If you shine a laser into diamond centered and perpendicular to the table face, the beam will be split and returned out of the table to show a pattern of dots (some bright, some dim). The pattern is unique to each diamond.

Gemprint.com uses this to fingerprint diamonds. You can see some images of "gemprints" here: https://www.gemprint.com/light-performance.html

This is precisely the sort of randomly insightful comment I’ve come to love from HN.

I agree. I actually found the GP link more interesting than the posted link. Not that the original wasn't interesting.

I did something similar in college in the 90's, except using cut crystals. One laser in, hundreds of beams out (faded, of course). You know the kind of faceted crystal balls that some people put in their cars and hang from the rearview mirror that cast rainbows everywhere when the sun hits it. Looked cool with a fog machine in the dorm.

Sounds like a Swiss watch company could use that for proving authenticity.

Company Seals are also a pretty big deal in East Asia, which are usually a red stamp. If several could be combined (e.g. CEO, CTO, CFO seals all generating a unique interference pattern), that would be a provable signature that the document is legitimate.

Looks similar to an X-ray crystallography diffraction pattern, except that requires X-rays.

X-rays are just one frequency of light. You can use any.

Also you can use electrons to get electron crystallography which is (arguably) more powerful as you can reform the original image after, it requires WAY smaller crystals, can see lighter atoms, and more.

You could use neutrons as well

> The pattern is unique to each diamond.

Oh boy, don’t let the NFT folk hear about this

Gemprint... the OG NFT.

thats really neat, i wonder if you can use this print during manufacturing to get a more perfect cut

Now I wanna know whether this can be reversed, constructing a 3d model from the dot pattern!

Using Differentiable Rendering (reconstructing scene setup parameter(s) from an image), possibly, although it would depend on the number of unknown params.

Mitsuba2 had an example of matching caustic patterns by modifying geometry I think...

My gut tells me it would work like a hash function, the diamond -> dot pattern function being trivial, but the dot pattern -> diamond function taking exponentially more effort

There is nothing intrinsically hard about it. Crystallography is the science of reversing this dot pattern.

I would have loved to see ray-traced images of diamonds tho.

I'm glad you posted this. I felt like that kid that ate paste in school and was too afraid to post that it would have been better with pics.

I was lost as soon as the article started talking about a maximum of 3 reflections... and nothing in the picture gave a hint as to why/what was meant.

I wanted to see images also - no matter how crude they be.

The math is sadly above my knowledge, but how cool!