NIR camera sensitivity limit

[Stefano]

Where is the upper sensitivity of silicon sensors? How deep can they go into the near infrared? Discussions about this topic started here: https://www.ultravio...r-verification/ and here: https://www.ultravio...__fromsearch__1.

[Andy Perrin]

As I said, once silicon goes transparent the sensitivity is exactly zero. (The pixels are leaky photon buckets, and a transparent pixel is a bucket with no bottom.) And transparency happens at a specific energy given by the formulas for absorption in semiconductors here:

https://en.m.wikiped...irect_band_gaps

(Scroll down to the section on absorption.)

 

 

Silicon is an indirect bandgap semiconductor, so we use the second formula at the link, and since the phonon energy is very small, that implies that the bandgap is the place where it becomes completely transparent (h*nu = Eg). However, the bandgap varies with processing, doping, temperature, and voltage. So if you want an exact number for the cutoff you still have to measure.

[Stefano]

I read on Wikipedia that silicon is transparent to red light at liquid helium temperatures... imagine seeing that. If the bandgap is 1.14 eV, the maximum wavelength detectable is 1087.58 nm, but since the formula is more complicated than that, you can’t just calculate it. A silicon filter would be a great way to see how far you can go.They aren’t very easy to find, but try to mount a polished slab of silicon on your lens, and then point your camera at a bright light source, such as the sun or a halogen lamp. You may see something beyond 1100 nm.

[Andy Perrin]

As I said in the other threads, I do seem to get some light from 1100-1150nm. No need to point at the sun. That number 1.14eV for the bandgap is very approximate. As I said, the actual bandgap varies with temperature, voltage, doping, and processing. And on top of that, the thickness of silicon in the sensor also matters. Even if absorption coefficient is very tiny, if the well is deep enough you will still get a few photons. But it won’t change the absolute cutoff at the bandgap where alpha is 0. I’m afraid 1150nm is the practical limit, Stefano.

[nfoto]

I read long ago in a Kodak technical paper that photography with film was limited to below 1150 nm.

[dabateman]

With my Lp1100 blocked I do see something. But I think I am just pass the 1100nm edge. I like Andy's optimistic 1150nm is probably the maximum. Less than that might be more realistic.

Unless you change the sensor design, as in Foveon sensors seem to be very IR sensitive. They seem to be the most sensitive to Ir that I have worked with.

From the thesis:

DETECTING NEAR-UV AND NEAR-IR

WAVELENGTHS WITH THE FOVEON IMAGE

SENSOR

by

Cheak Seck Fai

December 2004

 

Using a calibrated source to 1200nm, this is the Foveon sensitivity for the pre Meril sensors.

[Andy Perrin]

Even Foveons are silicon. I think at least one of the channels is very thick for Foveon by the nature of the design, so that helps milk a few more photons at the long end, but the location where alpha goes to zero is determined by the bandgap of the material only, and so all the additional thickness buys you is better sensitivity closer to that limiting number.

 

The reason for using germanium on silicon for the TriWave is that the Ge band gap is 0.66eV at room temp, which converts to 1240/0.66 = 1880nm as the absolute cutoff. At absolute zero, the bandgap of Ge is 0.74eV, which gives 1240/0.74=1680nm. The sensor is cooled to -80C = 193K, so it is actually between those two limits.

[dabateman]

Andy not arguing with you. I just present the Sigma sensor data as best case and look it drops off to zero at or just after 1100nm.

So thats your silicon limit.

[Andy Perrin]

Andy not arguing with you. I just present the Sigma sensor data as best case and look it drops off to zero at or just after 1100nm.

So thats your silicon limit.

Got it.

[Stefano]

So, if you want to go beyond 1100-1150 nm... Things become expensive, low resolution, low sensitivity... a whole new mess. I still have to try my “phosphoresce erasing” method, I think it can work. The difficult part about seeing in long wavelengths is that each photon carries a low energy, so it can’t do things like fluorescence, or promoting electrons to a higher energy state (unless the bandgap of your semiconductor is low enough). With UV you have plenty of energy, and if your sensor can not detect them directly, you can always use them to excite something, because they have more than enough energy to do so.

[Andy Perrin]

True that. But it’s not completely true that you can’t do up-conversion with the proper phosphor, as I showed on here, it’s just very inefficient and requires intense light to work. So yes there is always a trade off. The triwave has good sensitivity but low resolution. That can be overcome by making panoramas for stationary scenes.

[Stefano]

I read long ago in a Kodak technical paper that photography with film was limited to below 1150 nm.

I think that infrared films were limited at roughly 900 nm. Maybe that paper was about the maximum theoretical limit for compounds in general (you need a bit of energy to induce chemical changes in a crystal, it is already surprising that (with the right film) a weak infrared photon can do that). If you see a remote control LED you may be beyond 900 nm, since it usually peaks at 946 nm but has a pretty broad spectrum.

 

Anyway, film or digital, 1150 nm seems to be the absolute limit.