Search

According to the trend towards high-resolution CMOS image sensors, pixel sizes are continuously shrinking, towards and below 1.0μm, and sizes are now reaching a technological limit to meet required SNR performance [1-2]. SNR at low-light conditions, which is a key performance metric, is determined by the sensitivity and crosstalk in pixels. To improve sensitivity, pixel technology has migrated from frontside illumination (FSI) to backside illumiation (BSI) as pixel size shrinks down. In BSI technology, it is very difficult to further increase the sensitivity in a pixel of near-1.0μm size because there are no structural obstacles for incident light from micro-lens to photodiode. Therefore the only way to improve low-light SNR is to reduce crosstalk, which makes the non-diagonal elements of the color-correction matrix (CCM) close to zero and thus reduces color noise [3]. The best way to improve crosstalk is to introduce a complete physical isolation between neighboring pixels, e.g., using deep-trench isolation (DTI). So far, a few attempts using DTI have been made to suppress silicon crosstalk. A backside DTI in as small as 1.12μm-pixel, which is formed in the BSI process, is reported in [4], but it is just an intermediate step in the DTI-related technology because it cannot completely prevent silicon crosstalk, especially for long wavelengths of light. On the other hand, front-side DTIs for FSI pixels [5] and BSI pixels [6] are reported. In [5], however, DTI is present not only along the periphery of each pixel, but also invades into the pixel so that it is inefficient in terms of gathering incident light and providing sufficient amount of photodiode area. In [6], the pixel size is as large as 2.0μm and it is hard to scale down with this technology for near 1.0μm pitch because DTI width imposes a critical limit on the sufficient amount of photodiode area for full-well capacity. Thus, a new technological advance is necessary to realize the ideal front DTI in a small size pixel near 1.0μm.