Thursday, October 17, 2013

Image Sensors for EUV Photolithograthy

Extreme UV photolithoraphy challenges extend to image sensor design for mask and wafer inspection. KLA-Tencor's patent application US20130264481 "Back-illuminated sensor with boron layer" by Chern Jehn-Huar, Ehsani Ali R., Delgado Gildardo, Brown David L., Chuang Yung-Ho Alex, and Fielden John gives a nice overview of these challenges:

"Inspection systems are typically in use 24 hours per day with only short stoppages, the detectors are exposed to large doses of radiation after only a few months of operation.

A photon with a vacuum wavelength of 250 nm has energy of approximately 5 eV. The bandgap of silicon dioxide is about 10 eV. Although it would appear that such wavelength photons cannot be absorbed by silicon dioxide, silicon dioxide as grown on a silicon surface must have some dangling bonds at the interface with the silicon because the silicon dioxide structure cannot perfectly match that of the silicon crystal. Furthermore, because the single dioxide is amorphous, there are likely also some dangling bonds within the material. In practice, there will be a non-negligible density of defects and impurities within the oxide, as well as at the interface to underlying semiconductor, that can absorb photons with deep UV wavelengths, particularly those shorter than about 250 nm in wavelength. Furthermore, under high radiation flux density, two high-energy photons may arrive near the same location within a very short time interval (nanoseconds or picoseconds), which can lead to electrons being excited to the conduction band of the silicon dioxide by two absorption events in rapid succession or by two-photon absorption. EUV photons have very high energies (13.5 nm in wavelength corresponds to photon energy close to 92 eV) and are capable of breaking silicon-oxygen bonds as well as strongly interacting with defects and contaminants in the oxide. Electron and charged-particle detectors typically have to detect electrons or charged particles with energies of a few hundred eV or higher. Energies greater than 10 eV can readily break silicon-oxygen bonds.

As indicated above, high-energy photons and particles can break bonds and ionize atoms in a silicon dioxide layer. Because silicon dioxide is a good insulator, free electrons created in the silicon dioxide may have lifetimes of ms or longer before recombining. Some of these electrons may migrate into the semiconductor material. These electrons create electric fields within the silicon dioxide and between the silicon dioxide and semiconductor. These electric fields can cause electrons created in the semiconductor by absorption of photons to migrate to the surface of the semiconductor and recombine, thereby resulting in lost signal and reduced detector quantum efficiency. Near continuous use of the instrument means that there may be little, or no, time for recovery of the detector, as new free charges are created as fast as, or faster than, they can recombine.

High-energy particles and photons can also cause irreversible changes to the silicon dioxide. Such changes can include reconfiguration of the bonding of atoms or migration of small atoms within the silicon dioxide. At normal operating temperatures of the detector, which are typically in a range from around room temperature to about 50° C., these changes will not recover. In particular, it is known that conventional silicon photodiodes used as EUV detectors degrade in efficiency with use.

The silicon dioxide layer on the surface of semiconductor detectors significantly reduces the efficiency of those detectors for low-energy (less than about 2 kV) electrons. Some low-energy electrons are absorbed by the silicon dioxide, thereby causing the silicon dioxide to charge up and deflect subsequent arriving electrons. Because a native oxide will always form on an exposed silicon surface, silicon detectors necessarily must have some oxide on their surface. Growing or depositing an alternative dielectric material (instead of the oxide) on the surface of the semiconductor results in a much higher density of defect states at the semiconductor to silicon dioxide interface. These defects reduce the quantum efficiency of the detector, especially for photons or charged particles absorbed close to the surface of the semiconductor.
"

The proposed solution is to use BSI sensor with pure boron deposited on the backside to form a shallow pn-junction. The metallic boron layer effectively shields the junction from all surface defects that might exist in the oxide or AR coating on the back surface:

10 comments:

  1. Or, take any silicon sensor, put about 1 micron of CsI on it, get 5 photons at 550 nm per incident x-ray photon (probably collect 2) and completely block the x-rays from hitting the silicon.

    ReplyDelete
    Replies
    1. 1um CsI can not stop all the X-ray photons. You have to put a FOP with high X-ray absorption coefficient to stop all the X-ray photons. Please advise me where you can find 1um CsI deposition facility? Thanks.

      -yang ni

      Delete
    2. The absorption length for 92 eV photons in crystalline CsI is about 50 nm so 1 micron attenuates by about 10^9. Not zero, true. Possible sources include RMD in Massachusetts, HTE Labs in California and Applied Scintillation Technologies in the UK. Hamamatsu might do it on their own devices.

      Delete
    3. Thanks Dave !
      How compare CsI to Lumigen for UV conversion please ?
      -yang ni

      Delete
    4. Lumogen has an absorption length of about twice that of CsI if it is in a dense film. I do not know enough about the physics of Lumogen to say much more but I suspect that it has to be in some sort of dispersion medium to work properly. Also, it probably only emits one visible photon per absorbed x-ray photon so the photon yield would be lower. On the other hand, it is certainly much easier to apply. Yellow fluorescent markers work in the near UV. Check to see if the ink fluoresces under13.5 nm illumination. Maybe BASF has 13.5 nm data.

      Delete
  2. I would like to give the advice to KLA-Tencor to check out the PhD thesis of Lei SHI (TU Delft). In his 4-years PhD project he did exactly the same as what is written down in claim 1 of the patent.

    ReplyDelete
    Replies
    1. Albert, if the sensor is packaged in vacuum condition after the SiO2 removing, is this a viable way?

      Delete
    2. I am not sure I know what a membrane is. Is it a thinner layer suspended by a thicker layer (like the illustration), or is it like a thin layer of silicon in BSI?
      And what does "pure" mean? If it has diffused into the silicon then it is not pure anymore. And how does this compare to the 20
      + years of work done at JPL on backside devices for short wavelength illumination?
      Well, it is just an application. Hopefully the examine will consider all these things.

      Delete
    3. There is an as deposited pure boron layer and if the temperature is high enough ( 750-800C) an in diffused part. The JPL work is MBE this is probably via epi reactor.

      Delete
  3. The layer of pure B is just a few atom layers thick, not more than a few nm. It should be as thin as possible, but if it becomes to thin, the boron seems to cluster and areas without boron coverage show up. That is not what you want to have. All these details are written down in Lei Shi's PhD thesis.
    BTW, there are already DUV-photodiodes used in commercially available equipement ! IT REALLY WORKS !

    ReplyDelete

All comments are moderated to avoid spam.