Improved Cu Recess Control via Surfactant-Modulated Electrochemical Kinetics in Post-CMP Cleaning
ABSTRACT
Hybrid Cu bonding has become crucial in fine-pitch 3D semiconductor integration for achieving ultra-high interconnect density. To ensure high bonding yield, the Cu pads must be oxide-free and extremely planar, with low surface roughness and tightly controlled Cu recess, before bonding [1]. Even minute oxides on the Cu surface can degrade bond quality, so pretreatments with mild organic acids, such as citric acid, can improve the yield of Cu-Cu bonding by removing oxides [2]. However, DI-water-based wet exposure can still induce Cu loss during rinsing due to aqueous etching, lead to recess formation in fine features [3]. During wet cleaning, Cu and the barrier metal can be electrically coupled through the electrolyte, and the galvanic current is determined by both the potential difference and the polarization kinetics of the anodic and cathodic reactions [4]. Triton X-100 was selected to improve wetting and residue lift-off on hydrophobic Cu(I)-BTA [5]. This study aims to formulate a post-CMP cleaner that minimizes Cu recess while maintaining effective removal of particles and surface oxides.
Patterned wafers containing 8 μm Cu pads embedded in a TEOS dielectric with a 60 nm Ti barrier were used to evaluate Cu recess, while electrochemical measurements were conducted on separate Cu and Ti blanket wafers. Three alkaline citric acid-based cleaning solutions were examined, including 0.5 wt% citric acid alone and the same solution with 400 or 800 ppm Triton X-100, and all solutions were adjusted to pH 12 using KOH. Electrochemical behavior was characterized using open‑circuit potential (OCP), electrochemical impedance spectroscopy (EIS), and potentiodynamic polarization (PP). Cu recess depth, surface morphology, chemical changes, and wettability were analyzed using atomic force microscopy (AFM), field-emission scanning electron microscopy (FE‑SEM), Fourier transform infrared spectroscopy (FT‑IR), and static water contact angle measurements, respectively. Post‑CMP wet exposure was simulated using a repeated contamination and cleaning sequence applied to patterned wafers. Fifteen cycles were carried out for each chemistry, where each cycle involved immersing the sample in an alkaline barrier-metal slurry for 1 minute under 200 rpm stirring, rinsing with deionized water, and immersing in a pH 12 cleaning solution for 1 minute under 200 rpm stirring.
Although the galvanic driving force is often discussed in terms of the potential difference (ΔE), ΔE alone is insufficient to describe galvanic corrosion behavior, as surfactant‑induced changes in reaction kinetics can dominate (Fig. 1a). Upon the introduction of Triton X‑100, the Cu polarization behavior was only marginally affected, whereas the electrochemical response of Ti changed significantly, primarily on the cathodic branch. At 800 ppm Triton X-100, the cathodic kinetics of Ti were strongly inhibited, as reflected by an increased cathodic Tafel slope, which shifted the mixed-potential intersection to a lower galvanic current density (Fig. 1b). In contrast, at 400 ppm near the critical micelle concentration (CMC), surfactant coverage was insufficient to effectively block cathodic reaction sites on the Ti surface. As a result, although the Cu–Ti potential gap did not increase relative to the citric‑acid baseline, cathodic inhibition was weakest at 400 ppm, leading to the highest galvanic current density and the most pronounced Cu dissolution. Consistent with these electrochemical trends, AFM measurements after 15 contamination-cleaning cycles revealed net changes in Cu recess of 13.7 nm for the citric acid alone , 17.7 nm for the 400 ppm Triton condition, and 11.5 nm for the 800 ppm Triton X-100 condition (Table 1.). The 400 ppm condition exhibited the largest Cu recess, corresponding to the highest galvanic current density and the smallest Ti cathodic Tafel slope, whereas the 800 ppm condition showed the smallest recess together with the lowest galvanic current density and the largest cathodic Tafel slope. These results demonstrate a clear inverse correlation between Cu recess and cathodic kinetic inhibition on Ti. A larger cathodic Tafel slope indicates slower cathodic reaction kinetics, limiting sustained galvanic current and suppressing Cu loss during wet cleaning. Overall, the 800 ppm Triton X‑100 condition, corresponding to approximately twice the CMC, provides the most effective suppression of Cu recess variation among the cleaning chemistries investigated. In conclusion, Cu recess during the post-CMP cleaning process was effectively controlled by the addition of a surfactant, which inhibits the cathodic kinetics on the barrier metal.
BIOGRAPHY
Arim Woo is a Master’s student in Materials Science and Chemical Engineering at Hanyang University, Korea. Their research focuses on wet processing and surface engineering for semiconductor applications, particularly post-CMP cleaning and hybrid Cu bonding. They investigate electrochemical reactions at Cu/barrier interfaces, with an emphasis on understanding and suppressing galvanic corrosion between Cu and Ta-based materials to improve bonding reliability in advanced packaging such as HBM.
Their work combines electrochemical analysis (OCP, polarization, and galvanic measurements) with surface characterization techniques including AFM, KPFM, and contact angle measurements. They aim to develop cleaning chemistries that minimize Cu recess and dishing while maintaining effective particle removal. Through this, they seek to establish more controllable and reliable post-CMP cleaning strategies for next-generation semiconductor integration.