Ild0 Pre-Metal Dielectric Pmd Gap-Fill Planarization Process
Once the contact etch stop liner has been blanketed over the finished transistor, the entire gap between and around the gate stacks still has to be filled solid with dielectric before any contact hole can be cut, and that fill is the interlayer dielectric most engineers just call ILD0 or the pre-metal dielectric (PMD). This is not a thin conformal film like the liner underneath it; it is a thick gap-fill oxide, typically 400-800nm as deposited and planarized down to a controlled target thickness of roughly 150-300nm above the gate tops after polish, and its job is to physically occupy every nanometer of space between adjacent gate stacks, over the gates themselves, and above the source/drain regions so that the contact-hole litho and etch have a uniform, void-free dielectric volume to pattern into. The choice of deposition chemistry for that fill is dictated almost entirely by the aspect ratio of the gaps it has to fill, because the spacing between tightly packed gate stacks at advanced nodes can present aspect ratios well above what a simple conformal deposition can handle without leaving a seam or a void behind.
Historically, that gap-fill role was played by borophosphosilicate glass, BPSG, deposited by APCVD or low-temperature CVD and then subjected to a reflow anneal at roughly 800-900°C that let the glass physically soften and flow, rounding off the sharp corners left by the as-deposited film and improving planarity before any mechanical polish. BPSG reflow worked well on relaxed, large-pitch nodes where the thermal budget of an 800-900°C anneal late in the flow was tolerable, but as junctions became shallower and thermal budgets tightened, that reflow temperature became incompatible with preserving sharp, shallow source/drain profiles, and reflow-based planarization was phased out in favor of depositing a lower-temperature gap-fill oxide and relying on chemical-mechanical polishing (CMP) to do the planarization job that reflow used to do. Modern nodes typically use either high-density-plasma (HDP) oxide or plasma-enhanced TEOS (PE-TEOS), chosen for moderate-aspect-ratio gaps, or a flowable-CVD-style gap-fill such as the ozone-TEOS sub-atmospheric chemistry used in the sacvd-process when the gate pitch is tight enough that a sputter-and-redeposit film risks pinching off and leaving a void at the top of the gap before it fully fills from the bottom.
Which gap-fill chemistry gets chosen is a direct function of how aggressive the aspect ratio between neighboring gate stacks has become, and the three mainstream options trade thermal budget, void risk, and CMP behavior against each other in ways that track the node timeline closely.
| Property | BPSG reflow (legacy) | HDP oxide / PE-TEOS | SACVD / flowable oxide |
|---|---|---|---|
| Gap-fill at high aspect ratio | Poor without reflow softening; reflow itself limits minimum pitch | Good at moderate aspect ratio; sputter/etch/redeposit cycling needed above that | Best; bottom-up viscous flow fills narrow, deep gaps without pinching off |
| Thermal budget | High; reflow anneal ~800-900°C | Moderate; no reflow step required | Low to moderate; densify anneal well below BPSG reflow temperature |
| Particle / void risk | Low once reflowed, but reflow itself can shift dopant profiles | Moderate; seam or keyhole void risk rises sharply past its aspect-ratio limit | Low; flow mechanism specifically avoids pinch-off voids |
| CMP compatibility | Good; reflow pre-planarizes before polish | Good; standard oxide CMP slurries apply directly | Good; typically paired with a densify anneal before CMP |
| Adoption era / node | Large/relaxed-pitch nodes, phased out as junctions shallowed | Moderate-pitch nodes, still common where aspect ratio allows | Tight-pitch advanced nodes, same ozone-TEOS chemistry as sacvd-process |
CMP planarization after the gap-fill deposition is what actually sets the final, usable thickness of the ILD0 layer, and it has to do so uniformly across the die because contact-hole litho depends on a flat, predictable surface to focus and expose correctly. A dished or non-uniform ILD0 surface after polish translates directly into focus and depth-of-field problems for the contact litho step, and because the contact-hole etch is timed or endpointed against the known starting thickness of the dielectric above the CESL, any thickness variation left behind by an uneven polish risks either under-etching, leaving some contacts not fully open to the silicide, or over-etching elsewhere and punching through the CESL where the dielectric happened to be thinner than planned. The target thickness after CMP is chosen as a balance between leaving enough dielectric margin for a controllable, well-behaved etch and not adding so much oxide that the contact-hole etch has to run excessively long, since a longer etch increases the chance of CESL erosion wherever the liner is thinnest.
Read against the rest of this FEOL-to-contact arc, ILD0/PMD gap-fill is the step where the transistor stops being a collection of discrete, exposed structures and becomes a single embedded volume that the entire contact module then has to cut through. Gate formation built the stack; spacer formation set the sidewall geometry; source/drain epitaxy and silicide formation finished the active device and gave it a low-resistance contact surface; cesl-stress-liner-etch-stop-process then laid down the liner that both tunes channel strain and defines where every later etch has to stop; and ILD0/PMD gap-fill now buries that whole assembly in dielectric, using HDP oxide, PE-TEOS, or a SACVD-style flowable fill chosen by gap aspect ratio, historically finished with a BPSG reflow anneal and now almost always finished by CMP alone. Only once that dielectric volume is deposited void-free and planarized to a controlled thickness can the contact-hole litho and etch proceed, stopping cleanly on the CESL and clearing the way for tungsten-plug-contact-fill-process or cobalt-contact-fill-process to land the metal that finally closes the circuit from silicide to interconnect; a void anywhere in this fill, or a BPSG-era thermal budget applied to a shallow-junction node, turns this gap-fill step from a formality into the single most likely place for a contact-level defect to originate.