Post-Etch via Trench Ash Strip Clean
Once the dual-damascene via and trench etch lands on the cap, the freshly cut sidewalls and via bottom are not clean dielectric and copper — they are coated with etch byproduct polymer, a fluorocarbon-rich residue deposited by the same CxFy/CHF3/C4F8-based chemistry that cut the low-k IMD, plus whatever photoresist or hard-mask remnants survived the etch, and all of that has to come off before barrier/seed deposition without attacking the exposed low-k sidewalls or corroding any copper the via punched down onto. The industry-standard sequence runs a dry plasma ash first to strip the bulk of the polymer and any remaining organic hard mask, then a wet chemical clean to pull the thin residual film and ionic contamination the ash cannot reach, because neither step alone gets the via/trench surface clean enough for void-free barrier/seed coverage: ash alone leaves fluorine- and metal-containing residue in the via bottom that shows up later as high contact resistance, while wet clean alone cannot touch a polymer layer that thick without excessive dielectric attack or impractically long dip times. The two failure modes this step has to manage simultaneously are low-k sidewall damage — oxidizing or reducing plasma species pulling carbon out of a carbon-doped oxide (CDO) film and leaving it hygroscopic — and copper-bottom corrosion or galvanic attack wherever the via already punched through the etch-stop cap onto a prior Cu line, so the ash chemistry, the wet clean chemistry, and their order all get chosen around whichever risk dominates for a given via/trench geometry and exposed-Cu condition.
| O2 plasma ash | H2/N2 forming-gas ash | Dilute HF wet clean | Semi-aqueous wet clean | Proprietary residue remover | |
|---|---|---|---|---|---|
| Primary role | Fast bulk strip of photoresist and thick fluorocarbon polymer | Lower-damage bulk strip, favored once low-k is exposed | Removes thin oxide/residual film, dissolves fluorinated byproduct salts | Removes polymer and particulates with less aqueous attack on exposed Cu | Formulated specifically to lift post-etch residue without touching CDO carbon or bare Cu |
| Low-k damage risk | High — oxidizing O radicals deplete methyl (Si-CH3) termination, k can rise from ~2.5 toward 3.0+ | Low — reducing H2 chemistry avoids the oxidation pathway, the standard "damage-free" ash | Low at typical <1% HF concentration and short dip, but repeated dips accumulate carbon loss | Low to moderate, better than straight aqueous HF for exposed low-k | Low by design, but efficacy and damage trade off against formulation cost |
| Cu-bottom corrosion risk | Low — dry plasma does not directly attack exposed Cu absent halogen residue | Low — reducing chemistry is generally Cu-compatible | Moderate — bare HF can etch exposed Cu and barrier remnants without an inhibitor package | Low — semi-aqueous formulations commonly include Cu corrosion inhibitors | Low — inhibitor chemistry built in specifically for via-bottom Cu exposure |
| Residue removal effectiveness | High for bulk organic/fluorocarbon polymer | High for bulk polymer, slightly slower strip rate than O2 | High for thin ionic/fluoride residue, weak on thick polymer | High for particulate and polymer, gentler overall | High when matched to the specific etch chemistry's byproduct |
| Typical production use | First-pass strip when no Cu is exposed at the via bottom | Default ash once low-k sidewall exposure or Cu-bottom exposure is a concern | Legacy/simple flows, now mostly combined with an inhibitor additive | Common pairing after H2-based ash on exposed-low-k, exposed-Cu vias | Increasingly standard combined ash+wet sequence at advanced low-k/Cu nodes |
The low-k damage mechanism at the center of this step is carbon depletion: carbon-doped oxide films such as Black Diamond-class OSG rely on Si-CH3 methyl termination to keep their dielectric constant low, and an oxidizing ash — straight O2 plasma in particular — strips that methyl termination from the exposed sidewall, leaving a porous, now-hydrophilic surface that absorbs atmospheric moisture and pushes the local k-value up from a pristine ~2.5 toward 3.0 or higher in the damaged skin layer, which is exactly the capacitance and leakage penalty the low-k film was deposited to avoid. H2/N2 forming-gas and other reducing ash chemistries were developed specifically as "damage-free" alternatives because they strip the fluorocarbon polymer and resist without the oxidizing radicals that drive carbon depletion, typically holding the sidewall k-shift to a few tenths rather than the half-point-or-more jump seen after an aggressive O2 overash; the trade-off is a slower strip rate, so production flows often still use a short O2 or mixed-chemistry step for the thickest polymer followed by a damage-minimized reducing finish. The wet clean that follows carries its own version of the same risk, since dilute HF at too high a concentration or too long a dip time can etch into an already carbon-depleted sidewall and widen the via or trench CD, which is why production recipes hold HF concentration under roughly 1% and increasingly favor semi-aqueous or proprietary post-etch-residue removers formulated to pull fluoride and metal-ion residue while leaving both the low-k carbon skeleton and any exposed Cu at the via bottom untouched.
This ash-strip-clean step is the direct handoff from via-etch-stop-dual-damascene-integration: the via and trench geometry that etch defined now has to reach barrier/seed redeposition with its sidewalls dielectrically intact and its via-bottom Cu, where exposed, free of both residue and corrosion, or the next level inherits a contact-resistance or leakage problem baked in before a single barrier atom is deposited. The low-k damage risk this step manages is the same risk osg-process and airgap-process were built to avoid in the first place — a carbon-depleted, moisture-absorbing sidewall skin undoes exactly the k-value reduction those integration schemes spent a deposition step achieving, so a damage-mitigated ash-and-clean sequence is not optional polish but the only thing standing between a low-k dielectric stack on paper and a low-k dielectric stack that still measures low-k after etch. With the via/trench surface clean and the carbon skeleton preserved, the process hands off to barrier/seed PVD redeposition, where a thin Ta/TaN barrier and Cu seed layer go down on exactly the surface this step was responsible for protecting.