Copper CMP Slurry: Complete Guide — Chemistry, Process Optimization & Advanced Node Applications

1. What Is Copper CMP Slurry?

Copper CMP slurry is a carefully engineered aqueous dispersion used in the chemical mechanical planarization (CMP) of copper interconnect layers during semiconductor wafer fabrication. It combines abrasive particles, chemical oxidizers, corrosion inhibitors, complexing agents, and surfactants to achieve simultaneous chemical modification and mechanical removal of metallic copper — producing the planar, defect-free surfaces required by advanced multilevel interconnect architectures.

Unlike purely mechanical polishing, CMP exploits the synergy between chemistry and mechanical action. The slurry’s chemical agents transform the hard metallic copper surface into a softer, more easily abraded compound — typically copper oxide or a copper-organic complex — which abrasive particles then remove through contact with a rotating polishing pad. This dual action is what makes sub-nanometer planarity achievable across an entire 300 mm wafer.

Copper CMP slurry is fundamentally distinct from oxide CMP slurries (used for silicon dioxide, shallow-trench isolation, or interlayer dielectrics) or tungsten CMP slurries. Each metal and dielectric system demands a unique chemistry tailored to its electrochemical behavior. For copper specifically, the formulation must simultaneously address four engineering challenges that do not exist in oxide CMP:

• Softness and ductility: Copper’s mechanical compliance makes it prone to dishing — a concave recess that develops when the pad deflects into copper-filled trenches under polishing pressure.
• Electrochemical reactivity: Copper oxidizes readily in aqueous environments. Without a corrosion inhibitor, uncontrolled chemical dissolution attacks not just the overburden but also the recessed copper in completed damascene features.
• Barrier metal integration: Copper lines are surrounded by a thin barrier metal (Ta, TaN, Co, or Ru) that must be selectively cleared in a second polishing step without exposing or eroding the underlying dielectric.
• Dielectric fragility: At advanced nodes, copper is embedded in ultra-low-k (ULK) dielectric materials with elastic moduli as low as 5–10 GPa — far more fragile than conventional SiO₂ and intolerant of high mechanical pressure.

Balancing these constraints within a single liquid formulation — while maintaining colloidal stability, long shelf life, and lot-to-lot reproducibility — is the central engineering challenge of copper CMP slurry development. The best formulations available in July 2026 achieve copper removal rates exceeding 300 nm/min, within-wafer non-uniformity (WIWNU) below 3%, and defect densities in the low single digits per square centimeter.

2. Damascene & Dual-Damascene Process Background

To understand why copper CMP slurry is formulated the way it is, one must first understand the damascene process — the deposition and planarization sequence that gives copper interconnects their structure. The word “damascene” refers to an ancient metalworking technique of inlaying metal into recessed patterns in another material, and the semiconductor process follows exactly the same principle.

Single vs. Dual Damascene

In the single damascene process, a trench or via pattern is etched into the interlayer dielectric (ILD). A thin barrier/adhesion layer (typically TaN/Ta) is then deposited by physical vapor deposition (PVD), followed by a copper seed layer and electrochemical deposition (ECD) of bulk copper that overfills the features. CMP then removes the copper overburden, leaving metal only in the etched recesses. Single damascene forms either a metal line layer or a via layer — but not both at once.

The dual damascene process is the dominant approach in production BEOL (back-end-of-line) processing from the 180 nm node onward. Both the trench (metal line) and the via (connecting plug to the layer below) are etched into the dielectric in a single patterning sequence before barrier and copper deposition. A single CMP step then clears both the via and line levels simultaneously, halving the number of barrier deposition and CMP cycles compared to sequential single damascene. This process efficiency is a major driver of cost-per-wafer reduction at high-volume logic and DRAM fabs.

The Planarization Challenge in Copper CMP

After electrochemical deposition, a 300 mm production wafer carries a copper overburden of 400 to 1,000 nm above the top of the trench features, with significant topographic variation across the wafer driven by pattern density differences in the underlying circuit layout. Dense arrays of narrow metal lines produce thinner copper deposits above the features (due to the “filling effect” in ECD), while isolated wide pads produce thicker deposits. This creates a post-ECD topography of 50–200 nm peak-to-valley variation before CMP even begins.

The planarization target is to remove this overburden completely and stop precisely at the barrier layer, without dishing the copper lines, without eroding the surrounding dielectric, and with endpoint uniformity across the entire 300 mm wafer diameter. The total thickness to remove may exceed 800 nm, while the tolerable endpoint window — the thickness range within which polishing must stop — may be as narrow as 20 to 30 nm. This 40:1 ratio between removal depth and endpoint tolerance is what makes copper CMP one of the most demanding precision manufacturing processes in the semiconductor industry.

Pattern Density Effects on Copper CMP

CMP is inherently sensitive to the local pattern density of the circuit being polished. Dense arrays of copper lines present more conducting material per unit pad-contact area, creating different removal kinetics than isolated wide copper regions. In practice, this leads to two characteristic topographic nonuniformities:

• Within-die (WID) variation: Removal rates vary across different areas of a single die due to pattern density differences, causing some regions to be over-polished (risking dishing and erosion) while others are under-polished (leaving residual metal).
• Within-wafer (WIW) variation: Mechanical pressure and slurry velocity distributions across the wafer radius produce systematic radial removal rate profiles that must be compensated through multi-zone carrier head pressure control.

Modern copper CMP slurry formulations incorporate passivation agents and carefully controlled oxidizer concentrations that self-regulate removal rates across different pattern densities — reducing dishing on isolated lines and erosion in dense arrays simultaneously. The interaction between slurry chemistry and pattern geometry is discussed further in the defect control section below.

3. Copper CMP Slurry Chemistry: Oxidizers, Abrasives & Inhibitors

A production-grade copper CMP slurry contains four functional chemical components, each playing a distinct and interdependent role in the overall polishing mechanism. The formulation scientist’s challenge is to optimize all four simultaneously across a process window that must remain stable over the slurry’s full shelf life and be reproducible from lot to lot.

3.1 Oxidizing Agents: The Chemical Engine

The oxidizer is the component that makes chemical mechanical planarization “chemical.” By converting the hard metallic copper surface into a softer copper oxide or hydroxide, the oxidizer dramatically lowers the effective hardness of the material being abraded — enabling mechanical removal at much lower pressures than would be needed to scratch bare copper directly.

Hydrogen peroxide (H₂O₂) is the oxidizer of choice in the overwhelming majority of production copper CMP slurries. At concentrations of 1–5 wt%, H₂O₂ is a highly effective copper surface oxidizer that produces only water and oxygen as decomposition products — meaning zero risk of metallic or halide contamination in the copper film or the surrounding ILD. Its concentration can be tuned with great precision, and its oxidation kinetics are well characterized across a wide pH range.

The principal limitation of H₂O₂ is its thermodynamic and catalytic instability. Copper ions released during polishing act as a Fenton-type catalyst, accelerating H₂O₂ decomposition and depleting the oxidizer from the slurry. For this reason, copper CMP slurries using H₂O₂ as the oxidizer are almost universally delivered via point-of-use (POU) mixing systems, where the slurry concentrate and the oxidizer are combined at the polisher dispense arm just before application to the pad. This eliminates shelf-life concerns for the oxidizer component while allowing the concentrated slurry base (without H₂O₂) to be stored for months.

Alternative oxidizers include potassium iodate (KIO₃), which offers much greater intrinsic stability than H₂O₂ and has been used in a number of commercial formulations. Its main disadvantage is the introduction of iodate anions that may leave surface residues requiring more aggressive post-CMP cleaning. Ferric nitrate (Fe(NO₃)₃) provides a strong oxidizing potential but carries unacceptable iron contamination risk for CMOS processes. As of 2026, H₂O₂ with POU delivery remains the industry standard.

3.2 Abrasive Particles: Hardness, Size, and Surface Chemistry

The abrasive phase of copper CMP slurry provides the mechanical component of material removal. The selection of abrasive type, particle size, size distribution, and surface chemistry is one of the most consequential formulation decisions in copper CMP slurry design — with direct implications for removal rate, defectivity, and selectivity between copper and the barrier metal or ILD.

Colloidal silica (SiO₂) abrasives (Mohs hardness ≈ 7) are the preferred abrasive for Step 2 (barrier clearing) and for advanced-node applications where the low-k dielectric and narrow copper lines demand minimal mechanical stress. Colloidal silica can be synthesized with very narrow particle size distributions (σ/mean < 0.15), excellent colloidal stability across a wide pH range, and surface chemistry that can be functionalized to control zeta potential and improve selectivity. Its relatively low hardness compared to copper oxide (Mohs 3.5–4) and barrier metals means it removes the softened copper oxide layer efficiently without inducing sub-surface damage in the underlying copper.

Fumed and colloidal alumina (Al₂O₃) particles (Mohs hardness ≈ 9) deliver significantly higher removal rates than silica due to their greater hardness and the high surface energy of their fractal agglomerate structures. They are well-suited for Step 1 bulk copper removal, where throughput is paramount. However, alumina carries a substantially elevated risk of surface scratching, and its harder particle-surface interaction is incompatible with the sub-1 psi polishing pressures required for ELK dielectric stacks at advanced nodes. In 2026, the trend in advanced-node process integration is away from alumina and toward engineered colloidal silica even for Step 1, accepting a modest throughput penalty in exchange for lower defect budgets.

Key abrasive particle specifications that must be controlled in production copper CMP slurry include: D50 (median particle diameter, typically 50–150 nm), D99 (99th percentile diameter, critical for scratch prevention, typically < 500 nm), zeta potential (governs colloidal stability and inter-particle interactions), and total particle count per mL (affects pad-contact frequency and removal rate).

3.3 Complexing Agents: Driving Dissolution and Preventing Re-deposition

Complexing agents (also called chelating agents or ligands) react with the copper ions liberated during oxidation and mechanical removal to form soluble copper complexes. This ion-capture function serves two critical purposes: it drives the thermodynamic equilibrium of the oxidation reaction forward (Le Chatelier’s principle), increasing the net chemical contribution to removal rate; and it prevents dissolved copper ions from re-depositing onto the freshly polished dielectric surface, which would leave metallic copper contamination that is electrically catastrophic in the transistor active regions below.

Glycine (aminoacetic acid, H₂NCH₂COOH) is the most widely used complexing agent in copper CMP slurries. It forms a stable bidentate Cu-glycine chelate complex at pH 4–9 — the operating range of most copper CMP formulations — with a formation constant (log K ≈ 8.2) large enough to drive rapid copper ion sequestration without being so large as to destabilize the BTA passivation film. Other complexing agents used in copper CMP research and some commercial formulations include citric acid, oxalic acid, tartaric acid, and various amino acid derivatives. Each has different pH-dependent stability, temperature sensitivity, and compatibility with the other slurry components.

3.4 Corrosion Inhibitors: Benzotriazole and Its Alternatives

The corrosion inhibitor is the component most unique to copper CMP slurry — it has no equivalent in oxide or tungsten polishing. Without a corrosion inhibitor, the oxidizer in the slurry would dissolve copper not just from the overburden but also from the recessed copper lines in completed features, causing severe dishing and line thinning. The corrosion inhibitor suppresses this “static etching” by forming a protective film on the copper surface that is stable when there is no mechanical pad contact but is locally disrupted where the pad’s asperities make contact, enabling selective removal only at elevated points.

Benzotriazole (BTA), a five-membered aromatic ring containing three nitrogen atoms (C₆H₅N₃), is the industry-standard corrosion inhibitor for copper CMP. BTA adsorbs onto the copper surface and reacts with Cu⁺ ions to form an insoluble, polymeric Cu(I)-BTA coordination complex approximately 2–5 nm thick. This film has very low solubility in the slurry, excellent thermal stability up to at least 60°C (well above pad interface temperatures), and sufficient mechanical compliance that it does not significantly change the friction characteristics of the pad-wafer contact.

BTA concentration is one of the most sensitive tuning parameters in copper CMP slurry formulation. Typical production concentrations range from 0.001 to 0.1 wt%. At concentrations below the minimum effective level, static etch rates rise above 2 nm/min and dishing becomes uncontrollable. At concentrations above the optimal range, the BTA film becomes mechanically resistant enough to suppress the intended removal — reducing overall copper removal rate and increasing polishing time. The optimal BTA concentration window narrows at advanced nodes where the combination of low polishing pressure and tighter dishing tolerances demands very precise inhibitor control.

As of July 2026, several regulatory environments are scrutinizing chemical classes structurally adjacent to PFAS (per- and polyfluoroalkyl substances), and some fabs operating under aggressive environmental compliance programs are seeking BTA-free copper CMP slurry alternatives. Candidate alternative inhibitors include 2-mercaptobenzimidazole (MBI), 1-phenyl-5-mercaptotetrazole (PMTA), and various triazole and imidazole derivatives. JEEZ offers BTA-free copper CMP slurry variants for customers with these requirements, validated to equivalent dishing and erosion performance on patterned wafers.

3.5 pH, Surfactants, and Slurry Stability

The pH of the slurry system governs the surface charge (zeta potential) of abrasive particles, the speciation of copper complexes, the solubility of the BTA-Cu film, and the oxidation kinetics of H₂O₂ on the copper surface. Most copper CMP slurries operate in the pH range of 4–9. Acidic formulations (pH 4–6) favor higher removal rates due to better copper ion solubility and faster glycine complexation, but may increase galvanic corrosion risk at the Cu/barrier interface. Near-neutral to mildly alkaline formulations (pH 6–9) favor lower defectivity, better barrier metal compatibility, and improved slurry particle stability.

Surfactants and dispersants — typically nonionic polyol or polyethylene glycol-based compounds — maintain colloidal stability of the abrasive phase by providing steric repulsion between particles. They also modify the wetting behavior of the slurry on the polishing pad, influence the transport of slurry into narrow damascene features, and can improve the uniformity of slurry film distribution across the pad surface.

4. Material Removal Mechanism: The Tri-Layer Passivation Model

The mechanistic framework most widely accepted in the copper CMP research community is the tri-layer passivation model, which describes material removal as a cyclic sequence of three coupled surface processes. Understanding this model is essential for interpreting process data and diagnosing removal rate and uniformity problems in production.

This tri-layer mechanism explains several important process behaviors that are otherwise difficult to rationalize:

• The selectivity of protrusion removal: Only protrusions experience sufficient contact stress to disrupt the BTA film; recessed features do not, making the process inherently self-planarizing at the scale of individual die topography.
• The non-linearity of removal vs. oxidizer concentration: Below a threshold oxidizer concentration, the oxide layer cannot form fast enough to sustain mechanical removal; above an upper threshold, the BTA film cannot suppress static etching even without mechanical contact. Both limits are observable as inflections in removal-rate vs. H₂O₂ concentration curves.
• The temperature dependence of static etch rate: Higher slurry temperatures accelerate both H₂O₂ decomposition and BTA film formation kinetics, requiring careful balance at elevated operating temperatures.

The static etch rate (SER) — the rate at which copper dissolves in the slurry in the complete absence of mechanical input — is one of the most important slurry qualification specifications. Production copper CMP slurries typically specify SER below 1–2 nm/min. The ratio of dynamic removal rate (with mechanical contact) to static etch rate, which can exceed 200:1 in optimized formulations, is a direct measure of the slurry’s mechanical selectivity and its ability to polish protrusions without dissolving recesses.

5. Key Process Parameters & Optimization

Copper CMP performance is determined by the interaction of slurry chemistry with machine settings — downforce, velocity, slurry flow, and pad conditioning. Even the most precisely formulated copper CMP slurry will underperform if these parameters are outside the slurry’s designed operating window.

5.1 The Preston Equation and Its Limitations in Copper CMP

The Preston equation provides the foundational model for CMP material removal rate:

where RR is the material removal rate (nm/min), K_p is the Preston coefficient (a material/slurry-dependent proportionality constant), P is the applied downforce pressure (kPa), and V is the relative velocity between the wafer and pad surface (m/s).

For copper CMP, the Preston equation gives useful first-order guidance but substantially oversimplifies the chemistry-mechanical interaction. The Preston coefficient K_p is not constant across process conditions — it varies with oxidizer concentration, BTA loading, pH, and pad conditioning state. At very high pressures (> 5 psi), the relationship becomes nonlinear as pad elastic contact mechanics change and the slurry hydrodynamic film between pad and wafer collapses. At very low pressures (< 0.5 psi, now required for ELK processes), chemical dissolution kinetics become the rate-limiting step rather than mechanical abrasion, and the Preston equation severely overestimates removal rate.

A modified model incorporating a chemical rate term — sometimes referred to as the Langmuir-Hinshelwood kinetic model for CMP — better captures the saturation behavior observed when oxidizer concentration exceeds the optimal window, and the inhibition behavior when BTA loading becomes excessive. This model treats the surface oxidation, passivation, and mechanical removal as parallel reaction pathways, each with its own rate constant and concentration dependence.

5.2 Downforce Pressure

Applied downforce is the most direct mechanical lever for controlling copper removal rate. Higher pressure increases the frequency and depth of pad-asperity contact events per unit time, increasing the rate of BTA film disruption and copper oxide removal. The practical constraint on downforce in modern copper CMP is the mechanical integrity of the ILD stack: ELK dielectric films with elastic moduli of 5–10 GPa cannot withstand polishing pressures above approximately 1.5 psi without risk of cracking or delamination at the ILD-etch-stop interface.

This constraint has driven a progressive reduction in copper CMP polishing pressure over successive technology generations: from 3–5 psi at the 130 nm node to 1.5–2 psi at 28 nm, to below 1 psi for the most pressure-sensitive ELK-containing stacks at 7 nm and below in 2026. Multi-zone carrier head systems on modern CMP tools allow independent pressure control across 4–7 concentric zones on the wafer, enabling compensation for radial removal rate profiles and WIWNU optimization without sacrificing overall process window.

5.3 Platen and Carrier Rotation Speed

Both the platen (polishing pad table) rotation speed and the carrier (wafer holder) rotation speed contribute to the relative velocity V in the Preston equation. Standard copper CMP operates both at 60–120 rpm. When the two rotate at different speeds, the velocity at any point on the wafer surface is the vector sum of the pad and carrier velocities, creating a kinematically complex velocity distribution across the wafer radius that is non-uniform and dependent on the specific rpm ratio chosen. Optimizing the rpm ratio — in conjunction with multi-zone downforce profiling — is a key lever for achieving < 3% WIWNU across the full 300 mm wafer.

5.4 Slurry Flow Rate and Point-of-Use Delivery

Slurry flow rate determines the supply rate of fresh chemistry at the pad-wafer interface and the removal rate of polishing byproducts (Cu²⁺ ions, BTA-Cu complexes, abraded slurry particles, and pad debris). Insufficient flow leads to slurry starvation: the oxidizer is locally depleted, byproducts accumulate, and removal rate and uniformity degrade. Typical flow rates for 300 mm copper CMP are 150–250 mL/min at the pad surface.

For H₂O₂-based slurries, point-of-use (POU) mixing systems are essential for preventing oxidizer decomposition before it reaches the pad. In a POU system, concentrated slurry base (without H₂O₂) is stored in bulk containers and is combined with a metered stream of H₂O₂ concentrate at the polisher dispense arm, typically within 30–60 seconds of pad application. This eliminates the shelf-life limitation on oxidizer stability while maintaining the convenience of a pre-formulated slurry concentrate for all other components.

5.5 Pad Conditioning

The polishing pad’s micro-texture — the distribution of asperity heights across its surface — is critical for both slurry transport and mechanical contact with the wafer. As polishing proceeds, pad asperities are plastically deformed and smoothed (a process called glazing), reducing the effective contact area and therefore both removal rate and its uniformity. In-situ conditioning with a rotating diamond conditioner disk continuously dresses the pad surface, re-opening grooves and resetting the asperity height distribution to a statistically steady state.

Conditioning aggressiveness — disc RPM, applied force, sweep speed, and sweep pattern — must be carefully optimized for each slurry chemistry. Different slurry formulations produce different pad wear characteristics, and a conditioning recipe optimized for one slurry may produce either over-conditioning (excessive pad consumption, pad debris generation, accelerated pad change cycles) or under-conditioning (pad glazing, removal rate drift) when used with a different slurry. This slurry-pad-conditioner interaction is one reason that copper CMP slurry qualification must be performed on the specific pad type and conditioner configuration used in production.

5.6 Temperature Control

The CMP reaction interface operates at elevated temperatures — typically 40–60°C at the pad-wafer contact zone — due to frictional heat generation and the exothermic chemical reactions. Platen chilled water temperature (normally set to 18–22°C) controls the baseline pad temperature, providing an indirect lever on slurry chemistry kinetics. Lower pad temperatures suppress static etch rates and oxidation kinetics, potentially improving WIWNU by reducing chemistry-driven non-uniformity; higher temperatures increase removal rate but also accelerate H₂O₂ decomposition and BTA film dissolution. Pad temperature monitoring (via embedded sensors or infrared pyrometry) is increasingly used in advanced process control (APC) systems for copper CMP at leading fabs.

6. Defect Control: Dishing, Erosion & Scratching

Copper CMP introduces a characteristic set of topographic and surface defects that directly impact interconnect electrical performance, downstream lithography, and device reliability. Understanding the root causes of each defect type — and the levers available to control them — is as important as achieving the target removal rate.

6.1 Copper Dishing

Dishing is the concave topographic depression that develops in a polished copper line or pad relative to the surrounding ILD surface after CMP. It arises because the polishing pad’s elastic compliance allows it to deflect into copper-filled trenches, especially for wide metal features, removing copper from the center of the trench even while the surrounding dielectric is already at or near its target thickness. Simultaneously, the chemical dissolution component of the slurry (static etching) removes copper uniformly across the trench profile — including in the recessed center that has no pad contact.

Dishing magnitude scales strongly with feature width: a 100 µm copper bond pad may dish by 50–100 nm after a standard copper CMP process, while a 100 nm logic interconnect dishes by only 2–5 nm under the same conditions. However, even 2–3 nm of dishing on a 10 nm-wide copper line at the most advanced nodes represents a 20–30% cross-sectional area reduction — a significant contribution to interconnect resistance increase and electromigration susceptibility.

Primary mitigation strategies for copper dishing include: optimizing BTA concentration to maximize static etch rate suppression, selecting harder pad stacks that minimize pad deflection into trenches, tightening polishing endpoint detection to minimize over-polish time, and using run-to-run advanced process control (APC) to maintain consistent endpoints across wafer lots.

6.2 Dielectric Erosion

Erosion is the thinning of the interlayer dielectric material in regions of dense copper metal arrays, relative to isolated ILD regions, after the barrier clearing step. It is driven by the pattern density effect: in dense copper arrays, the pad makes contact with both the copper lines and the narrow inter-line dielectric simultaneously, and the slurry — even if formulated for Cu:ILD selectivity — removes some dielectric along with the barrier metal during Step 2 polishing.

Erosion reaches its maximum in densely patterned regions with high copper density and narrow ILD spaces. Values of 10–30 nm of ILD erosion are common in production copper CMP processes on 45 nm and older technology nodes. At 7 nm and below, the ILD between copper lines may be only 10–15 nm wide, and erosion tolerances shrink to single-nanometer levels.

The primary formulation lever for erosion control is the Cu:ILD removal rate selectivity in the Step 2 slurry. A slurry with Cu:ILD selectivity exceeding 100:1 effectively stops removing ILD once the copper and barrier are cleared, dramatically limiting erosion in dense regions. Higher selectivity, however, must be balanced against the requirement to also remove the barrier metal efficiently — since high Cu:barrier selectivity and high barrier:ILD selectivity are often competing formulation objectives.

6.3 Surface Scratches and Particle-Related Defects

Scratches on the copper or dielectric surface after CMP arise from large abrasive particles, agglomerated particle clusters (a common failure mode when slurry pH or zeta potential shifts), hard contamination from the pad or conditioner disk, or pad debris embedded in the polishing interface. Even a single 100 nm scratch traversing a copper interconnect can increase local current density enough to initiate electromigration voiding under operating conditions, creating a latent reliability risk that may not manifest in functional testing.

Scratch density is characterized by automated brightfield and darkfield wafer scanning — typically at 0.1 µm or 0.05 µm sensitivity thresholds on production inspection tools. Advanced-node copper CMP processes typically target fewer than 10 scratches per wafer above the 0.1 µm threshold, with some logic fabs requiring fewer than 5 per wafer.

7. Two-Step Copper CMP: Bulk Removal vs. Barrier Clearing

Virtually all production copper CMP processes use at least two distinct polishing steps, each performed with different slurry formulations and polishing conditions. This two-step architecture is not an arbitrary process complexity — it reflects the fundamental incompatibility between the chemistry required for efficient bulk copper removal and the chemistry needed to stop cleanly on the barrier layer while minimizing ILD erosion.

Step 1: Bulk Copper Removal in Detail

The Step 1 slurry must be formulated for maximum copper removal rate efficiency. Alumina abrasives (colloidal or fumed, 30–100 nm primary particle size) provide the necessary mechanical aggressiveness. Oxidizer loading is relatively high — typically 2–5 wt% H₂O₂ — to ensure rapid surface oxidation across all pattern density regions. Step 1 polishing pressures of 1.5–4 psi are common, though this range is progressively tightening at advanced nodes where ELK dielectric stacks limit maximum allowable pressure.

The target for Step 1 endpoint is to stop polishing with 20–50 nm of copper remaining uniformly across the wafer, above the barrier metal surface. This copper “residual” is thin and uniform enough that Step 2 can clear it quickly and precisely without introducing significant dishing or erosion. Stopping too early leaves too much copper for Step 2 to clear within its process window; stopping too late risks direct exposure of the barrier metal, which creates a kinetic discontinuity in the polishing behavior (a sudden change in removal rate and friction coefficient) that disturbs planarity.

Endpoint detection for Step 1 most commonly uses motor current monitoring — the torque required to rotate the platen changes as the surface transitions from bulk copper to the thin residual copper layer above the barrier. Some production tools also use in-situ reflectance measurement (ISRM), which detects the change in optical reflectance as copper thins — copper has a distinct reflectance spectrum from the barrier metals, providing a thickness-sensitive signal. Eddy-current sensing — a non-contact magnetic induction technique — is used in closed-loop step-1 control systems where real-time copper thickness measurement is needed for run-to-run APC feedback.

Step 2: Barrier Clearing and Final Planarization

Step 2 slurry chemistry must simultaneously accomplish three things: remove the thin copper residual from Step 1, clear the barrier metal (Ta, TaN, Co, or Ru) with sufficient rate to not become the throughput-limiting step, and stop on the ILD without excessive erosion. The fundamental challenge is that these three objectives have conflicting chemistry requirements:

• Efficient copper removal requires moderate-to-high H₂O₂ and BTA concentrations similar to Step 1.
• Efficient barrier metal removal requires a harder abrasive or a chemistry that attacks Ta/TaN (which are kinetically resistant to most of the chemical agents that attack copper).
• Low ILD erosion requires high Cu:ILD and barrier:ILD selectivity, which pushes toward a gentler, lower-pressure process.

Modern Step 2 copper CMP slurry formulations resolve these tensions through carefully engineered abrasive surface chemistry (colloidal silica with specific surface functionalization), near-neutral pH (6–8, which moderates both copper dissolution and barrier attack), and barrier-specific additives that enhance Ta/TaN mechanical softening without introducing ILD-attack pathways. The result is a Step 2 slurry with Cu removal rate of 50–150 nm/min, barrier removal rate of 10–30 nm/min, and ILD removal rate of <1 nm/min — giving effective Cu:ILD selectivity exceeding 100:1 in optimized formulations.

In advanced-node processes using cobalt (Co) or ruthenium (Ru) liners instead of Ta/TaN, Step 2 slurry chemistry requires complete re-engineering. Both Co and Ru are far more susceptible to oxidative corrosion in the standard Step 2 pH and H₂O₂ conditions, and their electrochemical potentials create significant galvanic corrosion risks at the Cu/Co and Cu/Ru interfaces. JEEZ offers dedicated Co/Ru-compatible Step 2 barrier slurry variants within the CuS product series.

8. Advanced Node Challenges: Copper CMP at 7 nm and Below

As semiconductor technology nodes have progressed from 28 nm through 7 nm, 5 nm, 3 nm, and into the 2 nm-class era of 2026, the challenges confronting copper CMP slurry have become qualitatively more complex. Several structural and materials changes at advanced nodes fundamentally alter the constraints on slurry formulation and process integration.

8.1 Ultra-Low-k (ELK) Dielectric Fragility

Advanced-node BEOL ILD materials — commonly called extra-low-k (ELK) or XLK dielectrics — achieve dielectric constants below 2.5 through controlled porosity (20–40% void volume fraction) in an organosilicate glass (OSG) or SiCOH matrix. This porosity reduces the dielectric constant but simultaneously reduces the film’s mechanical stiffness dramatically: elastic moduli of 5–10 GPa are typical for modern ELK dielectrics, compared to 70 GPa for thermally grown SiO₂ and 160 GPa for silicon.

The fragility of ELK films sets a hard upper bound on CMP polishing pressure. Applied pressures above approximately 1.5 psi risk cohesive fracture within the porous ELK layer or adhesive delamination at the ELK/etch-stop interface (where a dense SiCN or SiN cap layer interfaces with the porous ELK, creating a mechanical impedance mismatch). At the most pressure-sensitive process levels in leading-edge 3 nm-class logic, copper CMP must be performed at pressures below 0.7 psi on the pad.

At these ultra-low pressures, the mechanical contribution to material removal becomes secondary, and the chemical contribution must be enhanced to compensate. This drives slurry formulations toward higher oxidizer concentrations, more aggressive complexing agents, and lower BTA loadings than would be used in conventional copper CMP — creating a tighter formulation window that is harder to maintain with lot-to-lot consistency.

8.2 Cobalt and Ruthenium Liner Compatibility

The Ta/TaN barrier/liner architecture that dominated BEOL from the 130 nm node through approximately 28 nm is being replaced in advanced-node processes. At 7 nm and below, Ta/TaN liner thickness must scale below 2 nm to maintain an adequate copper filling fraction in narrow trenches — but sub-2 nm Ta films cannot reliably prevent copper diffusion into the ILD under operating bias-temperature stress conditions.

Cobalt (Co) and ruthenium (Ru) have emerged as leading replacement liner materials. Their bulk resistivities (Co: 6.2 µΩ·cm, Ru: 7.1 µΩ·cm) are lower than Ta and TaN for very thin film geometries, and their compatibility with selective CVD deposition enables thinner, more conformal liners. However, both metals introduce significant challenges for copper CMP slurry:

• Co and Ru have much higher standard electrode potentials relative to copper than Ta/TaN, creating galvanic couples at the Cu/liner interface when both are exposed to the oxidizing slurry environment during Step 2.
• Acidic pH and H₂O₂ — standard conditions for Step 1 copper CMP — can rapidly corrode thin Co and Ru liners, producing metal ion contamination and surface pitting at the liner edges.
• BTA, which provides excellent copper corrosion inhibition, has much weaker inhibition efficiency on Co and Ru surfaces, requiring alternative or supplementary inhibitor systems.

JEEZ’s advanced-node CuS barrier slurry series incorporates Co/Ru-compatible inhibitor packages and near-neutral pH formulations (pH 6.5–7.5) specifically developed to protect Co and Ru liners while maintaining adequate Cu removal and low defectivity in Step 2 processing.

8.3 Narrow Damascene Trench Challenges

At 7 nm and below, copper interconnect line widths are below 20 nm and via diameters are below 15 nm. These dimensions introduce copper CMP challenges that have no analog at earlier nodes:

• Tightened dishing tolerances: Even 2 nm of dishing on a 15 nm-wide copper line represents a 13% cross-sectional area reduction — sufficient to shift via resistance above specification or create an electromigration lifetime failure. The acceptable dishing budget is now of the same order as the slurry’s static etch rate per minute, meaning any over-polish exposure is catastrophic.
• Restricted abrasive access: Abrasive particles larger than 30–40 nm cannot physically enter via features with diameters below 15 nm. This changes the effective abrasive contact mechanics inside features compared to flat open areas, creating systematic within-feature polishing rate non-uniformity that is difficult to model from first principles.
• Surface roughness requirements: Ra surface roughness targets on polished copper below 0.3 nm rms are standard at 7 nm and below. Achieving these roughness levels requires very narrow abrasive particle size distributions (D99/D50 < 2.5) and careful control of pad asperity contact stress distributions.

8.4 Hybrid Bonding Copper Pad CMP

Perhaps the most novel copper CMP application emerging in 2026 is the planarization of copper bonding pads for hybrid bonding — a direct copper-to-copper and dielectric-to-dielectric bonding technique used in wafer-to-wafer (W2W) and chip-to-wafer (C2W) advanced packaging. In hybrid bonding, CMP is required to polish the copper bonding pad surface to nanometer-scale planarity, with the unusual requirement of intentionally producing a slight copper recess (typically −1 to −3 nm) relative to the surrounding dielectric surface.

This controlled recess is essential for the thermocompression bonding mechanism: when two wafers are aligned and pressed together at elevated temperature, the dielectric surfaces bond first (preventing misalignment), and the recessed copper then expands under thermal stress to make intimate metal-to-metal contact. The dimensional tolerance on this recess — typically ±0.5 nm wafer-level standard deviation — is more demanding than any conventional copper CMP application. JEEZ has developed a dedicated CuHB slurry series for hybrid bonding pad CMP that achieves reproducible copper recess in the 1–3 nm range with wafer-level σ below 0.5 nm.

8.5 Slurry Innovation Trends in 2026

The leading trends in copper CMP slurry development as of July 2026 include: low-pressure formulations (<1 psi capability) using enhanced chemical activity to compensate for reduced mechanical input; BTA-free formulations using azole-derivative and mercaptobenzimidazole-based inhibitor systems for PFAS-adjacent chemical compliance; AI-assisted formulation screening using molecular dynamics simulations to predict inhibitor adsorption geometry and stability; POU delivery systems with real-time H₂O₂ concentration feedback loops for closed-loop oxidizer control; and dedicated Co/Ru liner-compatible barrier slurry chemistries for GAA (gate-all-around) NSFET and backside power delivery network (BSPDN) integration schemes.

9. Post-CMP Cleaning After Copper Polishing

CMP cleaning is not an optional post-process step — it is an integral part of the copper CMP sequence that determines the final defect level and surface chemistry of the polished wafer. Inadequate post-CMP cleaning is one of the most prevalent causes of yield loss in copper interconnect fabrication, and the cleaning process must be co-optimized with the slurry formulation for maximum effectiveness.

Types of Residues After Copper CMP

The post-copper-CMP wafer surface carries four distinct categories of residues, each requiring different cleaning chemistry and mechanical approaches:

1. Slurry abrasive particles: Colloidal silica or alumina particles embedded in the BTA-Cu surface film or trapped in pad debris on the wafer surface. These are typically the largest contributor to post-CMP particle counts and require both mechanical (brush) and chemical (citric acid dissolution of the BTA-Cu matrix holding them in place) removal.
2. BTA-Cu complex film: The Cu(I)-BTA passivation layer formed during polishing remains on the copper surface after the polisher. While only 2–5 nm thick, it can block via formation in the next metallization level if not completely removed. Dilute citric acid (0.1–1.0 wt%) or dilute ammonium hydroxide solutions effectively dissolve BTA-Cu complexes without attacking the copper surface itself.
3. Oxidizer byproducts: H₂O₂ decomposition generates OH• radicals that can react with residual organic species on the surface, creating hard-to-remove oxidized organic deposits. A brief DI water flush immediately after polishing stops the oxidizer reaction and limits this byproduct accumulation.
4. Copper ion contamination: Cu²⁺ ions from the slurry adsorb electrostatically onto negatively charged ILD surfaces and can penetrate into porous ELK dielectrics. In the transistor active regions below the BEOL stack, copper is a deep-level trap that causes MOSFET leakage and reliability degradation at concentrations as low as 10¹¹ atoms/cm². Chelation with citric or oxalic acid during cleaning converts adsorbed copper to soluble Cu-citrate or Cu-oxalate complexes that rinse away completely.

Post-Copper-CMP Cleaning Methods

PVA (polyvinyl alcohol) cylindrical brush scrubbing is the standard mechanical cleaning method after copper CMP. Counter-rotating PVA brushes make conformal contact with the wafer surface under controlled compression (~3–5 mm deflection) while cleaning chemistry is dispensed through the brush or onto the wafer directly. The scrubbing action combines mechanical particle dislodgement with chemical dissolution of the BTA-Cu matrix. Standard cleaning chemistry is 0.05–0.5 wt% citric acid, which simultaneously chelates copper ions, dissolves BTA-Cu complexes, and helps maintain particle suspension in the rinsate.

Megasonic cleaning uses high-frequency acoustic energy (800 kHz–2 MHz) transmitted through DI water or a cleaning chemical to dislodge sub-micrometer particles through controlled acoustic streaming and transient cavitation. Megasonics is highly effective for particles larger than 50 nm but must be applied carefully when ELK dielectrics are present: at ultrasonic power densities above approximately 1–2 W/cm², acoustic pressure waves can initiate delamination at mechanically weak ELK-etch-stop interfaces.

DI water (DIW) rinse quality directly influences post-CMP residue levels: resistivity above 18 MΩ·cm and particle counts below 100/mL at 0.05 µm are required for production-grade post-CMP rinsing. Final IPA (isopropyl alcohol) rinse and spin-dry are used to prevent water marks on the hydrophobic copper surface.

The leading-edge post-CMP cleaning trend in 2026 is dry-in/dry-out integrated cleaning modules — physically coupled to the CMP polisher — that minimize air exposure of the freshly polished copper surface between polishing and cleaning, reducing native oxide regrowth and particulate re-deposition from the fab environment.

10. How to Select a Copper CMP Slurry

Selecting the right copper CMP slurry for a given application requires systematic evaluation across multiple performance dimensions. Process engineering and procurement teams must align on a shared set of performance metrics, qualification criteria, and total cost of ownership factors before committing to a production slurry.

10.1 Key Performance Specifications

10.2 Slurry Qualification Protocol

A rigorous copper CMP slurry qualification typically proceeds through three evaluation stages, each requiring increasing investment but providing increasing confidence in production performance:

10.3 Total Cost of Ownership (TCO) Considerations

The purchase price of copper CMP slurry is rarely the dominant factor in total cost of ownership. The key TCO drivers that often favor a higher-specification (and initially higher-cost) slurry include:

• Slurry consumption per wafer: A slurry achieving target removal rate at lower flow rate (or at higher dilution ratio) directly reduces consumption cost. Even a 10% reduction in slurry flow rate translates to significant savings at high-volume production scale.
• Pad lifetime: Slurry chemistry affects pad glazing rate and pad wear. A slurry formulated for gentler pad interaction — while maintaining removal performance — can extend pad lifetime significantly, reducing pad cost per wafer and minimizing throughput loss from pad change cycles.
• Defect-driven rework: At advanced-node production costs exceeding $10,000 per wafer, even a 0.5% yield improvement attributable to lower defect density pays for a significant slurry price premium almost immediately. The total defect cost — including inspection, rework, and scrapped wafers — should always be included in comparative TCO models.
• Post-CMP cleaning chemistry compatibility: Slurries that require aggressive or specialized cleaning chemicals to achieve adequate cleanliness add to operational chemical cost and may introduce cleaning chemistry compatibility constraints on the cleaning tool and adjacent process steps.

11. JEEZ Copper CMP Slurry Products

JEEZ Copper CMP Slurry Product Lines

JEEZ’s copper CMP slurry portfolio is continuously updated to address emerging process requirements at advanced nodes. Our application engineering team provides on-site support for slurry integration, process window optimization, and defect troubleshooting at customer fab sites worldwide. Sample quantities for process evaluation are available with short lead times.

---

12.常见问题