The label says 40 nm, your instrument says 180 nm, the microscope says 40 nm. All three are right

A supplier's datasheet says 40 nm. You disperse the powder, run it on the dynamic light scattering instrument down the corridor, and get 180 nm. You put a drop on a grid, take it to the electron microscope, and there they are — clearly, unmistakably, about 40 nm across.

Nobody in that story is wrong, and nobody is being dishonest. The three numbers disagree because they are answers to three different questions, and the reason nanomaterials characterisation confuses people so reliably is that all three answers get written down using the same word: size.

Underneath that confusion is a hard physical fact. You cannot look at a nanoparticle. Every number anyone quotes about one is the output of an instrument that measured something else and converted.

Why light gives up at 200 nm

An optical microscope cannot resolve two points closer together than roughly half the wavelength of the light illuminating them. This is Abbe's diffraction limit, and it is not an engineering shortcoming that a better lens or a steadier stage could fix — it is a property of waves. With visible light around 400 to 700 nm, the practical floor is about 200 nm.

A 40 nm particle is five times below that floor. Under the best light microscope ever built it is not a small blurred dot; it is nothing at all. Nor can you get there by magnifying harder — empty magnification enlarges the blur along with everything else.

There is a well-known exception worth naming so it is not mistaken for a general answer. Super-resolution fluorescence techniques, recognised with the 2014 Nobel Prize in Chemistry, genuinely beat the diffraction limit — but they do it by making individual fluorescent labels blink or switch on and off and locating each one's centre precisely. That is a superb tool for tagged biological structures and no use whatsoever for a bottle of zinc oxide, which does not fluoresce and cannot be labelled one particle at a time.

So for materials, the answer is to stop using light.

Electrons see smaller because their wavelength is smaller

An electron accelerated through a large voltage behaves as a wave with a de Broglie wavelength thousands of times shorter than visible light — at the 100 to 300 kV used in electron microscopes, a few picometres. Diffraction stops being the constraint entirely; what limits real instruments is the quality of the magnetic lenses, and modern aberration-corrected microscopes resolve below an ångström, which is smaller than the spacing between atoms.

Two instruments do most of the work, and they show genuinely different things.

Scanning electron microscopy (SEM) rasters a focused beam across the specimen surface and collects the electrons knocked loose from it. The image is a map of surface topography, with an enormous depth of field, which is why SEM images of powders look like landscape photographs of a strange terrain. It shows shape, texture, how particles sit together, and the overall character of a batch. Non-conducting samples charge up under the beam and have to be sputter-coated with a few nanometres of gold or carbon first — that coating is itself a small addition to whatever you then measure.

Transmission electron microscopy (TEM) sends the beam through a specimen thin enough to be transparent to electrons, and the image is a projection through the object rather than a view of its surface. This is what resolves individual nanoparticles, their internal structure, and at high resolution the lattice fringes of the crystal planes themselves. It is the closest thing to actually seeing the particle. It is also demanding: the sample must be very thin, the preparation is fiddly, and the instrument is expensive to buy and to run.

Both are usually paired with energy-dispersive X-ray spectroscopy (EDS), which reads the characteristic X-rays the beam kicks out and tells you which elements are present in the spot you are looking at — the answer to "is that dark speck my material or a piece of catalyst residue?"

What the microscope quietly does to your sample

Electron microscopy is the most trusted technique in this field and it has three limitations that matter enormously in practice.

It works in vacuum. Liquids boil away, so a dispersion cannot be imaged as a dispersion. You dry a droplet onto a grid — and drying is not a neutral act. As the liquid recedes, capillary forces drag particles together and deposit them in clusters and coffee-ring edges. Clumps in a TEM image may be real agglomerates from the bottle, or they may be artefacts the drying created a minute earlier, and telling those apart takes care rather than a glance. Cryogenic TEM, which freezes the specimen vitreously instead of drying it, exists precisely to avoid this and is correspondingly more work.

The beam changes things. Electrons deposit energy. Beam-sensitive materials — polymers, organics, some hydrated oxides — can shrink, mask, crystallise or ablate while you are watching, which means the thing being photographed is not quite the thing that arrived.

You look at almost nothing. A gram of 40 nm oxide contains on the order of 10¹⁵ particles. A careful session might image a few hundred. If those were selected because they were nicely dispersed and in focus — which is exactly how people choose fields of view — the resulting "size distribution" describes a photogenic minority. A published distribution drawn from fifty particles is an anecdote with error bars.

That combination is why microscopy is definitive about what a particle looks like and unreliable on its own about what is in the bottle.

Why light scattering says 180 when the microscope says 40

Dynamic light scattering (DLS) takes the opposite approach: leave the particles in liquid, and infer their size from how they move.

Suspended particles are jostled by solvent molecules and undergo Brownian motion, and smaller particles jostle faster. Shine a laser through the suspension and the scattered light flickers as particles move relative to each other; the rate of that flicker gives a diffusion coefficient, and the Stokes–Einstein relation converts a diffusion coefficient into a diameter. It takes seconds, needs no vacuum, and measures the material in the state you will actually use it.

Two features of that measurement explain the entire discrepancy in the opening paragraph.

First, what it reports is the hydrodynamic diameter — the size of the sphere that would diffuse the way this object diffuses. That includes the particle, its bound solvation shell, and any surfactant, dispersant or polymer stabiliser attached to it. A 40 nm oxide core with a bound layer and a stabiliser genuinely diffuses like something larger than 40 nm. The microscope, which sees only the dense core against a dried background, does not include any of that.

Second, and much more dramatically, DLS is intensity-weighted, and in the Rayleigh regime scattered intensity scales with roughly the sixth power of diameter. Doubling the diameter scatters sixty-four times as much light. A single 200 nm aggregate therefore contributes as much signal as a million 20 nm particles do. A sample that is 99.9% beautifully dispersed primary particles with a trace of aggregates will report a large average, because the trace dominates what the detector sees.

That sounds like a flaw and is better understood as a specialisation. DLS is superbly sensitive to exactly the thing that usually matters in an application — the presence of a large tail, whether loose agglomerates or permanently fused aggregates — and correspondingly poor at describing the fine fraction. It also struggles to separate two populations of similar size, and its polydispersity index is a rough summary rather than a distribution.

A microscope tells you how big the particles are. Light scattering tells you how big the things moving around in your liquid are. Those are different questions, and the gap between the two answers is not an error — it is the aggregation state, which is often the number you actually needed.

The other numbers on the datasheet, and what they really mean

BET surface area measures how much nitrogen adsorbs onto the powder at low temperature, and converts that to square metres per gram. Divide it into the material's density and you get an equivalent spherical diameter. It is cheap, robust and directly comparable between suppliers — but the conversion assumes smooth, non-porous, separate spheres. Give it a porous or rough material and the enormous internal surface produces an "equivalent size" far smaller than any particle in the sample. As a comparison between grades it is excellent; as a literal diameter it should be read with the assumption in mind.

The Scherrer equation applied to X-ray diffraction peak broadening reports something subtly but importantly different: crystallite size, meaning the size of a coherently diffracting domain. A single particle can be made of many crystallites. So the Scherrer number is a lower bound on particle size, and when it comes out well below the TEM value the honest conclusion is usually that the particles are polycrystalline, not that somebody made a mistake. Strain and instrument broadening also inflate the apparent width, which is why careful work separates them.

Nanoparticle tracking analysis films individual particles' Brownian motion and sizes each one, giving a number-weighted distribution and a concentration — much better than DLS at resolving mixed populations, at the cost of a narrower working range. Centrifugal sedimentation separates by settling speed and has the best resolution of the lot for multimodal samples. And zeta potential, though not a size measurement at all, predicts whether the dispersion will stay dispersed — a magnitude above roughly 30 mV in either direction indicates enough mutual repulsion to resist aggregation, which is the thing that decides whether today's DLS result will still hold next month.

The rule this all adds up to

A size without a technique is not a specification. "40 nm" means one thing from TEM, another from DLS, another from BET and another from Scherrer, and a supplier or a paper that does not say which one is quoting is not giving you information you can act on.

For anyone buying or reporting on nanomaterials, three habits cover most of it. Ask which technique produced the number, and on what sample preparation. Ask for at least two techniques that fail differently — microscopy plus light scattering is the standard pairing, because one is blind to the tail and the other is dominated by it. And treat a large gap between them as data rather than as a problem, because that gap is telling you about aggregation, about the coating, or about the drying, and all three are things worth knowing.

Why it matters for students and researchers

Characterisation is where a great deal of otherwise good nanomaterials research quietly goes wrong, and almost always in the same way: a number is reported without the method, the weighting or the sample preparation that produced it. A z-average from DLS presented as "the particle size" is probably the single most common such slip in the literature, and it is not a small one, because the sixth-power weighting means that number can be dominated by a population that is a vanishing fraction of the sample by count.

The deeper lesson generalises past this field. Every measurement is a model plus a signal. Stokes–Einstein assumes hard spheres in a continuum; BET assumes a particular adsorption behaviour and non-porous geometry; Scherrer assumes strain-free crystallites. When results disagree, the productive first question is not which instrument is broken but which assumption this sample violates — and that question usually teaches you something real about the material.

The open problems are worth knowing about too. In-situ and liquid-cell electron microscopy, which images particles in liquid rather than dried, is advancing quickly and is changing what is known about how particles nucleate and aggregate. Reference materials and interlaboratory comparisons for nanoparticle sizing remain thinner than the field's reliance on these numbers would suggest. And automated image analysis promises to replace the hand-counted fifty-particle histogram with something statistically defensible, provided the sampling problem underneath it is taken seriously rather than automated away.

Frequently asked questions

Why can't I just use a very good optical microscope?

Because the limit is the wavelength of light, not the quality of the optics. Two features closer than about half a wavelength blur into one no matter how good the lens, so visible light bottoms out near 200 nm. Increasing magnification beyond that enlarges the blur without adding detail. Getting below it requires either a much shorter wavelength, which is what electron microscopy does, or a trick that locates individual fluorescent emitters one at a time, which works for labelled biological samples and not for a powder.

Which size number should I put in my report?

All the ones you measured, each labelled with its technique — and if you must lead with one, lead with the one that matches your application. If the material will be used as a dry powder or a coating, the microscopy and BET numbers describe it. If it will be used dispersed in a liquid, the hydrodynamic diameter is what your process will actually experience, coating and all. Reporting a single unlabelled number is what makes work impossible to compare.

My DLS reading changes every time I measure. Is the instrument faulty?

Usually not. DLS is extremely sensitive to large objects, so a stray dust particle, an air bubble, a fingerprint on the cuvette or incomplete redispersion can move the result substantially. Drifting steadily upward across repeats normally means the sample is genuinely aggregating in the cell while you watch, which is information rather than noise — and a zeta potential measurement will usually tell you whether that was to be expected.

Are the particles in a TEM image really clumped, or did drying do that?

Both are common and you cannot tell from one image. Comparing several preparations at different dilutions helps, because drying artefacts get worse as concentration rises while real agglomerates do not. Comparing against a light scattering result on the wet dispersion helps more. Cryogenic TEM, which vitrifies the liquid instead of evaporating it, settles the question properly and is the right method when the answer matters.

What is the difference between crystallite size and particle size?

A crystallite is a region of continuous, coherently ordered lattice; a particle is the physical object, which may contain many crystallites separated by grain boundaries. X-ray line broadening reports the crystallite; microscopy reports the particle. For a single-crystal nanoparticle the two coincide, and for a polycrystalline one the X-ray number is smaller — legitimately, not erroneously.