Experimental residual stress mapping in bonded semiconductor devices

Recent advances in the microelectronics industry have led to increasing demand for on-line quality control throughout the entire fabrication process. The complexity of modern devices requires an increasing number of processing steps that amplifies the need for reliable inspection methods. It has become particularly important to develop techniques that quantify the residual stresses generated during each processing step. Localized stresses arising from defects at bonded interfaces can cause downstream processing failures. We conducted a broad range of experiments using controlled defects in order to better understand stresses that arise from bond defects. These defects consist of silicon dioxide 'mesas' that are patterned onto a silicon wafer that is subsequently bonded to a 'smooth' wafer. The defect dimensions were chosen to represent particles commonly found in a cleanroom. Each bonded pair was inspected using an infrared grey-field polariscope (IR-GFP) to obtain quantitative measures of the residual stress fields surrounding the 'mesas'. High-resolution x-ray topography (XRT) measurements were taken on the same samples to validate the measurements made using the IR-GFP with a standardized tool. While the XRT is capable of much higher spatial and stress resolution than the IR-GFP, both methods agreed qualitatively. These experiments show that, as expected, the debond radius increases with mesa height and decreases with increasing bond energy. The maximum residual stress around trapped particles was found to vary between 1.60 and 3.43 MPa for wafers treated using the standard RCA cleaning method.