Performance Simulation and Architecture Optimization for CMOS Image Sensor Pixels Scaling Down to 1.0 $\mu\hbox{m}$

As the pixel dimensions of complementary metal–oxide–semiconductor sensors are approaching the wavelength of visible light, significant diffraction effects occur in the pixel architecture region, resulting in decreased optical efficiency and increased spatial crosstalk. By introducing the finite-difference time-domain approach, the performance of the typical 1.75-, 1.35-, and 1.05-$\mu\hbox{m}$ pitch pixels is simulated and analyzed, respectively, in this brief. Several new approaches, which are beneficial to overcome the physical limitations of the conventional pixel architecture and providing better device characteristics for the sub-2-$\mu\hbox{m}$ pixels, are developed by analyzing the simulation results. Among them, the optimization in microlens can decrease the optical power loss above the color filter array to a certain extent, and the dielectric stack height reduction and a novel metal light funnel structure can maximize the optical efficiency and minimize the spatial crosstalk. Finally, the performance comparisons demonstrate that the optical characteristics of the optimized 1.05- $\mu\hbox{m}$ pixel are comparable to those of the conventional 1.75-$\mu\hbox{m}$ pixel.