DLS, NTA or TEM: Which Particle Sizing Method Should You Trust?

Three techniques, three numbers, one sample

Send the same nanoparticle dispersion for three measurements and ask each for a particle size. It is entirely ordinary for dynamic light scattering to come back with the largest number, nanoparticle tracking analysis with a smaller one, and transmission electron microscopy with the smallest of the three. Nobody has made a mistake. The three instruments measured three different physical quantities and weighted them three different ways, and the pattern of the spread between them is itself a measurement — of coating thickness, of solvation, and of how much aggregate is in the vial.

The practical problem is that specifications, certificates and purchase decisions treat "particle size" as one number. This guide is about which number you actually want.

What each technique measures

Dynamic light scattering

DLS illuminates a suspension with a laser and records the fluctuation of scattered intensity over time. Those fluctuations are caused by Brownian motion; the rate at which the intensity autocorrelation function decays gives the translational diffusion coefficient, and the Stokes-Einstein relation converts diffusion coefficient to a hydrodynamic diameter — the diameter of a hard sphere that would diffuse at the same rate.

Three consequences follow. The result includes anything that moves with the particle: adsorbed ligands, polymer coatings, the electrical double layer, the solvation shell. It is a property of the particle in that medium and changes with ionic strength. And it is a bulk ensemble measurement over the whole illuminated volume, which is what makes it fast and what makes it blunt.

Nanoparticle tracking analysis

NTA — also called particle tracking analysis — also derives a hydrodynamic diameter from Brownian motion and the Stokes-Einstein relation, but it does so particle by particle. Individual scattering centres are visualised in a thin illuminated sheet and their trajectories tracked frame by frame. Each tracked particle yields its own diffusion coefficient and therefore its own size.

That changes the weighting completely. NTA produces a number-weighted distribution directly, and it also produces a particle concentration, which DLS cannot. The cost is throughput and range: only particles that scatter enough to be tracked are counted, so the low end depends on the material's refractive index — gold is visible far smaller than a polymer — and only a few hundred to a few thousand particles are typically tracked per measurement.

Electron microscopy

TEM and SEM give a projected geometric size of individual, dried particles, measured directly from images. No hydrodynamic assumption, no refractive index, no model — you are looking at the object. The distribution is number-weighted because you counted particles.

The costs are equally direct. The sample is dry and in vacuum, so the coating that DLS sees may have collapsed or become invisible. Drying onto a grid can drive particles together, so aggregation seen in a micrograph is not necessarily aggregation in the vial. The field of view is tiny, so the statistics come from how many particles you were willing to count. And a low-contrast organic shell around a metal core may simply not appear.

The sixth-power problem

This is the single most important thing to understand about DLS, and it explains most of the disagreements above.

For particles much smaller than the wavelength of light, scattered intensity scales approximately with the sixth power of diameter. Two particles differing by a factor of ten in size differ by a factor of a million in scattered intensity. DLS reports an intensity-weighted distribution by default, so a vanishingly small number of aggregates dominates the signal.

Work it through. Suppose one particle in a thousand has formed a ten-fold aggregate. By number it is 0.1% of the sample. By scattered intensity it contributes on the order of a million times as much as a single particle, so it contributes roughly a thousand times the total intensity of the remaining nine hundred and ninety-nine — it dominates the measurement entirely. This is why a clean-looking dispersion can return a DLS mean far above the TEM mean, and why the honest conclusion is "there are a few aggregates" rather than "the particles are bigger than we thought".

DLS software will convert intensity distributions to volume and number distributions. Those conversions require the material's refractive index and absorption, assume spherical particles, and amplify noise at the small end — dividing by the sixth power of a poorly determined diameter. Treat converted number distributions from DLS as indicative and never as a substitute for a counting technique.

Side by side

Question	DLS	NTA	TEM / SEM
Quantity measured	Hydrodynamic diameter	Hydrodynamic diameter	Projected geometric size
Native weighting	Intensity	Number	Number
State of sample	In suspension	In suspension	Dried, under vacuum
Sees coatings and solvation	Yes	Yes	Only if it has contrast
Sees shape	No	No	Yes
Gives concentration	No	Yes	No
Sensitivity to a few aggregates	Very high	Moderate	Low
Resolving a bimodal population	Poor unless well separated	Good	Good
Time per sample	Minutes	Tens of minutes	Hours to days
Material needed	Small volume, dilute	Small volume, dilute	Very little
Standard	ISO 22412	ISO 19430	Counting practice varies

Which one answers your question

• "Has my batch aggregated since last week?" DLS, without hesitation. Its sensitivity to a small aggregate population is a defect when you want a size and an asset when you want a stability alarm. Repeat measurements over time on the same instrument are the strongest routine stability test available.
• "What is the true mean diameter of my cores?" Electron microscopy, counting enough particles — several hundred at minimum — from several areas of the grid.
• "How thick is the polymer shell?" The difference between the hydrodynamic diameter and the electron-microscopy core diameter, which is one of the few places where using two techniques together gives you a quantity neither gives alone.
• "How many particles per millilitre?" NTA, or single-particle ICP-MS. DLS cannot answer this and any software that appears to is inferring it from assumptions.
• "Is my sample bimodal?" NTA or microscopy. DLS resolves two populations only when they are separated by a wide factor, and even then the intensity weighting distorts the relative amounts badly.
• "Are my particles rods, not spheres?" Microscopy. Both scattering methods return an equivalent sphere and will happily give you one for a rod.
• "Will this be stable in buffer?" DLS as a function of ionic strength, alongside zeta potential — see what the zeta number means and when it lies.

Reconciling a disagreement

When the three numbers disagree, the pattern of disagreement is diagnostic:

• DLS high, NTA close to TEM. A small aggregate population, weighted up by intensity. Expected, and usually benign, but worth watching over time.
• DLS and NTA both well above TEM, by a similar margin. A real hydrodynamic layer — coating, ligand shell, or extended double layer at low ionic strength. Check whether the margin shrinks as salt is added.
• DLS below TEM. Unusual, and generally means the TEM count was biased by drying-induced clustering being measured as single objects, or that only the larger particles were counted because the small ones were hard to see.
• Wide DLS polydispersity index with narrow TEM histogram. The particles are fine; the suspension is not. Look at dispersion protocol and storage.

The polydispersity index from a cumulants analysis is the standard DLS width statistic. As a rough guide, values below about 0.1 indicate a narrow, well-behaved distribution, values through the middle of the range indicate moderate polydispersity, and high values mean the cumulants fit itself is questionable and the distribution should not be reported as a single mean at all. ISO 22412 sets out the cumulants procedure and what may be reported from it.

Getting the measurement right

• Filter the dispersant, not the sample. One dust particle is an aggregate as far as DLS is concerned.
• Dilute into the same medium. Diluting into pure water changes the ionic strength and therefore the double layer, the aggregation state, and the answer.
• Report the medium, pH and concentration with every hydrodynamic number. Without them it cannot be reproduced.
• Watch concentration. Too dilute and NTA has nothing to track; too concentrated and DLS suffers multiple scattering and particle interaction, both of which bias the result.
• Count enough particles for microscopy, from more than one region, and publish the histogram rather than the mean alone.
• Let the sample equilibrate thermally. The Stokes-Einstein relation contains viscosity, viscosity depends strongly on temperature, and an unequilibrated cell drifts.

What to put in a specification

Name the technique, the medium and the weighting, and give a distribution rather than a mean. "Hydrodynamic diameter 78 nm, intensity-weighted, DLS per ISO 22412, in 10 mM phosphate buffer pH 7.4, PDI 0.09" is a specification. "Size: 78 nm" is a number waiting to be misunderstood — and it is the line that most often goes wrong on a certificate of analysis, which the CoA guide covers in full.

Related guides

• Where sizing sits among the other measurements: the characterisation workflow.
• The long-form DLS and zeta guide: dynamic light scattering and zeta potential.
• Choosing between the two microscopes: TEM vs SEM.