In the organic electronic world, it is often argued that silicon can not do NIR. But is it true. In this brief article, I touch upon the latest results in NIR sensing in silicom CMOS imagers and explain the basic ideas behind it.
The chart above- from OmniVision- shows the latest sensitivity of silicon to NIR. It shows an EQE 50% @ 940nm with 2.9um pixel!
OmniVision was a pioneer here. Several years ago it annonuced its first vesion of the the so-called Nyxel pixel, which is the brandname for their NIR-sensitive Si pixel.
The basic idea is build up DTI or Deep Trench Isolation, which is an important innovation in the CMOS imaging industry. It is DTI - and improvements therein- which have really enabled and sustained much of the pixel scaling in the CMOS imaging industry in recent years.
A problem with pixel scaling was that cross-talk, meaning that as pixels shrunk in size there would be optical and electric cross talk between them. with DTI, the pixels are propely isolated from each other, minimising cross-talk, and enabling pixels to be shrunk.
DTI is very advanced today. You can see the latest examples below. Just consider the aspect ratio! In general, with DTI and pixel scaling, the active area of the CIS (CMOS Image Sensor) is becoming thicker- the innovation here between the different gens of the Nyxel appears to be the same: a deeper pixel which gives more chance of light absorption .
As for NIR sensing in Silicon, the basic idea is to lenghten the propogation path of light in silicon. Since silicon has a low absorption in NIR, the only way to increase total absorption is to increase the total apparent thickness. With DTI, the light is trapped within the pixel, bouncing off the walls multiple times, so as to increase its absoption chance. Furthermore, in this approach, the top of the pixel is somewhat roughened, causing light scattering as the light enters the pixel, further increasing the propogation path.
The image below shows the first version of Nyxel, demonstrating how the NIR sensitivity has already improved. Note that this technology is not just limited to Nyxel, and all other players like OnSemi also have similiar offerings.
One notable application area for NIR is in automotive sensing. This is especially useful for LIDAR sensors because many LIDARs operates at 904nm or longer wavelenghts.
Note that this still does not give a technological path to SWIR. For this, InGaAs, Ge, or other hybrid (QD or ORganic CMOS) approaches will be needed
