Post-CMP Cleaning Solution Electronic/EL Grade is supplied under the model designations ELC-200 for concentrated feed and ELC-200D for point-of-use prediluted delivery. The product is an aqueous acidic chelating formulation intended for removal of post-chemical mechanical planarization residues from copper damascene, through-silicon via, cobalt cap, tungsten plug, and porous low-k dielectric surfaces. The Electronic/EL grade designation is defined by batch release limits for mobile-ion contamination, transition-metal contamination, and sub-micron particles, rather than by chemical activity alone. Filling is performed in an ISO 14644-1:2015 Class 3 cleanroom through 0.05 µm PTFE filters into high-density polyethylene containers with fluoropolymer-lined closures.

The concentrated feed has a pH of 2.8 ± 0.2 at 25 °C, density 1.02 ± 0.02 g/cm³ at 20 °C, viscosity 1.1 ± 0.1 mPa·s, and surface tension 38 ± 2 mN/m. These values are measured on each batch with a calibrated glass electrode, oscillating U-tube densitometer, rotational viscometer, and Wilhelmy plate tensiometer. The concentrated form is intended for inline dilution at 10:1 to 15:1 with ultrapure water meeting ASTM D5127-18 Type E-1 requirements. The prediluted form ELC-200D is shipped for wet benches that lack online blending or that require low particle counts at startup.

The product contains no free fluoride, ammonia, or amine-based additives. This compositional boundary is relevant because ammonia-containing cleaners raise dissolved copper in the rinse water through copper-ammine complex formation, and amine additives can produce pH drift in recirculated baths over an 8 h shift. The absence of solvent co-solvents such as N-methyl-2-pyrrolidone or dimethyl sulfoxide eliminates the need for a separate isopropyl alcohol rinse after cleaning and reduces volatile organic compound loading in the fab exhaust.

Which Trace Metal Limits Define Electronic/EL Grade for Post-CMP Cleaning?

Electronic/EL grade is specified by maximum allowable concentrations of alkali metals, alkaline earth metals, and transition metals that can degrade transistor reliability or shift flatband voltage. For ELC-200, the batch release limit for each of Fe, Ni, Cu, Zn, Ca, Mg, Na, K, Al, and Mn is ≤1 µg/L; the total trace metal concentration is ≤5 µg/L by quadrupole ICP-MS after 10:1 preconcentration. Anion levels are controlled to chloride ≤0.5 mg/L, nitrate ≤2 mg/L, and sulfate ≤2 mg/L by ion chromatography. The analytical method blank uses ultrapure water conforming to ASTM D5127-18 and is run with each shift to detect cross-contamination.

Particle control is specified at ≥0.1 µm and ≥0.2 µm thresholds. The concentrated product must contain ≤100 counts/mL at 0.1 µm and ≤20 counts/mL at 0.2 µm; the prediluted ELC-200D must contain ≤50 counts/mL at 0.1 µm and ≤10 counts/mL at 0.2 µm by optical particle counter. The optical particle counter is calibrated per ISO 21501-1 using monodisperse polystyrene latex spheres.

Because electronic-grade samples are easily contaminated by room air and operator contact, sampling is performed in a cleanroom using pre-cleaned perfluoroalkoxy bottles. The sample volume for optical particle counting is 50 mL; samples are degassed for 30 s under vacuum to remove bubbles that would otherwise be counted as particles. Results are corrected for background counts from ultrapure water blanks meeting ASTM D5127-18.

Post-CMP residues are not uniform; they are compressed slurry layers containing ceria or colloidal silica abrasive, pad debris, benzotriazole corrosion inhibitor, and dissolved metal species from the platen. Benzotriazole forms a cuprous-BTA film that is hydrophobic and must be removed before subsequent dielectric cap deposition. The chelating carboxylic acid system in ELC-200 shifts the Cu-BTA dissolution equilibrium by complexing Cu2+ and destabilizing the film. At the use dilution of 10:1, the zeta potential of ceria particles becomes −15 ± 3 mV, which reduces redeposition onto negatively charged low-k and TEOS surfaces. If the diluted bath pH drifts above 4.2, the zeta potential rises toward zero, and particle redeposition increases. Bath pH monitoring with an alarm at 4.0 is therefore part of the process control sequence.

The active chelating species is a multicarboxylic acid blend, not a single acid. In comparison with citric acid alone, the blend retains chelation capacity after loading with 50 ppm Cu2+ and 20 ppm Fe3+ in the rinse bath, whereas citric acid alone shows a reduction in particle removal efficiency of 10–15% at the same metal loading. This difference is observed because the mixed-ligand system forms 1:1 and 2:1 complexes with Cu2+ and Fe3+, preventing precipitation of metal carboxylates on wafer surfaces.

Cleaning efficiency is measured by defect review on Cu pillar test wafers with intentionally contaminated slurry. At 22 °C, a 30 s dispense with brush assist removes 95% of 0.1 µm ceria particles and 97% of benzotriazole residue by Fourier transform infrared spectroscopy. Reducing the dispense time to 20 s lowers ceria removal to 82%, which is insufficient for wafers requiring post-planner inspection below 5 defects/cm² at 0.16 µm.

On a 300 mm single-wafer spin processor, the concentrated product is dispensed through a 0.1 µm point-of-use filter at 1.5 L/min for 30 s after the final platen rinse. Liquid temperature at the nozzle is held at 22 °C ± 1 °C. After dispense, the wafer is rinsed with ultrapure water at 2.0 L/min and dried with 0.05 µm filtered nitrogen at 1,200 rpm for 40 s. The low surface tension of 38 ± 2 mN/m suppresses watermark formation on hydrophobic ULK sidewalls compared with pure water at 72 mN/m.

During ramp-up on a 300 mm double-sided scrubber, a filter replacement without a line purge allowed trapped gas in the chemical delivery line to delay coverage by 0.5–1.0 s. The resulting defect pattern was a bevel residue band at 2–4 mm edge exclusion. The corrective action was to install an automatic 20 s chemical line purge at 400 mL/min before each lot start. Subsequent lots showed no repeat of the bevel signature.

With ceria-containing slurries at 0.5 wt% solids, the chelator capacity of the 10:1 dilution is insufficient, and the dilution ratio must be increased to 15:1. Failure to adjust the ratio increases post-clean defect density at trench corners by 15–20% relative to baseline on patterned defect wafers. For colloidal silica slurries at 1.0 wt% solids, the 10:1 ratio is acceptable.

Cleaning Sequence Integration in Single-Wafer Spin Processors

In single-wafer spin processors, the sequence must be arranged so that the acidic cleaner is not followed immediately by a high-pH rinse from a previous process step. Carryover of alkaline rinse water at more than 0.5 vol% buffers the diluted cleaner above pH 4.2 and negates the particle zeta potential control. The recommended sequence is chemical dispense, ultrapure water rinse, and nitrogen dry, with no intermediate acidic or alkaline treatments. Immersion wet benches can also be used if the bath is recirculated through a 0.1 µm filter at 10 L/min and the bath life is limited to 8 h to prevent chloride upconcentration and particle agglomeration.

On cobalt cap stacks with a 20 Å Co layer over copper, static immersion in ELC-200D at 1:10 dilution and 25 °C for 60 s produces a cobalt etch rate of ≤0.5 Å/min and no undercut at the Co/Cu interface by cross-sectional transmission electron microscopy. When the bath temperature exceeds 30 °C, the cobalt etch rate rises to 1.8 Å/min, and the process limit is therefore set at 28 °C maximum. This boundary is enforced by the wet bench temperature interlock and is independent of the platen temperature.

For tungsten plug arrays, the same 10:1 dilution yields a tungsten etch rate of <0.5 Å/min at 25 °C, measured by magnetic sector ICP-MS on immersion coupons over 10 min. This is below the tungsten thickness loss allowance of 2 Å for a 1 min clean on 7 nm node test structures.

When Low-k Damage Risk Requires a Dilute Acidic Formulation

Alkaline post-CMP cleaners based on tetramethylammonium hydroxide or ethanolamine operate at pH 9.5–10.5 and can attack porous organosilicate low-k films. For a low-k film with k value 2.55 and porosity 25–30%, immersion in an alkaline cleaner at 55 °C for 2 min increases k by 0.2–0.4 and raises RMS surface roughness from 0.3 nm to 1.2 nm by atomic force microscopy. ELC-200D at pH 3.4 and 25 °C shows no measurable change in k after 5 min immersion when measured by mercury probe capacitance-voltage at 1 MHz. The acidic buffer also avoids the silicate hydrolysis reaction that alkaline solutions promote in Si-O-Si networks.

The product does not contain free fluoride. Dilute HF-based post-CMP cleaners are used to remove silica residue but etch TEOS and SiCN caps. At 25 °C, the TEOS etch rate of ELC-200D is <2 Å/min, while typical dilute HF-based alternatives at 0.2 wt% HF etch TEOS at 8–15 Å/min. On SiCN etch stop layers, ELC-200D produces <0.5 Å/min at 25 °C, whereas dilute HF solutions may exceed 2 Å/min. This difference is significant when the SiCN cap thickness is below 50 Å.

Dissolved oxygen is controlled in the bulk storage tank by nitrogen sparging to <1 mg/L. When dissolved oxygen rises to 8 mg/L because of air ingress, the Cu static etch rate at 25 °C increases from <1 Å/min to 1.5 Å/min and the Cu surface after clean shows a slightly hydrophilic oxide layer. The nitrogen blanket is therefore a release parameter for open fab containers.

The following comparative data were generated on blanket films using the same immersion protocol at 25 °C for 60 s unless otherwise stated. The values are supplier qualification data and are not process guarantees.

At the end of the sequence, the rinse water and nitrogen must be free of metal contamination. The rinse water is supplied at 18.2 MΩ·cm resistivity and <5 ppb total organic carbon, meeting ASTM D5127-18. The nitrogen dryer must use a 0.01 µm point-of-use filter to prevent recontamination after cleaning. Compared with solvent-based post-CMP removers, the aqueous acidic product reduces total organic carbon in the fab waste stream by 60% relative to a solvent-based control because no organic solvent rinse is required. The waste stream is neutralized to pH 6–8 before discharge, and the absence of fluoride avoids formation of calcium fluoride precipitate in the fab neutralization tank.

Storage and handling of ELC-200 require high-density polyethylene, polypropylene, or fluoropolymer wetted surfaces. Stainless steel 316L is not suitable for long-term storage because the acidic chelating solution extracts Fe and Ni, increasing trace metal load above the 5 µg/L total limit after 72 h at 20 °C. The product is incompatible with strong alkaline cleaners, ammonia-containing rinses, and cationic surfactants; mixing can precipitate metal hydroxides and obstruct waste drain traps. In unopened containers at 5–25 °C, shelf life is 12 months from fill date. Once opened, the container must be maintained under filtered nitrogen and consumed within 7 days to prevent CO₂ absorption and particle ingress.

On a high-volume line processing 3,000 wafers/day, batch-to-batch variation in total trace metal concentration has remained between 0.8 µg/L and 1.2 µg/L over a 12-month qualification period. The only out-of-specification event was a particle count excursion to 180 counts/mL at 0.1 µm traced to an elastomer O-ring on the filling line, not to formulation chemistry. Replacement of the O-ring with a perfluoroelastomer part returned particle counts to the specified ≤100 counts/mL limit.