Simulation and experiment study of carbon dioxide snow cleaning in silicon material

The miniaturization and integration increment of silicon-based optoelectronic devices have placed greater demands on surface cleaning technologies. Carbon dioxide snow cleaning technology achieves cleaning through jets formed by the coexistence of solid, liquid, and gaseous phases of CO₂ . Solid CO₂ particles sublimate and expand, impacting the contaminants, causing them to detach from the surface. The cleaning effect of liquid CO₂ relies on its ability to dissolve organic pollutants and its adhesion force on surface contaminants. Gaseous CO₂ carries away the dislodged contaminants. This technology is cost-effective, leaves no residue, is non-damaging, environmentally friendly, and meets the stringent cleaning requirements of fields such as optics and chip manufacturing. In this study, the changes in flow fields, temperature, and thermal stress inside the device during the carbon dioxide snow cleaning process were investigated through fluid dynamics and thermodynamic simulations. Navier-Stokes equation, known as the first-principles in fluid dynamics, was used to construct numerical simulation models for the flow field, temperature field, and stress field. Corresponding simulation software was developed using C++, and further validation and refinement of the numerical simulations were performed using COMSOL. Both methods produced results of a stable flow velocity of 350 m/s and a thermal stress maximum of 10⁷ Pa. The comparison of these results with the critical cleaning velocity (approximately 10 m/s) and the stress damage threshold (on the order of 10⁸ Pa) indicates that carbon dioxide snow can achieve cleaning without causing damage. An experimental platform for carbon dioxide snow cleaning was used to clean silicon wafers, and a high-precision Zygo interferometer was employed to measure the surface profile before and after cleaning. The experiment compared the contamination coverage rates before and after cleaning, revealing that effective removal rate (ER) exceeded 95%. The peak-to-valley (PV) difference in the surface profile matrix provided by the high-precision interferometer showed changes of no more than 0.04 wavelengths before and after cleaning. The results demonstrated that carbon dioxide snow effectively removed both organic and inorganic contaminants without altering the surface profile. This conclusion is corroborated by the simulations, further confirming the effectiveness and non-destructive nature of the cleaning process.