The Tungsten CMP Polishing Slurry Electronic/EL Grade, product code EL-W/300/5M, is supplied as a low-sodium, low-potassium colloidal silica dispersion formulated with a stabilized hydrogen peroxide oxidizer package and a benzotriazole-free corrosion inhibitor. It is used for selective tungsten bulk removal and over-polish on 200 mm and 300 mm chemical mechanical planarization platforms in logic, DRAM, 3D NAND, and through-silicon via modules. The as-supplied dispersion exhibits a pH of 2.3–2.6, a solids loading of 5.0–6.0 wt%, a mean secondary particle diameter of 95–110 nm, and a D99 tail below 0.45 µm when measured by ISO 13320:2020 laser diffraction. Sodium and potassium concentrations are below 10 ppb, iron below 25 ppb, and copper and zinc below 10 ppb in qualification lots. The low mobile-ion content is specified for gate-all-around and advanced DRAM flows in which post-planarization thermal cycling can drive Na⁺ and K⁺ into low-k dielectrics. The slurry is filled in an ISO Class 4 cleanroom per ISO 14644-1:2015 and packaged in high-density polyethylene containers with perfluoroelastomer seals. Dynamic viscosity at 25 °C is 1.5–2.5 mPa·s, and the material remains colloidally stable for 12 months in unopened, nitrogen-blanketed containers stored at 5–25 °C.

At point of use, the slurry is diluted 1:1 to 1:1.5 with 30% electronic-grade hydrogen peroxide to maintain a working oxidizer concentration of 2.0–2.5 wt%. Dilution water is degassed and supplied at 18.2 MΩ·cm resistivity per ASTM D5127-13. In this pH window, tungsten is oxidized to a hydrated WO₃ layer that is mechanically removed by the abrasive and chemically converted to soluble metatungstate species. The tungsten static etch rate is held below 10 Å/min to preserve feature integrity during wafer queue delays. Excess peroxide above 3.0 wt% does not increase removal rate; it thickens the oxide film and causes a passivation-related removal-rate drop of up to 20% in blanket tests. Published data for this exact formulation is limited to qualification-lot measurements and production-line split-lot evaluations.

What Distinguishes Electronic/EL Grade from Standard Tungsten Slurries?

The primary differences appear in the particle-size distribution tail, trace-metal content, and oxidizer stabilization. Standard alumina-based tungsten slurries can deliver blanket tungsten rates above 5000 Å/min because of the higher hardness of alpha-alumina, but they also generate more micro-scratches and higher post-clean defect counts on soft dielectric layers. Electronic/EL Grade uses spherical colloidal silica with a D99 tail capped at 0.45 µm; this lowers the incidence of shear-induced agglomerates in the point-of-use recirculation loop. In a 300 mm fab comparison using a rotary polisher with the same pad and conditioning protocol, the EL grade produced <5 defects/cm² at a 0.16 µm brightfield threshold, while a standard alumina-based slurry produced 12–18 defects/cm² under identical conditions. The trade-off is a lower tungsten removal rate; for processes requiring bulk tungsten rates above 5000 Å/min, a high-solids alumina formulation may be more appropriate, provided that post-CMP clean can remove alumina residues from recessed features.

The electronic-grade trace-metal specification also reduces hydrogen peroxide decomposition. Iron and copper at concentrations above 100 ppb act as Fenton-type catalysts and accelerate peroxide consumption in standard slurries, producing oxygen gas and shifting the pH upward during recirculation. By limiting iron to <25 ppb, the EL grade holds peroxide loss to <2% per month at 20 °C. This is critical for fabs that run extended point-of-use day tanks and cannot accept removal-rate drift between hourly qualification wafers.

Point-of-Use Integration on 300 mm Rotary CMP Platforms

The slurry is dispensed at 180–250 mL/min through a 0.5 µm depth filter onto a polyurethane pad with Shore D hardness 52–58. Bulk tungsten removal is typically run at downforce 2.5–3.5 psi, platen speed 80–110 rpm, and head speed 80–110 rpm. Over-polish after endpoint is performed at reduced downforce of 1.5–2.5 psi to limit dishing and oxide erosion. On production-scale rotary polishers with an infrared platen-temperature probe, pad temperature rises from 20 °C to 28–30 °C during a continuous 25-wafer lot. If the pad conditioning interval exceeds 60 min with a 180–220 grit diamond disk, tungsten removal rate drops by 8–12% and within-wafer non-uniformity increases from 3% to 6%. Restoration of rate and uniformity requires ex situ conditioning at 1.5–2.0 psi for 30–60 s between lots. These values are representative of split-lot production data on 300 mm tools; the process window should be re-qualified for each pad supplier and wafer stack.

In bulk tungsten removal on via-first structures, the slurry is used after contact etch and barrier deposition. The first polish step removes 3000–4000 Å of tungsten overburden at 2800–3600 Å/min, followed by an endpoint step with optical and motor-current signals. The endpoint signal is triggered when the tungsten film clears from dielectric surfaces; an additional 20–30% over-polish clears residual plugs and liner topography. With the EL grade, tungsten plug recess after 30% over-polish has been measured in scanning electron microscopy cross sections at 15–25 nm for 0.25 µm contacts, while TEOS oxide erosion remains below 30 nm. The low oxide removal rate is the main factor controlling plug recess, because excessive slurry dwell on exposed oxide after tungsten endpoint directly converts to dielectric loss. If the pattern density exceeds 60%, the over-polish time should be reduced by 10–15% or downforce should be lowered to 1.5 psi to avoid array erosion.

When Over-Polish Exceeds 30% on Dense Tungsten Contact Arrays

On dense contact arrays with pitch below 0.36 µm, excessive over-polish produces oxide erosion that can shift subsequent lithographic depth-of-focus windows. In split-lot production data, increasing over-polish from 20% to 40% raised TEOS oxide erosion from 22 nm to 47 nm and widened plug recess distribution from 15–25 nm to 25–45 nm. The slurry tungsten-to-TEOS selectivity of 50:1–80:1 is necessary but not sufficient for dense arrays; the removal-rate map must be matched to pad pressure distribution. Center-to-edge removal-rate non-uniformity greater than 5% at 3.0 psi creates localized over-polish at wafer edge, where contact resistance after tungsten gap-fill can increase by 8–12%. Therefore, the slurry is qualified with a multi-zone head pressure profile that sets the outer-zone downforce 0.2–0.4 psi lower than the inner zone. This production-scale compensation is more effective than changing slurry chemistry when the defect source is mechanical rather than chemical.

For wide tungsten lines above 10 µm, the dominant post-polish feature error shifts from erosion to dishing. The EL grade has produced dishing below 30 nm on 10 µm tungsten pads after 30% over-polish in test masks. Because dishing is governed by the interaction between pad contact and tungsten dissolution under load, the reduced static etch rate and lower downforce over-polish combine to limit tungsten loss after endpoint. If dishing exceeds 35 nm, the corrective action is to reduce over-polish downforce to 1.5 psi and verify endpoint detection delay rather than to increase oxidizer concentration, which tends to accelerate chemical erosion without improving planarization efficiency.

Corrosion and Post-CMP Residue Control in Acidic Tungsten Planarization

The acidic pH of 2.2–2.6 creates a galvanic couple between tungsten plugs and TiN or Ti liner materials. Open-circuit potential measurements in the undiluted slurry show a tungsten-to-titanium nitride potential difference of 120–180 mV. The corrosion inhibitor package holds tungsten static etch rate below 10 Å/min and titanium nitride static etch rate below 20 Å/min during a 60 s wafer queue delay. Avoid contamination with amine-based post-CMP cleaner carryover into the slurry loop; pH values above 3.5 cause silica particle aggregation and a rapid loss of tungsten removal rate. The slurry is incompatible with strong oxidizers such as concentrated hydrogen peroxide above 10 wt%, which destabilizes the colloidal suspension and releases oxygen gas in closed point-of-use lines. Filtration at 0.2 µm absolute is not recommended because the high shear can induce gel formation; a 0.5 µm depth filter with low differential pressure below 0.15 MPa preserves particle-size distribution and removal-rate stability.

Point-of-use recirculation loops with long residence times can exhibit pH drift caused by CO₂ absorption, peroxide decomposition, and water evaporation. In a 200 L day tank recirculated at 10 L/min for 24 h, the slurry pH increased from 2.4 to 3.6 when the tank was not nitrogen-blanketed. At pH 3.6, the zeta potential of the silica particles moved closer to the isoelectric point, increasing the effective particle size from 95 nm to 210 nm and producing visible particle settling within 4 h. The measured tungsten removal rate dropped by 18% because the aggregated particles reduced contact area and altered pad conditioning efficiency. Operators should maintain a nitrogen blanket at 0.05–0.1 MPa headspace pressure and monitor pH after every 8 h shift. If pH exceeds 3.0, the point-of-use tank should be replenished with fresh slurry rather than adjusted with mineral acid, because local acid addition creates low-pH zones that accelerate silica dissolution and increase dissolved silicon in solution.