10
\$\begingroup\$

silicon_wafer. Image taken from Can somebody identify this 12" silicon wafer?

So this silicon wafer looks multicolored (and beautiful). But how does it get multicolored like a rainbow? What is the reason for this phenomenon?

\$\endgroup\$
8
  • 2
    \$\begingroup\$ It's a trick of the light. Nothing more, nothing less. \$\endgroup\$
    – Andy aka
    Jul 3, 2021 at 10:26
  • 3
    \$\begingroup\$ It's called iridescence. \$\endgroup\$
    – Andy aka
    Jul 3, 2021 at 10:29
  • 4
    \$\begingroup\$ I’m voting to close this question because it's about the physics of light and not about EE. \$\endgroup\$
    – Andy aka
    Jul 3, 2021 at 10:43
  • 6
    \$\begingroup\$ @Andyaka end users may not care about how electronic components are made but plenty of electrical engineers do. \$\endgroup\$
    – uhoh
    Jul 4, 2021 at 1:55
  • 2
    \$\begingroup\$ Plenty of EEs care about how to cook food and how beer is made and what TV programmes are worth watching and if there is life on Mars but, all those subjects are irrelevant to this site. \$\endgroup\$
    – Andy aka
    Jul 4, 2021 at 8:57

1 Answer 1

14
\$\begingroup\$

detail of rainbow pattern on wafer

If it's an old image or an image of any wafer made by a old process or with large node technology that does not have "dummy patterning" then the only thing that will make the scribe lanes show color is thin film interference.

enter image description here Source

However, starting at roughly 0.25 microns (early 1990's) dummy patterns were added everywhere (including the scribe lane) to wafers at topographical layers (as opposed to implants) to handle problems with processes that suffered from microloading and other pattern density-dependent processes. These included reactive ion etching and chemical-mechanical planarization (or polishing) as well as problems coating with the thinner photoresists necessary for 248 nm Deep UV photolithography which was introduced around the same timeframe. While optical proximity correction is a different think entirely, the adding of larger features like this is often folded in to the overall OPC corrections flow of mask pattern generation.

If that was the case for this wafer, then this could be either thin film or diffraction effects (a bit like DVDs do)

If it's thin film then the film has to be thick enough to produce different colors at slightly different angles for this photo, which often is the case after an oxide passivation covered with a nitride cap. The top, often silicon-rich nitride has an index of refraction of 2.0 or higher, and so Fresnel reflection is higher and the air-nitride interface acts more like a top "mirror" than an air-oxide (or spin-on glass) circa 1.5 index interface would.

But with technology nodes of dimension so much smaller than a wavelength of light, how could it be diffraction?

Yes, once a feature is below a half-wavelength in air, you can not see colorful diffraction patterns for that wavelength looking at it at that wavelength, even edge-on with a flashlight.

But the dummy patterns don't always have to be minimum resolvable feature size, it depends on the details of the process and the step, but some can be much larger.

\$\endgroup\$
2
  • 2
    \$\begingroup\$ An observation. Colors arising from iridescence can be distinguished from colors arising from diffraction by the presence/absence of the extra-spectral color magenta. Magenta in not a monochromatic color, and doesn't occur in regular diffraction. The soap bubble in your answer clearly has magenta. I am not sure, but I would say that the silicon wafer does not, meaning the colors come from diffraction, rather than thin film effects. \$\endgroup\$ Jul 26, 2021 at 3:12
  • \$\begingroup\$ @MathKeepsMeBusy That's a fascinating point! I'd never thought of that. Yes multi-order thin film interference can have overlapping peaks, say n=3 red + n=4 blue, but a large pitch grating that allows higher order diffraction (with a spacing $d$ of several wavelengths) can do that too. So perhaps we can further suggest that this is diffraction from features on the order of a wavelength or two at most. \$\endgroup\$
    – uhoh
    Jul 26, 2021 at 3:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.