Among various abrasive systems, silica sol (colloidal dispersion of nano-silica particles in water or solvent) is widely used in the polishing of dielectric materials such as silicon wafers, silicon dioxide, and silicon nitride due to its moderate hardness, good dispersibility, and low risk of scratching equipment. However, semiconductor-grade applications impose extremely stringent requirements on silica sol: trace metal impurities can diffuse into devices, causing leakage or threshold voltage drift; non-uniform particle size or the presence of large particles can lead to micro-scratches on wafer surfaces, directly reducing yield; poor colloidal stability results in unstable polishing rates, affecting batch-to-batch consistency. Therefore, how to prepare silica sol that simultaneously meets the demands of "ultra-high purity, monodisperse particle size, controllable morphology, and long-term stability" has become a common challenge for both materials science and the semiconductor industry.
Challenge 1: Removal of trace metal impurities
Metal impurities are a major factor causing wafer surface defects and device failure. Metal ions such as Na, Fe, Al, Ca, Mg, Cu, and Pb can remain on the wafer surface after polishing, compromising device insulation and causing leakage, or diffuse into the silicon substrate during high-temperature processes, leading to parameter drift. Consequently, the metal impurity content in silica sol used for chip CMP slurries is typically required to be below 1 ppm, and for advanced processes, even below 1 ppb per individual metal. In the preparation of silica sol, however, trace metals are introduced not only from raw materials (silicon powder, water glass, silicate esters) but also from reaction vessels, pipelines, and additives. Conventional filtration and ion exchange cannot thoroughly remove impurities at the ppb level.
Challenge 2: Precise control of particle size monodispersity
When a polishing slurry with a broad particle size distribution is used, large silica particles tend to create scratches on the silicon wafer surface and cause fluctuations in polishing rate or localized over-polishing. Therefore, the particle size and its uniformity are critical. Typically, semiconductor-grade silica sol requires particle sizes in the range of 10–50 nm; for high-end processes (e.g., 5 nm node and below), even finer control is needed, around 10–30 nm. However, during the growth of silica particles, secondary nucleation and agglomeration easily occur, making truly monodisperse particles difficult to achieve. Moreover, in large-scale production, minor fluctuations in parameters such as temperature, pH, and feed rate can disrupt particle size uniformity, imposing extremely high demands on process precision.
Challenge 3: Controllable particle morphology
Although spherical monodisperse silica sol used as abrasive can achieve good surface quality, spherical silica particles tend to roll and have small contact areas, leading to low polishing efficiency. In recent years, leading foreign companies have focused on developing non-spherical, edge-free, smooth-surfaced silica abrasives such as dumbbell-shaped, cocoon-shaped, and ellipsoidal particles. These particles offer advantages such as high specific surface area, softness, and low scratch tendency, showing great promise for semiconductor CMP polishing. However, the preparation of such morphologies remains challenging.
Challenge 4: Ensuring long-term stability
Long-term stability is fundamental for the industrial application of silica sol. Semiconductor polishing slurries need to be stored for 6–12 months or more, requiring that the silica sol does not gel, stratify, or undergo particle growth under broad pH (8–11) and high solid content (30%–40%) conditions. The technical difficulty lies in the high surface energy of nanoparticles, which makes them prone to agglomeration due to reduced electrostatic repulsion or hydrogen bonding. Additionally, temperature changes and impurity ion contamination accelerate colloidal destabilization. Stability control becomes exponentially more difficult at high solid contents, necessitating surface modification and system optimization to enhance long-term stability.
Mainstream preparation methods of electronic-grade silica sol for polishing slurries
Currently, the main methods for preparing high-purity silica sol are ion exchange, silicon powder hydrolysis, and sol-gel (silicate ester hydrolysis). These three methods differ significantly in raw material selection, product purity, particle size control, and production cost, making them suitable for different levels of semiconductor polishing requirements.
1. Ion exchange method
Also known as the water glass method, this is the most mature and widely used process. It uses industrial water glass (sodium silicate) as raw material, which is passed through a cation exchange resin to remove Na⁺, and then through a weak-base anion exchange resin to remove chloride and other impurities, yielding a dilute silica sol and a high-purity active silicic acid solution. A stabilizer is then added to adjust the pH to 8.5–10.5, and through nucleation and particle growth reactions, a monodisperse, size-controllable silica sol is produced, which is finally concentrated and purified by ultrafiltration or centrifugation.
Advantages: Suitable for large-scale industrial production, low raw material cost, controllable particle size down to 10–20 nm, and after deep purification, metal ion content can be controlled to the ppm level, meeting the needs of low-to-mid-end semiconductor polishing.
Disadvantages: Water glass raw materials contain many metal impurities, making purification difficult; strict control of reaction concentration, pH, temperature, etc., is required to avoid non-uniform particle size or gelation; the process also generates large volumes of salt-containing wastewater, and resin regeneration is costly, with high environmental pressure.
2. Sol-gel method
This method uses high-purity tetraethyl orthosilicate (TEOS) or tetramethyl orthosilicate (TMOS) as the silicon source. In an alcohol solvent, hydrolysis and polycondensation are catalyzed by acid or base to produce SiO₂ nanoparticles, followed by solvent exchange and concentration to obtain high-purity silica sol. By precisely controlling reaction conditions, molecular-level doping can be achieved, producing silica nanoparticles with uniform size, controllable morphology, and high purity. TEOS is generally preferred for industrial production and laboratory research due to its lower cost, lower toxicity, higher safety, and slower, more controllable hydrolysis rate. TMOS hydrolyzes very quickly, leading to vigorous reactions and faster formation of silica sol, which facilitates a highly cross-linked gel structure in a short time, but the reaction is difficult to control.
3. Silicon powder hydrolysis method
This method uses high-purity silicon powder as the raw material, which reacts with pure water under the catalysis of an inorganic or organic base (e.g., sodium hydroxide) to form hydrated silicic acid, which then polymerizes to form silica sol. The purity of the product depends on the purity of the silicon powder, enabling the preparation of extremely high-purity silica sol with very low impurity levels. At the same time, parameters such as SiO₂ particle size, viscosity, pH, density, and purity are easier to control compared to other methods. However, morphology control is difficult, and there is a risk of hydrogen explosion.