Holes Better than Microlenses

Peer-reviewed Progress In Electromagnetics Research journal (PIER-B) published a paper:

"Optical Performances of Lensless Sub-2micron Pixel for Application in Image Sensors"
R. Marinelli and E. Palange
Dipartimento di Ingegneria Elettrica e dell'Informazione, Universita dell'Aquila, Via G. Gronchi 18, 67100 L'Aquila, Italy
Progress In Electromagnetics Research B, Vol. 31, 1-14, 2011.

The paper proposes to use a gold film with sub-micron aperture holes on top of polymer-based lightguide instead of microlens. The simulation shows that light diffracted from the aperture, mediated by the generation of surface plasmon-polariton modes, is entirely confined within the lightpipe. Full 3D simulations on a Bayer cell composed of four pixels demonstrate that it is possible to improve optical efficiencies and crosstalk rejection between adjacent pixels by 30% and by a factor 12 respectively better than those ones obtained in similar pixels equipped with microlenses for pixel sizes close to 0.97um.

In simulations the radius of the circular section at the bottom of the lightpipe, i.e., the photodiode area, was fixed at 0.4 um, the pixel size D varied from 1.75 to 0.97 um and the ILD thickness, L, was set to 2.5 um. The metal film thickness T is 0.2 um.

The conclusion in the paper says:

"We have demonstrated that lensless pixels arranged in the Bayer cell configuration as used in commercial image sensor devices and making use of lightpipe structure have better optical performances respect to those ones obtained for pixels equipped with microlenses located at their entrance. By performing full 3D electromagnetic simulations of the light propagation inside the Bayer cell, we have shown that in terms of the normalized optical e±ciency and crosstalk, the lensless pixel geometry is always more efficient [with] respect to the microlens equipped counterpart. In particular, simulations demonstrate that the decrease of the normalized optical efficiency and the increase of the crosstalk effects as a function of the reduction of the pixel size is largely less effective for the lensless pixel geometry and this can foresee the real possibility to decrease the pixel size for values less than 1 um. Moreover, the proposed geometry is completely compatible with the actual large scale image sensor manufacturing process and overcomes the problems related to the fabrication of microlenses with large numerical apertures."