Post-CMP Cleaning

As the feature sizes of integrated circuits continue to shrink, the impact of chemical-mechanical polishing and subsequent cleaning processes in the back-end-of-line copper interconnect fabrication on device performance and long-term reliability has become increasingly prominent. Although electrical performance tests at both the wafer level and the package level may indicate compliance, residual copper species and surface microdefects can still gradually evolve into catastrophic failures during device operation. Therefore, elucidating the morphological evolution of copper residues and the mechanism behind the formation and amplification of micro-pit defects during post-CMP cleaning is of great engineering significance for developing efficient cleaning strategies and ensuring device reliability.

I. Forms of Copper Residues and Their Impact Mechanisms on Reliability

1.1 Multiform Characteristics of Copper Residues

After Cu CMP, copper residues do not exist in a single form; they mainly fall into the following three categories:

(1) Surface-free Cu²⁺ ions: These ions are physically adsorbed onto the surface of the dielectric layer in the form of hydrated ions. Their surface density typically ranges from 10¹⁰ to 10¹² atoms/cm², and they exhibit local enrichment in the edge regions of the copper wires.

(2) Copper enrichment within micro-pits: During CMP, nanoscale surface depressions—typically about 5–20 nm deep and 20–100 nm in diameter—are formed. These depressions often accumulate high concentrations of copper ions or metallic copper particles internally.

(3) Complexed residues: Stable complexes formed with additives such as benzotriazole (BTA) or glycine (glycine)—for example, Cu-BTA and Cu-glycine—can be difficult to remove completely during conventional cleaning processes.

1.2 Reliability Failure Mode Induced by Copper Residue

1.2.1 Short-term failure behavior

Time-Dependent Dielectric Breakdown (TDDB): Under the influence of an electric field, residual copper ions migrate toward the cathode, forming conductive pathways within the dielectric layer and significantly shortening the lifetime of the dielectric layer. Experiments show that as the level of copper contamination increases, the TDDB lifetime can drop sharply from over 10 years to failure within the initial testing phase.

Deterioration of interlayer signal integrity: Increased copper ions enhance the leakage conductivity of the dielectric layer, intensifying capacitive coupling effects between adjacent interconnects and leading to degraded high-frequency circuit performance.

1.2.2 Long-term Reliability Degradation

Reduced electromigration lifetime: Copper ions enriched at the surface and grain boundaries accelerate the diffusion of copper atoms along grain boundaries, reducing the electromigration lifetime by 30%–50%.

Increased stress-induced migration sensitivity: Micro-pits, acting as stress concentration points, promote void nucleation and growth during thermal cycling, ultimately leading to increased interconnect resistance and even open-circuit failure.

II. Sources, Magnification Mechanisms, and Control Strategies for Micro-Pit Defects

2.1 Generation of Micro-Pits and Definition of Process Responsibility

Impact of CMP Process: Primarily determines the initial nucleation sites of micro-pits (such as grain boundaries and defect sites) and their sub-nanometer-sized embryo sizes (typically <5 nm, below the conventional detection limit).

Impact of the cleaning process: Whether the dominant micro-pits are amplified by chemical or mechanical effects during subsequent cleaning to a detectable scale, thereby influencing their final morphology.

2.2 Magnification Mechanism of Micro-Pits During the Cleaning Process

• Chemical amplification mechanism:

The cleaning solution has an excessively high corrosion rate on copper and the substrate material (e.g., >10 Å/min).

The complexing agent exhibits overly strong selective complexation toward metal ions at the pit edges.

Atoms at the edges of micro-pits exhibit higher reactivity due to their coordination unsaturation.

• Mechanical amplification mechanism:

The brushing pressure induces local stress concentration at the edges of micro-pits, thereby promoting material removal.

High rotational speeds enhance interfacial mass transfer and reaction kinetics, thereby intensifying pitting corrosion.

2.3 Case Study: Practical Equipment Validation and Optimization of the 28 nm Cu Interconnect Cleaning Process

Problem Identification: During mass production, the defect density of a certain 28 nm Cu interconnect layer unexpectedly increased from 0.05/cm² to 0.3/cm². The primary defect type is micro-pits with a depth ranging from 15 to 25 nm.

• Diagnostic analysis:

The static cupping test showed that the cleaning solution has a corrosion rate of 8 Å/min, which complies with the process specifications.

However, dynamic testing conducted on mass-produced single-wafer cleaning equipment—simulating actual brushing conditions (brush pressure, rotation speed, and flow rate)—revealed local corrosion rates as high as 20–30 Å/min, significantly exceeding the permissible range.

The root cause lies in the fact that the synergistic effect of chemical corrosion and mechanical brushing is underestimated in real production environments, and laboratory cupping data fail to fully reflect the interfacial strengthening reactions induced by brushing.

• Optimization Measure—Two-Step Cleaning Method:

Step 1: Pure Chemical Passivation Treatment—Use a BTA solution with a pH of 5.5, immerse the surface without brushing for 45 seconds to achieve initial passivation.

Step 2: Gentle Brush Cleaning—Clean using a mildly chelating alkaline cleaning solution with a pH of 6.0 under gentle conditions: brushing speed of 150 RPM and pressure of 80 g/cm².

• Verification result:

The defect density has dropped to 0.08/cm²;

Under actual brushing process conditions, the surface corrosion rate stabilizes at approximately 10 Å/min.

The residual copper on the surface has been reduced below the detection limit.

Key takeaway: The determination of the post-CMP cleaning process window must be based on validation using actual production equipment and scrubbing conditions. Relying solely on static cup-test data may lead to misjudgments of the actual cleaning risks.

Conclusion

Controlling copper residues and suppressing micro-pits in the post-CMP cleaning process are critical steps for ensuring the reliability of nanoscale copper interconnects. This study highlights that the diversity of copper residues and the amplification effect of micro-pits during cleaning are the primary mechanisms leading to latent device failures. Process validation conducted on actual production equipment has demonstrated that a two-step cleaning approach can effectively balance the roles of chemical passivation and mechanical cleaning, significantly reducing defect density and metal residue. Future work should further focus on the systematic co-optimization of cleaning chemistry, brush-dragging dynamics, and CMP processes to meet the cleaning challenges posed by increasingly advanced process nodes.

For more technical details, please consult our technical advisor at: 13861996325!

▲ Technical Consultation ▲ Follow the Kexinwei WeChat Official Account