Backgrinding reduces silicon wafers to the thickness required for advanced packaging. However, the grinding process leaves a damaged surface layer whose roughness directly determines die strength, downstream bonding yield, and electrical performance of thinned devices. Controlling this roughness in production is non-trivial. The ground surface must typically meet roughness specifications below a few nanometres Ra. At the same time, measurement must keep pace with grinder throughput.
lineWLI Advantage
lineWLI is ideally suited to this task: its delivers nanometer-level roughness characterisation in a single shot with no mechanical scanning. Each measurement takes less than 10 miliseconds, making it fast enough for inline integration directly at the grinder output.
lineWLI measured the surface roughness of four wafer surfaces at six locations each. This diagram shows the location of the measurement points. Please note that in measurement points 1 and 2 the line is perpendicular to the radial direction whereas for measurement points 5 and 6 the line extends in the radial direction. Measurement points 3 and 4 lie in the center of the wafer This measurement setup detects if the roughness depends both on the location and on orientation:
Wafer Backside
The first measurement was on the backside of a 200 mm single side polished wafer. Optically the wafer had a matte surface finish that reflected light with a diffuse appearance. The results show a small radial but no directional dependency. The profile corresponds to position 1:
Wafer Frontside
Now, we measured the polished front side. The roughness was two orders of magnitude smaller than before. Remember, lineWLI measures each profile within 10 ms. So we can easily achieve sub-nanometer resolution in vibrationally noisy environments. To demonstrate we had the system on a table with no noise isolation and measured the same line on the wafer 40 times back to back. One standard deviation of the elevation measurement is 0.1 nm. This noise translates into a noise of the Roughness measurement of less than 10 pm!
Another Wafer Backside
We measured the backside of another 200 mm single side polished wafer but from another vendor. Even though the backside of this wafer had the same optical appearance to the naked eye, lineWLI measured a roughness that was 50% higher than the first wafer:
Sensor Wafer Backside
The final measurement was the backside of a wafer that had sensor dies on the front. The backside had a mirror-like appearance but with a visible haze in a swirling pattern. The roughness measurements show a dependence on the direction. Measurement position 1 and 6 and 2 and 5 show different values by more than a factor of two. To confirm the hypothesis of directionality, measurements 7 & 8 perpendicular to measurements 5&6 were made. They show roughness values in line with the measurements 1&2
The measurements perpendicular and parallel to the radial direction look much different. Left Position 2, right position 5
