, the two device types cannot share one liner stress value; either the liner's stress is modulated locally over each device type on a single film, or two physically separate liners are used, one over each polarity, in what the industry calls a dual-stress-liner (DSL) process.

Silicide formation complete
Gate and source/drain silicon converted to low-resistance NiSi/CoSi contact
Conformal CESL nitride depositionPECVD, 20-40nm, blanket over gate/spacer/silicide
Tensile recipe over NMOS
+1 to +1.5 GPa, high ion energy/silane flow
Compressive recipe over PMOS
-1.5 to -3+ GPa, lower ion energy, dense plasma
Dual-stress-liner integrationlitho mask + selective etch + strip + second liner deposit, or single-liner local stress modulation
Liner removed over one device type
Replaced with opposite-stress film for independently tuned strain per polarity
Finished CESL in place over both device types
Tensile over NMOS, compressive over PMOS
ILD0 contact-hole etchOxide etch chemistry selective to nitride
Etch stops cleanly on CESL
High SiO2:SiN etch selectivity prevents punch-through into silicide
Liner punched through locally
Only at contact-hole locations, rest of CESL stays intact as strain source
Ready for plug fill
Tungsten-plug-contact-fill-process or cobalt-contact-fill-process lands through the opening

Whichever integration scheme is used, the liner's job as etch stop and the liner's job as strain source have to coexist without one undermining the other, which is why the comparison between the two stress polarities matters as much as the comparison between integration schemes.

PropertyTensile CESL (NMOS)Compressive CESL (PMOS)
Typical stress magnitude+1 to +1.5 GPa tensile-1.5 to -3+ GPa compressive (higher-order DSL schemes push further)
Deposition condition driverHigher ion bombardment energy and silane flow raise Si-H/N-H bond contentLower ion energy, denser plasma, different bonding network favors compressive film
Primary mobility benefitElectron mobility boost from tensile strain along current-flow directionHole mobility boost from compressive strain along current-flow direction
Integration complexitySingle-liner selective-stress modulation is simpler/cheaper but gives weaker, shared tuningFull dual-liner litho+etch+strip+redeposit is costlier but gives stronger, independently tunable strain per device type
CESL Dual-Stress-Liner Process and the Contact Etch-Stop Handoff One conformal nitride film is both a mechanical strain source and the backstop every contact etch stops on 1 · After silicide Gate + S/D silicided 2 · Blanket CESL PECVD nitride, 20-40nm 3 · Dual-stress split NMOS tensile PMOS compr. 4 · Contact etch stops Oxide etch selective to CESL 5 · W/Co fill Plug lands on silicide Everywhere away from a contact hole the CESL stays intact, continuing its strain-engineering job for the life of the device Only at contact-hole locations is it locally punched through, after the oxide etch stops cleanly on it CESL stress vector vs. source/drain epitaxy stress vector Two independent, geometrically distinct strain mechanisms acting on the same channel CESL: vertical, from above CESL stress applied downward over the whole channel Epi: lateral, from the sides Embedded S/D epi stress pushes laterally into the channel Tensile CESL (NMOS) Compressive CESL (PMOS) Embedded S/D epi (lateral strain) W/Co plug CESL stress and S/D epi stress are independent, additive strain mechanisms, not duplicates of each other Same liner that boosts mobility also gates every contact etch that follows it

That vertical-versus-lateral distinction is the whole reason CESL stress counts as a second strain mechanism rather than a restatement of the first. The source-drain epitaxy strained-channel process pushes strain into the channel laterally, from the embedded SiGe or Si:C regions sitting beside the channel at the source and drain edges, acting through the crystal lattice mismatch at those two side boundaries. CESL stress is applied from directly above, as a blanket mechanical load bearing down on the gate and the entire channel region at once, so a tensile or compressive liner changes the strain state of the channel through an entirely different geometric path than the epitaxy does. Because the two mechanisms act along different vectors, their effects add rather than substitute for one another, which is why advanced nodes deploy both: embedded source/drain epitaxy for the lateral strain budget, and a tuned or dual CESL for an additional vertical strain budget layered on top, each independently tunable and each contributing its own share of the total mobility gain.

The two integration paths for getting that stress split between NMOS and PMOS trade cost for strain strength in a straightforward way. A single-liner, selectively-stress-modulated approach deposits one film and uses local plasma treatment, such as a masked ion implant or UV cure that relaxes stress in exposed regions, to push one device type's liner stress toward neutral or the opposite sign without physically removing and redepositing anything; it is cheaper and adds fewer mask and etch steps, but the achievable stress range on each device type is narrower because both started from the same base film. A full dual-stress-liner approach deposits the first liner blanket, then uses a lithography mask to protect one device type while a wet or dry etch strips the liner off the other, followed by a second PECVD deposition of the opposite-stress film and a final strip of the mask; this adds a complete litho-etch-strip-redeposit cycle, which is markedly more complex and costly, but it lets each device type's liner be optimized independently at its own stress extreme rather than compromising between two conflicting targets on a single film.

Once the dual- or single-liner stress profile is locked in, the liner has to switch roles instantly the moment the contact module starts, and that handoff is where the etch-stop function earns its name. The ILD0 contact-hole etch runs an oxide-selective chemistry that etches silicon dioxide far faster than silicon nitride, so as the etch front reaches the CESL at the bottom of each contact hole, the etch rate drops sharply and the process can be timed or endpointed to stop right there, leaving the nitride liner intact everywhere except at the contact-hole locations themselves, where a short, separate nitride-opening step punches through just that local patch. Everywhere else, away from the contacts, the CESL stays fully intact and continues doing its strain-engineering job for the life of the device; only at the handful of contact locations is it locally sacrificed to open a path down to the silicide. Once that local opening exists, tungsten-plug-contact-fill-process or cobalt-contact-fill-process deposits the metal that fills the hole and makes electrical contact to the silicide below, meaning the same film that boosted channel mobility for the whole transistor also determined, by its etch selectivity, exactly where and how cleanly that metal plug could land.

Read against the rest of this contact-prep arc, CESL deposition is both the last strain-engineering lever in the flow and the literal gatekeeper for the contact module that follows it. Gate formation established the high-k metal gate stack; spacer formation and its anisotropic RIE etch-back set the self-aligned sidewall geometry; source/drain epitaxy added the first, lateral strain mechanism through embedded SiGe or Si:C; silicide formation turned the exposed silicon into a low-resistance ohmic contact; CESL deposition then adds a second, vertical strain mechanism independent of the first while simultaneously serving as the etch-stop material the entire contact-hole etch depends on to avoid punching through into the silicide it sits on top of; and tungsten-plug-contact-fill-process or cobalt-contact-fill-process finishes the arc by filling the opening the contact etch made through that same liner. No step after silicide formation can bypass the CESL, which makes it simultaneously the last chance to tune channel strain and the first hard constraint the contact module has to etch through correctly.

Take contact etch stop layer CESL stress liner process further

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