During semiconductor manufacturing, processes such as thin-film deposition, etching, and metal interconnection progressively create topographic variations—peaks and valleys—on the wafer surface. If these irregularities are not promptly eliminated, they can compromise depth-of-focus control during subsequent photolithography, pattern transfer accuracy, and the fabrication of multilayer interconnect structures. Consequently, Chemical Mechanical Polishing (CMP) has become a critical process for achieving global wafer planarization in advanced integrated circuit manufacturing, with the abrasive particles in the polishing slurry serving as a key component that determines material removal efficiency, surface quality, and defect levels.

CMP Working Principle

CMP does not simply rely on hard particles to "grind away" the wafer surface; rather, it achieves material removal through the synergy of chemical and mechanical actions. Typically, components in the polishing slurry—such as oxidizing agents, complexing agents, and pH adjusters—first react with the wafer surface, inducing oxidation, hydrolysis, complexation, or other chemical changes in the outermost material layer to form a reaction layer that is relatively easy to remove. Subsequently, nano-abrasive particles dispersed in the slurry remove this reaction layer through the combined action of the wafer, the polishing pad, and the slurry itself. As fresh material surfaces are continuously exposed, the cycle of chemical reaction and mechanical removal repeats, ultimately achieving efficient, controllable material removal and planarization of the wafer surface. Therefore, harder is not necessarily better for CMP abrasives; their performance requires a balance among material removal rate, surface roughness, scratch defects, material selectivity, dispersion stability, and the difficulty of subsequent cleaning.

Currently, the most representative abrasives used in semiconductor CMP include silicon dioxide (SiO?), aluminum oxide (Al?O?), and cerium oxide (CeO?). These three materials differ significantly in terms of hardness, surface chemical properties, dispersion characteristics, and applicable materials, thereby offering distinct advantages for specific applications.

Silicon Dioxide: Wide Range of Applications and Outstanding Defect Control Capabilities

Silicon dioxide is a widely used type of abrasive in CMP slurries, with colloidal silica and fumed silica being the most common forms. Colloidal silica is typically produced via liquid-phase methods, allowing for relatively easy control over particle size, size distribution, and surface characteristics; in contrast, fumed silica is formed through high-temperature gas-phase processes and tends to exhibit chain-like or aggregated structures. In precision CMP for semiconductors, colloidal silica is particularly prevalent due to its superior dispersion stability and high particle uniformity.

Silica nanoparticles TEM

Silicon dioxide is characterized by moderate hardness and the ease with which its particles can be engineered at the nanoscale, alongside a well-established colloidal dispersion system. Compared to harder abrasives like aluminum oxide, it exerts a relatively gentle mechanical action on the wafer surface, thereby reducing the risk of deep and micro-scratches. By adjusting parameters such as particle size, particle size distribution, solids content, pH, surface charge, and surface functional groups, the interaction between the particles and the wafer can be further modified to control material removal rates and defect levels. Consequently, SiO? is not merely an abrasive providing mechanical action; it can also be viewed as a nanoparticle platform capable of being highly tuned through surface chemistry.

Silicon dioxide is suitable for CMP systems involving oxides, certain low-k materials, metals, and semiconductors, particularly for processes demanding high surface quality and strict defect control. However, as its mechanical removal capability is not the highest among the three mainstream abrasives, practical formulations often require the optimization of oxidizers, complexing agents, pH levels, and polishing process parameters to enhance overall material removal efficiency. In the context of advanced CMP, the primary advantage of silicon dioxide often lies not in simply pursuing the highest removal rate, but in achieving an optimal balance between high removal efficiency and minimal surface damage.

Aluminum oxide: Higher hardness and stronger mechanical removal capability.

Alumina is another classic CMP abrasive. Compared to silica, alumina possesses greater hardness and thus exerts stronger mechanical action on the wafer surface; it is traditionally used for the CMP of metals such as tungsten and copper, as well as in other processes requiring high material removal rates. Since enhancing the mechanical action of the abrasive can increase the material removal rate—within certain limits—alumina offers a distinct advantage in systems where processing efficiency is a priority.

Alumina nanoparticle TEM

However, high hardness also entails a greater risk of surface damage. If particle sizes are excessive, the particle size distribution is too broad, agglomerates are present, or polishing pressure is improperly controlled, alumina is more prone to causing surface defects such as scratches and micro-scratches. Consequently, modern alumina-based CMP systems increasingly prioritize nanoparticle size control, surface modification, and dispersion stability, rather than simply relying on the abrasive's inherent hardness to achieve high removal rates. For advanced semiconductor manufacturing, the core challenge in utilizing alumina abrasives lies in balancing robust material removal capabilities with the minimization of mechanical damage.

Cerium oxide: Combining mechanical action with unique interfacial chemistry.

Cerium oxide is a representative material used in CMP abrasives; its unique characteristic lies in the fact that, beyond the mechanical removal of silicon dioxide, it also engages in strong interfacial chemical interactions. Active cerium sites on the surface of cerium oxide particles interact with silanol groups on the SiO? surface to form Ce—O—Si bonds; consequently, the removal of silicon dioxide by cerium oxide does not rely solely on particle hardness and mechanical friction, but is further facilitated by chemical action.

Image of cerium oxide particles + diagram of the Ce–O–Si mechanism

It is precisely due to this characteristic that cerium oxide plays a pivotal role in STI (Shallow Trench Isolation) CMP. STI CMP typically requires the efficient removal of SiO? while minimizing the removal of the Si?N? stop layer; consequently, the SiO?/Si?N? removal selectivity is a critical factor. Cerium oxide exhibits a strong chemical affinity for SiO?; when combined with the modulation of other chemical components in the slurry, it enables high oxide removal rates and excellent material selectivity, making it one of the most representative abrasives used in STI CMP.

STI CMP process cross-sectional view

However, the strong interaction between ceria and SiO? also presents challenges for subsequent cleaning. Since CeO? particles tend to adhere to oxide surfaces—potentially leading to stubborn particle residues or cerium-ion contamination after polishing—post-CMP cleaning becomes particularly critical when using ceria-based systems. This underscores the fact that evaluating a CMP abrasive requires looking beyond its material removal performance during the polishing stage; one must also consider its impact on subsequent cleaning and defect control.

How do you choose between the three mainstream types of abrasives?

Overall, there is no absolute superiority or inferiority among silicon dioxide, aluminum oxide, and cerium oxide; rather, each offers distinct advantages tailored to specific process requirements. Silicon dioxide features relatively mild mechanical action and a mature dispersion system, making it suitable for applications requiring a balance between low defect levels and process stability. Aluminum oxide possesses higher hardness and stronger mechanical removal capabilities, making it ideal for scenarios demanding high material removal rates. Cerium oxide, leveraging unique interfacial chemical interactions with SiO?, demonstrates outstanding advantages in oxide CMP and STI CMP applications.

However, in actual CMP formulations, there is no fixed one-to-one correspondence dictating that a specific material must use a specific type of abrasive. Ultimate polishing performance is determined by a combination of factors, including abrasives, oxidizers, complexing agents, corrosion inhibitors, surfactants, pH levels, and polishing pads, as well as process parameters such as pressure, rotational speed, and flow rate. For copper CMP, different formulation systems can employ different types of abrasives; similarly, for SiO? CMP, different particles can be selected based on requirements for material removal rates, selectivity, and defect control. Therefore, the abrasive is merely a core component of the overall CMP chemical-mechanical system, rather than the sole factor determining performance.

Moving from "Hardness" to "Particle Engineering"

As requirements for defect levels and surface quality in advanced manufacturing processes become increasingly stringent, the research focus for CMP (Chemical Mechanical Planarization) abrasives is shifting. While early research prioritized rapid material removal, current efforts increasingly emphasize the properties of the nanoparticles themselves—such as size, particle size distribution, morphology, surface chemistry, and dispersion stability. Reducing abrasive particle size lowers the localized pressure exerted by individual particles on the wafer surface, thereby mitigating the risk of scratching; however, excessively small particles possess higher surface energy and are more prone to agglomeration. If nanoparticles (tens of nanometers in size) aggregate into large particles (hundreds of nanometers or even micrometer-scale), the risk of defects increases significantly.

Consequently, the development of advanced CMP slurries focuses not only on average particle size but also on strictly controlling the count of large particles, particle size distribution, and long-term colloidal stability. By manipulating Zeta potential, the ionic environment, and surface functional groups, researchers can minimize particle agglomeration while simultaneously altering the adsorption and reaction dynamics between the abrasive and the wafer surface. In recent years, composite abrasives, core-shell structured particles, and surface-functionalized abrasives—such as SiO?/CeO? composites and CeO?-coated particles—have emerged as key areas of research. The underlying strategy involves utilizing different materials to handle mechanical and chemical functions separately, thereby achieving a balance of high removal rates, low defect levels, and high selectivity.

With the entry of wide-bandgap semiconductors (such as SiC and GaN) and other novel materials into the advanced manufacturing landscape, abrasives like zirconia, titania, and nanodiamonds are garnering increased attention. Traditional SiO?-based systems often struggle to achieve optimal material removal efficiency when processing hard, chemically stable materials like SiC, necessitating the development of more efficient abrasives and surface chemistry systems. However, the high hardness of these novel abrasives also entails a higher risk of defects. Therefore, the future direction of development lies not merely in identifying harder materials, but in achieving more precise material removal through particle structure design and the modulation of interfacial chemistry.

Conclusion

The evolution of CMP abrasives—from silica and alumina to ceria—reflects the shift in semiconductor planarization technology from simple "mechanical grinding" to a process driven by the synergistic control of interfacial chemistry and nanoparticles. Silica has become one of the most widely used abrasives due to its excellent dispersion stability and relatively gentle mechanical action; alumina leverages its superior hardness to deliver high material removal rates; and ceria plays a pivotal role in STI and other oxide CMP processes thanks to its unique chemical interaction with SiO?.

For advanced semiconductor manufacturing, the ideal CMP abrasive is not necessarily the hardest or fastest-acting particle; rather, it is one that achieves a comprehensive balance among material removal rate, selectivity, surface roughness, defect control, dispersion stability, and post-CMP cleaning performance. Future CMP abrasives will increasingly trend toward nanostructuring, surface functionalization, hybridization, and customization for specific materials.