In the semiconductor devise manufacturing process especially in the post-CMP cleaning, PVA brushes are widely used for removing contaminants on a wafer. Normally, it is used with chemical solution, so the contaminants are removed with combined effects both chemical and mechanical actions. The cleaning characteristics using PVA brush of the mechanical effects are greatly depending on its lubrication condition. Therefore, many studies clarified the contact condition between the PVA brush and several surface in chemical solutions, and also engineers evaluated the friction or torques during PVA brushes operating. For the friction between PVA brushes and surfaces, our recent study clarified that the brush deformation was also important factor for deciding the friction. In this study, we observed contact area distribution during the PVA brushes scrubbing. We developed a new method for observing the real contact area of PVA brushes scrubbing by using total reflection light and evaluated the real contact rate with image analysis. We used prisms with two types of wettability and a collimating LED light. The reflected light rays were observed with a high-speed video camera. And contacting motions at the prism top surface covered with enough water were observed. The light rays are totally reflected at the surface with gas or liquid on it. In contrast, when there is a contacting solid, incident light transmitted to the solid because of high refractive index than the prism. Therefore, the transmitted part is shown as dark part on the image obtained high speed camera. In this study we performed mainly two experiments. The first experiments were for observing entire brush deformation. In these experiments, we observed contact area distribution during brush scrubbing to verify the assumption proposed by our recent study. We claimed that contact area distribution which was induced by the brush deformation was also important factor for the brush friction. The second experiments were to estimate the real contact rate using microscope. We evaluated the threshold to distinguish the real contact area from solid and fluid by difference of darkness. We used a spherical POM ball which has similar refractive index with PVA brushes. And we have determined the threshold for binarization according to Hertzian contact theory. Then we used the same threshold to evaluate the real contact rate from monochrome images of PVA brushes. Here the real contact rate defined by the area ratio between observation and real contact. Attached figure shows the typical image obtained contact area of PVA roller brush on a PMMA prism. The brush was roller type with skin layer and rotating at 100 rpm. In this case, the brush was compressed with 2 mm on the prism (d = 2 mm) and the dotted circles showed the undeformed size of nodules. From the figure, a gap between the two divided dark parts were clearly observed. And the front part forward of the gap was darker than the other parts. The darker points indicate the region where the PVA brushes more compressed. Therefore, we found that the front part of PVA brushes were more compressed compare with the rear part. From combined observation from the side during PVA brushes scrubbing, we also found that the front dark parts were the nodule side face of PVA brushes. In other words, PVA brushes were not only contacted with bottom area, but also its partly contacted with side face. And the side face was more compressed compare to the nodule bottom. Next, we evaluated the real contact rate. First, we confirmed that the contact area of compressing spherical ball well reproduced using Hertzian contact theory with single threshold. Then we applied this threshold for PVA brush images. As a result, the contact rate for PVA brush with skin layer was a few percent. In addition, no contact rate was observed using without skin layer brushes.