Microscopes
How We See Small
The Limit Abbe Found in 1873
By the end of the nineteenth century, optical microscopes had stopped improving. Not because the glass was imperfect, or the makers were out of ideas, but because Ernst Abbe worked out in 1873 that there was a floor and they had reached it. Two things closer together than roughly half the wavelength of the light used cannot be told apart, by any lens, ever.
The exact statement has a second term in it, and the second term is worth knowing. The floor is the wavelength divided by twice the numerical aperture, a number that measures how wide a cone of light the lens can gather and what the light travels through on the way in. That is why the best objectives are lowered into a drop of oil. Oil bends light more than air does, the cone the lens can accept widens, and the floor drops by about a third.
For visible light that is about two hundred nanometers. It is enough to see cells comfortably, bacteria clearly, and the larger structures inside a cell as vague shapes. It is nowhere near enough for a virus, a ribosome, a protein, or an atom. An enormous fraction of what matters in biology and chemistry sat permanently below the floor.
The limit is the same physics that sets what a telescope can resolve, arrived at from the other direction. Waves squeezed through an opening spread, and detail finer than that spreading is not merely hard to see – it never makes it into the image at all. The identical wall stands at the scale of galaxies.
Bring a Shorter Wave
The limit is set by wavelength, so the way through it is to illuminate with something whose wavelength is shorter. In 1924 de Broglie proposed that matter has a wavelength too, shrinking as the particle’s momentum grows. Electrons are easy to accelerate, so their wavelength is easy to make extremely small.
An electron accelerated through a hundred thousand volts has a wavelength of a few thousandths of a nanometer. That is a small fraction of the width of a single atom, and roughly a hundred thousand times shorter than visible light. The first electron microscope was built within a decade of the idea, and the wall that had stopped optics for nearly sixty years was simply walked around.
Electrons cannot be focused with glass, so the lenses are magnetic fields shaped to bend the beam the way a lens bends light. The whole column must be under vacuum, because electrons scatter off air. And the sample must be very thin, because electrons cannot pass through much of anything.
Those magnetic lenses turn out to be the new floor. Otto Scherzer proved in 1936 that a static, rotationally symmetric electron lens always has spherical aberration, and always of the same sign, so it cannot be cancelled by pairing lenses the way optical designers cancel aberrations in glass. The beam therefore never focuses as tightly as its wavelength would permit. Even instruments with modern multipole correctors resolve about fifty picometers, more than ten times coarser than the three-picometer wave they are working with. The wall moved. It did not disappear.
Those constraints are the price. A living sample cannot survive vacuum, cannot be sliced thin, and would be cooked by the beam regardless. So for decades electron microscopy meant looking at things that had been dehydrated, embedded in resin, sliced, and stained with heavy metals – and then wondering how much of the structure was real and how much was an artifact of the preparation.
Freeze It So Fast It Cannot Notice
The way out was to stop preparing the sample and start freezing it. Plunge a thin film of solution into ethane cooled by liquid nitrogen and the water has no time to organize into ice crystals, which would tear the structure apart. It solidifies into a glass instead, and the molecules inside are held exactly as they were in solution.
Each individual molecule can only take a tiny dose of electrons before it is destroyed, so each image is almost pure noise. The trick is that the film contains hundreds of thousands of copies of the same molecule, frozen in every possible orientation. Software identifies each one, works out which way it was facing, and averages them together. Noise falls, structure emerges, and a three-dimensional map is reconstructed from a crowd of nearly useless pictures.
Better detectors and better algorithms turned this from a niche technique into the dominant one over a few years around 2013, an episode the field openly calls the resolution revolution. It won the chemistry Nobel in 2017, and it is now routine to see individual atoms in a protein that has never been crystallized. Much of what is known about how drugs bind to their targets comes from these maps.
Stop Looking and Start Feeling
There is a completely different way to beat the limit, and it is to stop using waves at all. Bring an extremely sharp tip close to a surface – within a nanometer – and something strange becomes measurable. Electrons tunnel across the gap even though classically they cannot, and the tunneling current is so violently sensitive to distance that changing the gap by the width of a single atom changes it by a factor of ten.
So you scan the tip across the surface and continuously adjust its height to hold the current constant. The record of how the tip had to move is the surface, atom by atom. This is scanning tunneling microscopy. The instrument was built in 1981 and was resolving the individual atoms on a silicon surface within two years. There is no image in the optical sense at all. There is a map, assembled from a very large number of individual measurements.
Atomic force microscopy came next and dropped the requirement that the sample conduct electricity. It senses the tiny force between the tip and the surface instead of a current, which means it works on insulators, on polymers, and on biological material in liquid. Living cells can be imaged while alive, and a single molecule can be pulled apart while the force needed is recorded.
The same tip can also push. Individual atoms have been dragged into position on a surface deliberately, which turned the instrument from a microscope into a very slow assembler and demonstrated that manipulation at the atomic scale is a technical problem rather than an impossible one.
Beating the Light Limit With Light
Electron microscopes and scanning tips both give up on light. Then, in the 2000s, people found ways to get past Abbe’s limit while still using an ordinary optical microscope on a living cell, which had been widely considered impossible.
The loophole is that the limit says you cannot separate two things emitting at the same time. It says nothing about things emitting at different times. So make almost all the fluorescent markers in the sample dark, and let a sparse random few switch on. Each isolated one produces a blurry blob, but the center of that blob can be located far more precisely than its width – the same way you can find the middle of a fuzzy dot to better than its size. Record the position, switch that set off, let another random set light up, and repeat many thousands of times.
Stack all the recorded positions and an image appears with detail ten to twenty times finer than the limit allows. A different approach reaches the same place from the other side: illuminate with a bright spot, then overlay a doughnut-shaped beam that switches off everything except the very center, shrinking the effective spot far below the diffraction size. Both were recognized with the 2014 chemistry Nobel, and both are now standard in cell biology.
None of These Are Photographs
The word microscope stops meaning what it used to somewhere in the middle of this page. Nothing produced by cryogenic electron microscopy, a scanning tip, or a super-resolution technique is a picture in the sense that a photograph is. Each is a reconstruction, assembled computationally from measurements that individually show almost nothing.
That is not a criticism, but it does move where the errors live. A reconstruction depends on its assumptions, and the failure mode is not a blurry image – it is a sharp image of something that was not there. The field takes this seriously and validates maps by splitting the data in half and checking the two halves agree independently, which is the same instinct behind any statistical threshold for discovery.
There is also a hard physical trade that never goes away. Seeing something small means hitting it with something energetic, and something energetic damages what it hits. Every technique here is negotiating between resolution and destruction, which is why cryogenic samples get one very brief look and why the highest-resolution images of biological material are all of things that were already dead and frozen.
It is worth naming where that trade actually comes from, because it is easy to file under the wrong law. A short wavelength means a large momentum, and a large momentum delivered into a molecule breaks it. That is de Broglie plus bookkeeping, a fact about probes and damage, and no amount of cleverness gets around it. The uncertainty principle is a separate and deeper statement, that a quantum state cannot hold sharp values of position and momentum at once whether or not anyone measures it. Both are floors. Only one of them is about what the instrument does to the sample.
Every Limit Has a Word to Attack
Every step above came from taking a limit seriously enough to find its exact statement, and then noticing which word in that statement could be attacked. Abbe’s floor had two terms in it, so attack both: shorten the wavelength, and widen the cone. The limit assumed lenses, so remove the lens. It assumed simultaneous emitters, so stagger them in time. None of these broke the physics. Each one found a condition in it that did not have to hold.
That is worth carrying beyond microscopy. A limit precisely stated is a much more useful object than a limit vaguely felt, because a precise one tells you where its own assumptions are.
