Ta-based liner CMP - 40nm

Because mechanically fragile low-k or advanced dielectrics set strict upper limits on the permissible mechanical down-pressure, the material removal strategy is fundamentally forced toward chemically dominated mechanisms .

Following the bulk Cu removal in the preceding Cu CMP step, the BEOL plana

rization process transitions to the barrier CMP phase to clear the Ta-based liner . The primary objective of this step is to completely remove the conductive Ta/TaN barrier layer from the field dielectric regions, thereby ensuring strict electrical isolation between adjacent MET3 interconnects . If residual barrier material is left on the field, it creates unintended conductive paths that severely degrade yield through inter-line short circuits . Furthermore, this step prepares the wafer surface for the subsequent Post CMP Cleaning and ILD 3-1 Deposition by minimizing surface topography and achieving global planarization . Unlike the tighter-pitch MET1 or MET2 layers, the MET3 layer in a 40nm BSI CMOS image sensor flow often incorporates wider global routing interconnects and optical shielding structures, making this specific step highly susceptible to geometry-dependent metal dishing . Fundamentally, the removal of the Ta-based liner operates via a controlled tribo-electrochemical coupled corrosion mechanism rather than pure mechanical abrasion . Because tantalum is chemically inert in standard acidic Cu slurries, barrier CMP slurries typically employ alkaline environments with specific complexing agents to slowly release oxidizers, such as H2O2 derived from percarbonate . This chemical environment induces the formation of weakly bound oxide and hydroxide sub-layers on the exposed Ta surface . In the polishing contact zone, mechanical friction from the pad and suspended abrasives (e.g. , colloidal silica) dynamically disrupts these passivated reaction layers, exposing fresh metal to allow the continuous electrochemical reactions to proceed . This synergistic process follows mixed potential theory, where parallel anodic metal dissolution and cathodic reduction reactions are coupled at the dynamic interface , with the mechanical removal component scaling proportionally to applied pressure and relative velocity as defined by Preston's law . Material and slurry selections for this step are driven by the critical need to invert the removal rate selectivity compared to the preceding Cu CMP step . Because copper is inherently softer and chemically more reactive than the surrounding dielectric, it is highly prone to over-polishing during the Ta clearance phase . To mitigate this, the barrier CMP slurry incorporates specific corrosion inhibitors, such as triazole compounds, which adsorb onto the recessed Cu surfaces to form a protective passivation film . Simultaneously, the slurry relies on specific pH regulators to modulate the surface zeta potential and carefully tune the interfacial electrochemical dissolution rates of both metals . By dynamically suppressing the potential differences between the chemically active Cu and the passive Ta, the slurry effectively minimizes the driving force for galvanic corrosion at the bimetallic interface . At the 40nm technology node, the local contact mechanics and slurry mass transport behaviors become highly non-uniform across dense patterned arrays, meaning blanket wafer selectivity metrics no longer reliably predict actual planarization performance . Because mechanically fragile low-k or advanced dielectrics set strict upper limits on the permissible mechanical down-pressure, the material removal strategy is fundamentally forced toward chemically dominated mechanisms . Consequently, controlling the effective "dishing radius" through optimized slurry chemistry and integrating layout-level line segmentation are mandatory co-optimization strategies to maintain cross-sectional metal volume and prevent catastrophic RC delay degradation in wide MET3 structures .

[High] Metal Dishing and Dielectric Erosion: Because copper is substantially softer than the silicon dioxide dielectric and the Ta barrier, its removal rate often exceeds that of the surrounding materials during the over-polish phase . This mismatch leads to a cylindrical thickness loss in the metal regions (dishing) and localized thinning of the oxide arrays (erosion), both of which directly reduce the interconnect cross-sectional area and severely degrade electrical RC performance .
[High] Galvanic Corrosion at the Cu/Ta Interface: The wet CMP slurry acts as a conductive electrolyte, creating a local micro-battery effect between the dissimilar Cu and Ta metals due to their mismatched electrochemical equilibrium potentials . If the slurry chemistry and pH regulators fail to sufficiently suppress this potential difference, rapid anodic dissolution of the less noble metal occurs, resulting in localized edge defects, trench voiding, and eventual electromigration reliability failures .
[Medium] Incomplete Barrier Clearance (Underpolish): As interconnect dimensions shrink, the effective slurry transport and local mechanical stress distribution change significantly, which can drastically reduce the specific Cu/liner removal-rate selectivity on dense patterned structures . If the local barrier removal rate drops excessively, residual Ta/TaN will remain on the field dielectric, causing critical inter-line leakage currents or hard electrical shorts .
[Low] Tribological Defectivity and Scratching: To effectively shear the chemically modified Ta-oxide surface layers, barrier CMP slurries must utilize relatively hard mechanical abrasives such as colloidal silica or alumina . If abrasive agglomeration occurs due to poor colloidal stability or suboptimal pH regulation , the resulting large particles can physically gouge the mechanically fragile dielectric, creating deep micro-scratches that trap conductive residues (Engineering Practice).

Following the bulk Cu removal in the preceding Cu CMP step, the BEOL plana

[High] Metal Dishing and Dielectric Erosion: Because copper is substantially softer than the silicon dioxide dielectric and the Ta barrier, its removal rate often exceeds that of the surrounding materials during the over-polish phase . This mismatch leads to a cylindrical thickness loss in the metal regions (dishing) and localized thinning of the oxide arrays (erosion), both of which directly reduce the interconnect cross-sectional area and severely degrade electrical RC performance .
[High] Galvanic Corrosion at the Cu/Ta Interface: The wet CMP slurry acts as a conductive electrolyte, creating a local micro-battery effect between the dissimilar Cu and Ta metals due to their mismatched electrochemical equilibrium potentials . If the slurry chemistry and pH regulators fail to sufficiently suppress this potential difference, rapid anodic dissolution of the less noble metal occurs, resulting in localized edge defects, trench voiding, and eventual electromigration reliability failures .
[Medium] Incomplete Barrier Clearance (Underpolish): As interconnect dimensions shrink, the effective slurry transport and local mechanical stress distribution change significantly, which can drastically reduce the specific Cu/liner removal-rate selectivity on dense patterned structures . If the local barrier removal rate drops excessively, residual Ta/TaN will remain on the field dielectric, causing critical inter-line leakage currents or hard electrical shorts .
[Low] Tribological Defectivity and Scratching: To effectively shear the chemically modified Ta-oxide surface layers, barrier CMP slurries must utilize relatively hard mechanical abrasives such as colloidal silica or alumina . If abrasive agglomeration occurs due to poor colloidal stability or suboptimal pH regulation , the resulting large particles can physically gouge the mechanically fragile dielectric, creating deep micro-scratches that trap conductive residues (Engineering Practice).

40nm BSI CMOS Image SensorPreview

Process Cross-Section

Step highlight

In depth

Risks & Challenges

40nm BSI CMOS Image SensorPreview

Process Cross-Section

Step highlight

In depth

Risks & Challenges