Post-CMP cleaning is the sequence of wet-chemical and mechanical steps performed immediately after chemical mechanical planarization to remove slurry residue, metallic contamination, organic films, and loosened particles from the wafer surface before it advances to the next deposition or patterning level. CMP slurries can leave abrasive particles adhered to the surface, dissolved metal species that may redeposit, organic additives such as benzotriazole that form persistent films on copper, and corrosion products generated during the polish itself. If these residues remain, they can nucleate defects in subsequently deposited films, raise via resistance, degrade dielectric reliability, and reduce yield. The cleaning module therefore operates under constraints set by the preceding CMP step and the requirements of the next film: the chemistry must remove each contaminant class without unacceptable attack on exposed metals, barriers, caps, or dielectrics, while mechanical and acoustic action must dislodge particles without exceeding the surface-damage budget.

The post-CMP cleaning module must address multiple contaminant classes simultaneously — abrasive particles, metallic ions, organic films, and corrosion products — and a cleaning recipe that removes one class effectively can worsen another if the chemistry and sequence are not integrated. Slurry abrasive particles (silica, alumina, or ceria, depending on the CMP step) adhere through combinations of van der Waals, electrostatic, chemical, and capillary interactions. Metallic species from the polished film, slurry, pad, or conditioner can adsorb or redeposit, while mobile contaminants that remain above the integration-specific limit can degrade dielectric reliability. Organic residues — corrosion inhibitors, surfactants, and polymer fragments — can form films that interfere with subsequent nucleation or adhesion. The cleaning sequence must therefore be tested as a complete train so that each step removes its target without reintroducing contamination removed earlier.

Brush scrubbing with porous PVA rollers is a widely used mechanical cleaning step, and its effectiveness depends on brush contact, rotation, fluid delivery, chemistry, and the condition of the brush itself. Depending on tool architecture, brushes clean the device side, backside, or both while chemistry at the brush-wafer interface changes particle adhesion and helps transport released material away. Contact must be sufficient to remove contamination without creating scratches or damaging fragile features. The porous brush can also load with particles and metal-containing residues and later become a cross-contamination source, so qualification must cover brush break-in, steady-state operation, cleaning, and end-of-life criteria tied to measured defect performance.

Megasonic cleaning couples high-frequency acoustic energy into the liquid, where acoustic streaming and cavitation-related effects can contribute to particle detachment and transport without direct brush contact. Frequency, delivered power, dissolved-gas content, transducer geometry, liquid depth, chemistry, and feature geometry jointly determine removal and damage; frequency alone does not define a safe window. Excess acoustic stress or bubble activity can damage fragile lines, caps, or low-k structures, while insufficient energy leaves adhered particles behind. Solution pH, ionic strength, and surfactants also change wetting and interfacial charge, so acoustic parameters and chemistry must be qualified together on representative structures rather than inferred from blanket particle-removal data.

The chemistry used in post-CMP cleaning can adjust interfacial charge, complex dissolved metals, remove organic residues, and control corrosion, but its effect depends on concentration, pH, dissolved oxygen, exposure time, and the complete exposed-material stack. Organic acids and chelators can reduce free copper-ion activity and help remove copper-containing residues. Alkaline chemistries can make silica and some oxide surfaces similarly charged, but the sign and magnitude of zeta potential must be measured for the actual particle, surface, and solution. Hydrogen peroxide can alter copper oxidation and passivation, yet whether it protects, roughens, or accelerates dissolution depends on the rest of the formulation and cannot be assigned from peroxide concentration alone. A first-order DLVO description writes the interaction energy as the sum of van der Waals and electrostatic double-layer terms,

but real post-CMP surfaces can also involve roughness, chemical bonding, steric forces, and non-DLVO interactions. Adjusting pH, ionic strength, and surfactant adsorption may reduce attraction or create an energy barrier to redeposition, but it does not guarantee net repulsion. In an idealized sphere-plane van der Waals contact at fixed separation and Hamaker constant, adhesion scales with particle radius as

where $R_p$ is particle radius. Removal forces scale differently and are sensitive to contact geometry and flow, which helps explain why particle-removal performance must be reported by size rather than as a single aggregate number.

Metallic contamination control after CMP requires both removal of surface metal species and prevention of adsorption or redeposition from the cleaning liquid and brush. Copper-ion speciation, complex formation, pH, dissolved oxygen, and the electrochemical potentials of exposed conductive materials govern whether copper remains soluble, adsorbs, or deposits. Chelating agents can reduce free-metal activity, but effectiveness depends on formulation and the chelator-to-metal loading throughout bath or point-of-use life. Surface-metal measurements such as VPD-ICP-MS or TXRF must meet process-specific limits tied to the reliability of subsequent dielectric or metallization levels; historical limits should not be carried forward without requalification.

Post-CMP cleaning of surfaces containing low-k dielectric materials adds the constraint that the cleaning chemistry must not increase the dielectric constant, roughen the surface, or extract carbon from the film, all of which degrade the electrical and mechanical properties the integration scheme depends on. Carbon-doped oxide and organosilicate glass dielectrics achieve their low permittivity partly through methyl groups that reduce film density and polarizability; alkaline cleaning solutions or oxidizing chemistries can strip these groups, raising the dielectric constant and increasing capacitance in the finished interconnect. Damage to the low-k surface also increases moisture uptake, which further raises the dielectric constant and can compromise adhesion of the next deposited layer. The compatibility of the cleaning recipe with the specific low-k material must be demonstrated by measuring the dielectric constant and leakage current after cleaning, not only by verifying particle and metal specifications. This constraint often narrows the pH and oxidizer windows available for cleaning, forcing the process engineer to accept less aggressive particle or metal removal rather than risk dielectric damage, and the resulting trade-off must be resolved through the integration qualification rather than by optimizing the cleaning step in isolation.

Transfer wafer from CMP platen to cleaning module without allowing slurry residue to dry → Apply the qualified keep-wet or displacement rinse → Select noncontact acoustic cleaning, brush cleaning, or their qualified sequence from the surface and defect risks → Apply chemistry for particle release, metal complexation, organic removal, and corrosion control → Use staged DI-water displacement until residual chemistry meets specification → Apply the qualified spin, vapor-assisted, or other drying method → Inspect particles and scratches by size and location → Measure surface metals and organic residue against integration limits → Verify exposed-metal corrosion and low-k compatibility → Qualify the complete sequence across CMP consumable life, brush life, chemistry loading, and incoming variation

Drying after the final rinse is part of contamination control because residual droplets can concentrate dissolved material into watermarks and a moving contact line can redistribute particles. Spin drying removes liquid centrifugally, while IPA-assisted Marangoni drying uses a surface-tension gradient to withdraw the liquid film. Neither method is universally superior: watermarking, particle adders, chemical residue, topography, surface wettability, exhaust control, and safety requirements determine the qualified choice. The post-dry inspection—not the nominal dryer type—must demonstrate that cleaning gains survive the final module step.

Read post-CMP cleaning through a contamination-budget lens: every step in the sequence — megasonic exposure, brush scrub, chemical treatment, rinse, and dry — must reduce its target contaminant class below the specification without introducing new defects, corroding exposed metals, or damaging the dielectric, and the cleaning module as a whole must be qualified not only at its own output but against the yield and reliability of the levels built on top of it.