Monday, October 22, 2012

Atomically Flattened Surface Usefulness

Shigetoshi Sugawa from Tohoku University, Japan presented "Achievement, issues, and next steps of CMOS image sensors" at 6th International Workshop on Semiconductor Pixel Detectors for Particles and Imaging (PIXEL2012) held on September 3 – 7, 2012 in Inawashiro, Japan.

The presentation discusses pixel RTN limitations and improvements, HDR imaging with LOFIC, global shutter, fast frame rate, and more. One of the parts talks about atomically flattened surface advantages:


  1. It is interesting that the flatter surface still has 1/f noise but reduced considerably. So one might conclude some 1/f noise comes from surface roughness, and some from another source.

    The UV degradation, which I suppose comes from interface and SiO2 etc. trapping of carriers, seems to related to surface roughness, which makes sense from a field localization point of view (maybe). Interesting stuff.

  2. I think that the flatness reduces the total effective surface area in this case. Like a chemical capacitor, the roughness increases considerably the effective surface area.
    -yang ni

  3. YN, are you saying 1/f noise is related to surface area? That would suggest smaller transistors have smaller 1/f noise, for one thing. Note also the reduction factor is 10x in the data shown.

    1. Well, yes and no. Because when the generation center nb is small, you will have more isolated events. Of course when the surface is of better quality and less generation centers, there will be an improvement.

      This is maybe the reason of the transition from 1/f to RTS.

      -yang ni

  4. I am trying to understand the basic mechanism here. Is it because roughness "adds" energy levels in the band-gap that can act as traps (or fast states) that they are bad for 1/f noise? If so, is it like a stress induced mechanism that could be characterized for instance by Raman spectroscopy?

  5. There seem to be many contributors to 1/f noise. RTS, for example, is considered a contributing element. Interface coulombic scattering from charged traps is another. I heard another theory that it is inherent in turbulent fluid flow and thus could be present in a perfect device. There seem to be many processing conditions that affect 1/f noise and it seems surface flatness is among these. What is interesting is that it is a broad spectrum effect -- or at least a knee is not revealed in the data presented.


All comments are moderated to avoid spam.