Ultrasonic agitation generates cavitation bubbles and micro-jets that remove strongly adhered particles from the wafer surface, preventing contamination during subsequent bonding steps .
The ultrasonic cleaning step immediately follows the CIS Backside Si CMP2 process in the production of nanosca
le Backside Illuminated (BSI) CMOS Image Sensors . Chemical mechanical polishing (CMP) is considered one of the dirtiest processes in wafer manufacturing, leaving a large amount of abrasive particles, metal contaminants, and chemical residues on the wafer surface . For BSI devices, achieving an ultra-clean, defect-free backside silicon surface is critical because this interface directly determines the sensor's optical coupling efficiency and charge storage performance . This cleaning step ensures the complete removal of CMP residues, preparing a pristine silicon surface for the subsequent oxide hard mask deposition and backside passivation steps . The physical mechanism of particle removal relies on overcoming strong nanoscale adhesion forces . Following CMP, submicron particles initially adsorb to the wafer surface via hydrogen bonding, and over time, interfacial chemical reactions and diffusion mixing significantly increase their adhesion strength . According to the Hamaker model, van der Waals attraction dominates at these small separation distances, making simple fluid shear insufficient for particle detachment . Ultrasonic agitation introduces high-frequency acoustic waves into the liquid, producing cavitation bubbles and micro-jet effects that generate sufficient localized shear stress to dislodge these strongly adhered particles . To prevent particle redeposition, the cleaning chemistry is strictly controlled to modify the surface zeta potential, leveraging DLVO theory to maximize electrostatic repulsion between the detached particles and the substrate [P1, P4]. Material and method selection for this step carefully balances particle removal efficiency with the need to protect the fragile, thinned backside substrate . While contact-based PVA brush scrubbing is highly effective for many post-CMP applications [P1, P3], non-contact ultrasonic or megasonic cleaning is heavily relied upon to prevent mechanical scratching on the ultra-thin active silicon layer (Engineering Practice). Specialized equipment can utilize surface tension to elevate the cleaning fluid, allowing the wafer backside to contact the liquid locally while ultrasonic vibration is applied, which prevents the entire wafer from being immersed and protects front-side bonding interfaces . Process parameters, such as acoustic power and chemical concentration, interact to dictate the cleaning efficacy; higher chemical concentrations drive the complexation and desorption of metal contaminants, while acoustic power dictates the mechanical detachment force [P1, P4]. At the 40nm node, the extreme thinness of the backside silicon layer makes the process highly sensitive to mechanical stress and surface roughening . Excessive ultrasonic power can induce surface roughness on the substrate, which acts as a scattering center for incident light and degrades the overall quantum efficiency of the image sensor [P4, T1]. Therefore, the cleaning process window is tightly constrained, requiring a precise combination of chemical weakening of particle bonds and strictly regulated acoustic energy to ensure high yield without compromising the structural integrity of the thinned photodiode substrate .
Sign in to continue through all 417 steps