PVD Barrier Seed Redeposition

Once the via and trench emerge clean from ash-strip-clean, the next thing that touches that surface is metal, and it has to be the right metal in the right order: a thin tantalum or Ta/TaN barrier first, then a copper seed layer, both laid down by physical vapor deposition before the via ever sees an electroplating bath. The barrier exists because copper diffuses into silicon dioxide and low-k dielectric at ordinary interconnect operating temperatures, poisoning transistor threshold voltages and degrading dielectric breakdown if nothing stops it, so a continuous, pinhole-free Ta-based liner has to coat every sidewall and via bottom before copper ever arrives; the seed layer exists because copper electroplating is an electrochemical process that needs a continuous conductive surface to plate onto, and the Ta/TaN barrier itself is too resistive and too thin to carry the plating current evenly, so a separate 10s-of-nanometers-thick PVD copper film is deposited on top of the barrier purely to give the plating bath something uniform to nucleate on. Both films are deposited in the same class of tool, typically a multi-chamber PVD platform such as Applied Materials' Endura, where the wafer moves between a degas chamber, an argon-sputter preclean, a Ta/TaN barrier chamber, and a Cu seed chamber without breaking vacuum, because even a brief air break between barrier and seed deposition risks oxidizing the Ta surface and degrading seed adhesion.

Clean via/trench
Handed off from ash-strip-clean: dielectrically intact sidewalls, residue-free Cu at via bottom
PVD barrier deposition
Ar sputter preclean
In-situ argon ion etch removes native oxide from exposed Cu before barrier lands
Ta / Ta-TaN bilayer sputter
2-10 nm, line-of-sight PVD from a Ta target, often TaN first for adhesion/nucleation then pure Ta for low resistivity
PVD copper seed
Cu sputter, same vacuum
Tens of nm, continuous conductive film for electroplating nucleation, no air break from barrier step
Step-coverage limit
Line-of-sight PVD thins at via sidewalls and bottom as aspect ratio climbs, the scaling wall for this step
Ready for electroplating
Continuous Cu seed over continuous barrier handed to copper electroplating for bottom-up fill
Pure TaPure TaNTa/TaN bilayerALD TaN (post-PVD-era)
Bulk resistivity~130-200 µΩ·cm (rises sharply below ~5 nm film thickness from grain-boundary and surface scattering)~200-250 µΩ·cm, consistently more resistive than pure TaEffective stack resistivity between the two, tuned by relative thickness of each layerComparable to PVD TaN per unit thickness, but can be deposited thinner for the same barrier performance
Diffusion-blocking qualityGood once fully densified, but poor adhesion directly to Cu seed without an interface layerExcellent — the nitrogen disrupts the Ta grain structure into a more amorphous, harder-to-diffuse-through filmBest of both: TaN against the dielectric for adhesion and blocking, Ta against the seed for low resistivity and good Cu wettingExcellent, and more uniform thickness through the via than PVD
Step coverage at high aspect ratioDegrades fast — PVD is line-of-sight, sidewall and bottom coverage fall off steeply above roughly 3-4:1 aspect ratioSame line-of-sight limitation as pure TaSame limitation, now doubled since two sequential PVD steps each lose coverageConformal — ALD's surface-reaction mechanism coats sidewalls and bottom close to equally regardless of aspect ratio
Adoption era / statusEarly Cu interconnect generations, simple single-layer approachCommon where blocking quality mattered more than resistivity budgetBecame the mainstream PVD approach through most of the 2000s-2010s nodesPhased in as PVD step coverage ran out of headroom below roughly the 28-22 nm interconnect generations, eventually giving way to Co/Ru direct-plate schemes at the most advanced nodes
Barrier/Seed PVD: Conformal Coating at Low Aspect Ratio Ta/TaN barrier plus Cu seed lining a via sidewall and bottom before electroplating Via cross-section, low aspect ratio (~2:1) low-k dielectric Ta/TaN barrier Cu seed via bottom, prior Cu line At this geometry, line-of-sight PVD sputter flux reaches sidewall and bottom with relatively even thickness Coverage uniformity, sidewall vs bottom vs top top field sidewall via bottom Workable coverage at roughly 3-4:1 aspect ratio and below is why PVD barrier/seed was mainstream for over a decade

Step coverage is the physical limit that eventually ends PVD's run as the barrier/seed workhorse: PVD is a line-of-sight process, atoms sputtered off the target travel in straight lines to the wafer, so as a via's aspect ratio climbs — depth divided by width — the top opening starts to self-shadow the lower sidewall and via bottom from a growing fraction of the incoming flux, and the deposited film thins unevenly, with the bottom corner and lower sidewall losing thickness fastest. Below roughly a 3-4:1 aspect ratio this thinning is manageable with process tricks like collimated or long-throw sputtering and wafer biasing to re-angle the flux, but as interconnect vias scaled narrower and taller through successive nodes, aspect ratios pushed past that window and PVD coverage at the via bottom and lower sidewall became too thin and too non-uniform to guarantee a continuous barrier — a single pinhole in the barrier at the via bottom is a direct copper-diffusion path into the dielectric, and a seed layer that goes to zero thickness anywhere on the sidewall leaves a gap the plating bath cannot nucleate across, producing voided or incompletely filled vias. This is precisely the scaling wall that motivated the industry's shift toward atomic layer deposition for the barrier: ALD builds the film one self-limiting surface reaction at a time rather than by line-of-sight flux, so it coats sidewalls and via bottoms with near-equal thickness regardless of aspect ratio, trading PVD's higher throughput and lower cost for conformality the geometry now demands; at the most advanced nodes the industry has gone further still, replacing the PVD-seed-plus-electroplating sequence entirely with direct-plate schemes on cobalt or ruthenium liners that need no separate copper seed at all.

The Scaling Wall: PVD Step Coverage at High Aspect Ratio Line-of-sight flux self-shadows as vias get narrower and taller Low aspect ratio (~3:1) Continuous barrier, bottom intact High aspect ratio (~8:1) Thinned/discontinuous at bottom corner incoming sputter flux, mostly unobstructed top opening self-shadows lower sidewall and bottom A pinhole in the barrier is a direct Cu-diffusion path; a seed gap on the sidewall is a plating void waiting to happen ALD barrier deposition coats both geometries with near-equal thickness because it builds film by self-limiting surface reaction, not line-of-sight flux This is why PVD barrier/seed gave way to ALD barriers and, at the most advanced nodes, Co/Ru direct-plate liners

Barrier/seed PVD is the direct handoff from post-etch-via-trench-ash-strip-clean and the direct setup for copper electroplating: the ash-and-clean step delivered a dielectrically intact, residue-free via and trench, and everything this step does — the argon preclean, the Ta/TaN barrier sputter, the Cu seed sputter — is aimed at putting a continuous, pinhole-free conductive surface on top of that clean geometry without ever breaking vacuum between steps. Get the barrier continuous and the seed uniform, and the via hands off cleanly to copper electroplating for bottom-up fill; leave a pinhole in the barrier or a thin spot in the seed, and the defect that PVD's line-of-sight physics created at the via bottom or lower sidewall becomes the void, the diffusion spike, or the resistance outlier that shows up downstream in electrical test, long after the wafer has left the PVD chamber. That aspect-ratio-driven step-coverage limit is also why this single process step has an expiration date built into the technology roadmap: as vias kept scaling narrower and taller, PVD's line-of-sight physics ran out of headroom before copper interconnect scaling did, which is exactly why ALD barrier deposition and, further out, cobalt and ruthenium direct-plate liner schemes exist — they are the answer to a geometry problem this step cannot out-engineer forever.

Take PVD barrier seed redeposition further

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