In copper bulk CMP for 300 mm logic and 3D NAND interconnect layers, colloidal silica or fumed alumina slurries are maintained in a peroxide–glycine system at pH 4.0–6.0. An electronic/EL-grade polymeric dispersant is introduced at 0.2–2.0 wt% of liquid slurry mass to prevent aggregate formation when dissolved Cu²⁺, glycine, and benzotriazole raise the ionic strength above 0.5 M. The dispersant adsorbs on abrasive surfaces through carboxylate or sulfonate anchoring groups; the resulting electrosteric barrier must hold the D50 shift measured by dynamic light scattering per ISO 22412:2017 to less than 25 nm relative to the incoming abrasive lot. Zeta potential values are typically held between −25 mV and −40 mV using electrophoretic light scattering per ISO 13099-2:2012. If the absolute zeta potential falls below 20 mV, a rise in laser surface inspection count above 0.2 µm on a KLA Surfscan SP7 is the conventional lot rejection signal. In high-volume manufacturing, the slurry is filtered through a 0.1 µm polypropylene depth filter immediately before point-of-use distribution; dispersant-induced flocculation can blind this filter within 4–8 h.

The function is not limited to particle suspension. In copper bulk slurries, excess dispersant above 2.0 wt% competes with benzotriazole for copper surface sites, altering the passivation layer thickness and increasing dishing. A reduction below 0.2 wt% leaves the slurry understabilized at the high ionic strength produced by 1–3 wt% hydrogen peroxide decomposition. Formulation qualification therefore uses a combined metric: 24 h stability at 20 °C with no more than 5 % increase in D90 and no visible sediment. For EL-grade material, the dispersant entering the slurry formulation must be low in sodium and chloride. Cation impurities are commonly screened by ICP-MS after acid digestion using an ISO/IEC 17025-accredited method; total trace metals below 100 ppb are typical for commercially available EL-grade dispersant lots, with Fe, Cu, Ni, Cr, Ca, Mg, and K below 10 ppb each. The final device structures include copper interconnects with barrier layers, low-k dielectric spacing, and etch-stop layers; defects generated by dispersant failure translate directly into open or short yield loss at ≤ 7 nm design rules.

What role does an EL-grade dispersant play in Ta/TaN barrier slurry selectivity?

Barrier CMP for Ta/TaN liners in copper/low-k integration operates at pH 8.5–10.0 with colloidal silica at 2–6 wt% solids. The dispersant in this regime is selected for anionic stability in the presence of guanidine or imidazole corrosion inhibitors; it must not reduce the Ta/TaN-to-Cu removal rate ratio below typical production targets of 30:1. At high pH, the dispersant carboxylate groups are strongly dissociated, but competitive adsorption from dissolved silicate species can displace the dispersant from the silica surface. The resulting zeta potential drift is monitored by ISO 13099-2:2012; a shift of more than 15 mV after 24 h aging at 40 °C is treated as an incoming lot failure because it indicates weak anchoring. In barrier slurries, the dispersant must also remain compatible with the low-k dielectric film. Above 1.5 wt%, an excess anionic dispersant can increase the slurry viscosity by more than 20 %, which reduces planarization efficiency on the barrier layer and increases copper dishing in the adjacent line.

Formulation work on TaN elimination slurries frequently uses a two-step process: bulk barrier removal at 0.5–1.0 wt% dispersant, followed by a buff-stop step with a lower abrasive content and a higher dispersant-to-abrasive ratio to prevent particle reattachment to the copper surface. The terminal product is the completed copper interconnect after barrier layer removal and before dielectric cap deposition. Equipment-side evidence from 300 mm polishers shows that an under-dispersed TaN slurry produces edge ring defects that are detectable on post-CMP macro inspection; correction to the dispersant content returns edge exclusion wafer loss to below 3 mm. This application requires an EL-grade dispersant with low amine residue because amine-carrying dispersant grades can accelerate copper corrosion in the alkaline barrier step; in such cases, static etch rate on copper at pH 9.0 exceeds 5 nm/min and must be suppressed with additional BTA, complicating the formulation. For EU production sites, the blend is further assessed under Regulation (EC) No 1907/2006; any SVHC present above 0.1 wt% triggers communication obligations in the supply chain.

Tungsten plug planarization for DRAM and 3D NAND contact arrays operates at pH 2.0–3.0 with 5–10 wt% colloidal silica and iron-based oxidizer chemistry. Dispersant selection in this acidic environment centers on polyacrylic acid and copolymers with molecular mass 10,000–50,000 g/mol; below 10,000 g/mol the adsorbed layer is too thin to prevent bridging, while above 50,000 g/mol the high molecular weight increases slurry viscosity at the point-of-use filter. The slurry is maintained at a zeta potential of −15 mV to −30 mV despite the high ionic strength from Fe(NO₃)₃ and KIO₃; pH drift above 3.5 changes the iron redox balance and causes particle agglomeration. Dispersant content is normally held at 0.3–1.2 wt%, and the critical process window is narrower than in copper bulk CMP because excessive dispersant suppresses the tungsten removal rate by more than 20 % while insufficient dispersant produces plug seam attack and keyhole voids after tungsten fill.

Production-scale polishers with rotary or orbital kinematics have demonstrated that tungsten slurries with poor dispersant stability show a rise in D90 oversize from 50 nm to 300 nm within 6 h of slurry loop recirculation. This oversize fraction is retained on a 0.5 µm point-of-use filter and leads to filter housing pressure alarm. The terminal products are tungsten contacts and vias below 60 nm diameter; any post-CMP tungsten plug loss or oxide erosion creates contact resistance shifts. For EL-grade dispersants, Fe and Ni must be controlled below 5 ppb because the tungsten slurry is already iron-catalyzed; additional metal contamination shifts the oxidizer kinetics and alters the tungsten-to-oxide selectivity. The dispersant certificate of analysis should include an acid number reported as 100–300 mg KOH/g and an insoluble residue limit below 0.05 wt%.

When ceria-based STI slurries exceed 2 wt% solids loading, dispersant demand becomes nonlinear.

Shallow trench isolation planarization for logic and memory uses ceria abrasives with a primary particle size of 50–150 nm but a secondary aggregate size that can exceed 250 nm without adequate dispersant control. Because ceria has a higher density than silica and a point of zero charge near pH 6.5–8.0, the suspension is inherently unstable at the typical STI process pH of 3.5–6.5. Dispersant demand is not a linear function of solids loading. At 2 wt% ceria, 0.5–1.5 wt% of an anionic ammonium salt dispersant may maintain a D50 below 200 nm, but at 4–5 wt% ceria the required dispersant can rise to 3 wt% and simultaneously increase slurry viscosity from 2 cP to more than 8 cP. This viscosity threshold affects the slurry delivery system: rotary platen polishing tools with a ≤ 600 mL/min slurry flow rate exhibit non-uniform removal when the viscosity crosses 5 cP, because the slurry film thickness under the pad changes.

The dispersant in STI ceria slurries must also balance oxide removal rate against nitride removal rate. High dispersant adsorption on ceria suppresses particle contact and can lower oxide removal rate below 1,000 Å/min at 1.5–2.0 psi downforce, which is insufficient for through-time targets. Lower dispersant concentrations leave ceria aggregates that increase the nitride loss and reduce the oxide-to-nitride selectivity. A common lot release test therefore monitors the oxide-to-nitride removal rate ratio on 200 mm monitor wafers; a ratio above 30:1 is typical for advanced STI applications. The finished structures are isolation trenches filled with high-density plasma oxide; residual cerium particles embedded in the oxide sidewall after CMP are a known local contamination source. EL-grade dispersant must be filtered through 0.1 µm capsules before blending to remove insoluble crosslinked dispersant gel particles that would otherwise appear as carbon-rich defects on the wafer surface.

EL-grade dispersant lot acceptance in CMP slurry manufacturing commonly includes the following matrix. Each test is performed on a prediluted aqueous solution because neat dispersant viscosity exceeds the linear range of most QC rheometers.

Post-CMP residue control in advanced node cleaning formulations

After copper or tungsten CMP, particle residue is managed through aqueous cleaning formulations based on dilute ammonia-peroxide mixtures, organic acids, or hydroxylamine-based strippers at pH 7–10. An electronic-grade dispersant is introduced at 0.01–0.2 wt% into the cleaning bath to maintain abraded particles in suspension and prevent reattachment to the wafer. The mechanism is not the same as slurry stabilization; the cleaned wafer surface is hydrophobic after BTA passivation, so hydrophobic polymer segments in the dispersant compete with the wafer surface for particle adhesion. Cleaning bath conductivity is commonly held below 100 µS/cm for copper/low-k applications, which limits the dispersant concentration that can be used without causing copper corrosion. The dispersant must be compatible with the aqueous cleaning chemistry: in SC-1-like mixtures containing NH₄OH:H₂O₂:H₂O at 1:1:5 ratio, peroxide decomposition is accelerated by certain carboxylate-functional polymers above 0.05 wt%, reducing bath life to less than 4 h at 40 °C.

The post-CMP cleaning step is a narrow process window because the dispersant must not leave an organic residue on the wafer metrology pads. Residual dispersant films are detected as contact angle shifts from 10° to more than 30° on oxide or low-k surfaces; this is a release criterion in 10 nm and below technology. Formulations therefore use very short contact times of 30–120 s in single-wafer cleaning tools, with a rinse of ultrapure water at 18.2 MΩ·cm. For EL-grade dispersant, the absence of metal ions is critical in this application because the cleaner is not followed by a chemical step that can remove adsorbed cations. Packaging for cleaning additives requires fluoropolymer-lined containers; extracts from high-density polyethylene can introduce oligomers that increase wafer surface organic defects. The terminal output is a particle-neutral wafer entering the subsequent dielectric barrier deposition chamber.

Through-silicon via planarization for 2.5D and 3D packaging copper interconnects requires removal rates above 1 µm/min, which pushes slurry formulations to 10–20 wt% alumina or silica solids and pH 4.0–5.5. The EL-grade dispersant in this segment is selected primarily for shear stability because the slurry is recirculated through 1/2-inch PFA tubing at flow rates exceeding 400 mL/min and is repeatedly pumped by bellows or diaphragm pumps. Under these conditions, low molecular weight dispersants desorb under shear, while high molecular weight polymers undergo chain scission. A dispersant with a molecular mass of 20,000–100,000 g/mol and a shear-stable backbone such as a styrene–maleic acid copolymer is specified. The terminal via dimensions are 5–10 µm in diameter and 50–100 µm deep; large particles larger than 1 µm create via mouth blockages in the post-CMP cleaning step. Batch-to-batch variance in dispersant Na content above 500 ppb is tolerable for packaging fabs but must be controlled because it increases the zeta potential gradient across the slurry loop.

On production-scale 300 mm packaging lines, TSV copper slurries with inadequate dispersant content exhibit pad glazing and viscous slurry buildup on the retaining ring. The polishing tool is commonly a three-platen design with an in-situ optical endpoint; a slurry that thickens above 10 cP disturbs the endpoint signal by changing the pad-sweep temperature profile. Dispersant loading is set at 0.5–2.5 wt% and is tuned by measuring the slurry particle size distribution every 4 h with an in-line dynamic light scattering probe. The final product is a thinned wafer with filled copper TSVs ready for microbump metallization; dispersion defects in the slurry directly yield via protrusion or recess beyond the allowable ± 1 µm post-CMP dishing window. No further planarization correction is available after this step, so the EL-grade dispersant lot must pass a 7-day accelerated aging test at 40 °C before release.