[adge]

You can't actually "see" DNA itself because it is smaller than the wavelength of light. Those groves are 50nm and the wavelength of blue light is about 480nm.

So how then are the images generated?

It uses a technique called Fluorescence Microscopy: https://en.wikipedia.org/wiki/Fluorescence_microscope

You essentially stick some special molecules on the DNA that absorb light in one frequency and then they emit light in another frequency. So you blast the molecules with light of one frequency and then use a dichroic mirror to filter that light out and you only see the emissions and thus you see where the DNA is, but you don't "see" the DNA itself.

Like STM itself what we mean when we "see" something at those length scales is interesting. Scanning Tunnelling Microscopy is like a blind man reading braille - not really "seeing" anything but getting enough info to describe the picture.

source: I used to work in the same lab as Dan. Hi Dan!

[feralimal2]

This is a useful explanation - thanks!

What does this mean to the value of what we are "seeing"?

The example used before - that it is equivalent to a deaf man seeing a music visualisation - is apt. It is some sort of model, but not particularly close. It might still be useful, of course.

[Groxx]

Not "seeing a music visualisation" (was it edited maybe?), reading braille. STM involves sticking a probe microscopically close to the thing-being-scanned, and reading how the surface's atoms deflect it. Which is very close to "literally feeling" the surface, as all "touch" is just electrons repelling each other at a distance - this is just at a slightly larger distance. https://en.wikipedia.org/wiki/Scanning_tunneling_microscope

Maybe a more human-scale-friendly analogy would be "finding the hot burner by moving your hand near it". In STM, each atom is a hot burner. You can pretty accurately figure out the arrangement of burners on your stove without needing to see or touch it.