Home
Biography
CV
Research
Publications

Optimization of micron-size pixels in solid-state image sensors (VIS/near-IR)

My research on image sensing focuses on accurate modeling of light-matter interaction and the consequences of fundamental light properties on solid-state image sensor performance as image sensor technology scales.

The majority of solid-state image sensors are being fabricated in complementary metal oxide semiconductor (CMOS) technology. The steady progress in CMOS image sensor technology development can be measured by the steady increase in pixel count. The increase has been achieved by Moore ’s law which permits a regular decrease in pixel size. Increasing pixel count has the clear benefits of reducing cost and allowing integration of image sensors into small mobile devices where the form factor is limited (e.g., cell phones and PDAs). The issues involved in pixel scaling, however, differ from the general issue of technology scaling. Sensors must interact with light, and the properties of light do not scale. For example, the amount of light (irradiance) available at the image plane does not scale with pixel size.

Figure: Mean photon count incident on a pixel as a function of pixel size (μm) and photometric exposure (lux-sec). Counts are calculated for an equal-energy spectral source. The range of photometric exposure values corresponds to the darker half of a typical computer display imaged with an f/2.8 lens and a 10-ms exposure duration. The lines represent equal-photon counts. Spatially uniform photon noise becomes visible at an SNR of 33:1. This level corresponds to the bold dotted equal-photon line at 1,000 photons/pixel.

The figure measures the number of incident photons that can be captured by an ideal pixel (100% fill-factor, 100% quantum efficiency; no pixel induced noise); it also defines an upper bound on the pixel signal quality. Any loss of light due to scattering, imperfect detector quantum efficiency, the presence of color filters, incomplete fill-factor, will decrease the number of photon absorptions and decrease signal quality. In this research I apply advanced optical modeling techniques to analyze the optical path of CMOS image sensor pixels.

Figure: Electromagnetic field simulation (2D) of two neighboring pixels in a CMOS image sensor (left: pixel with green color filter, right: pixel with blue color filer) when illuminated with green light (λ=555 nm). Outlines depict the microlenses that are used to focus the light onto the photodetector in the silicon substrate at the bottom. Also visible is the dielectric stack that supports the CMOS backend technology (metal interconnect layers and vias). The amplitude of the y-component of the electric field is shown as a function of x and z (red and blue indicate large positive and negative values; green is the zero level). Note the finite spot size at the substrate level; it's dimensions are governed by diffraction due to the wave nature of light.

In older image sensor technologies, a geometrical optics analysis typically sufficed; but as technology scales and pixel elements become commensurate with the wavelength of light, more sophisticated modeling is required to achieve accurate results. Electromagnetic field modeling of the light incident on image sensor pixels more appropriately describes the light-matter interaction inside a pixel. Since the electromagnetic field distribution the pixel has a strong and direct influence on sensor performance and image quality, our approach offers the possibility of optimizing the design by modifications to the geometric and material parameters. This research is crucial to the continued development in the field of image sensing. In addition, it can be easily extended to any field in which light interacts with nano-size electronic integrated circuits, e.g., on-chip optical interconnects.

Featured publications:

Back