Defect detecting, positioning, and sizing by KLA-Tencor Surfscans (6200,6400,SP1-classic, and SP1-TBI)

Unpatterned wafer surface control in terms of defect and particle measurement is of paramount importance for the semiconductor industry. This metrology is used by the industry for product qualification, manufacturing tools monitoring, and defect root cause analyses. Light-scattering-based wafer surface scanners are widely used for these purposes. Yet, little work has been done to quantify the performance of these instruments. In this paper, we report the experimental measurements of defect XY positioning accuracy, sizing accuracy, and capture rate performed on several wafer surface scanners (KLA/Tencor Surfscans 6200A, 6200B, 6220, 6400, SP1-Classic, SP1-TBI). The XY-positioning and sizing performance of these instruments were characterized using a standard XY calibration wafer (KLA-Tencor Part No. 358770) and a bare silicon wafer with PSL spheres (0.72 μm). It was found that the capability of the wafer scanners to accurately measure defect locations was highly variable. Map-to-map comparisons (without correction for coordinate system misalignment) showed XY positioning accuracies varying from 50 μm to over ± 700 μm. For the 6x00 series tools, there was a clear dependence of mapping accuracy on the orientation of wafers in the wafer cassette. With optimized map-to-map coordinate system alignment, positioning accuracies varied from nominally ± 30 μm to over ± 130 μm, with the 6200 and SP1-TBI instruments providing the best performance. The counting performance of these instruments under either low or high throughput was characterized using a bare silicon wafer with environmental particles and a bare silicon wafer with PSL spheres (0.155μm). It was found that capture rate of PSL spheres was compatible for 6200 series and SP1 series. However, the experimental data indicate that SP1-TBI under-estimates the total number of environmental particles by about 60% when high throughput is used. The experimental data also shows that 6400 not only under-estimates the defect size but also reports less defects (about 10%) than other instruments. These results begin to establish some of the limitations of evaluated instruments as well as a deeper understanding of proper operation needed for more quantitative application.