There is not a "best" method. The characterisation depends on the type of material, the size, the size distribution, shape, stability on so on. Electron microscopy (both TEM and SEM) is generally a good way to get an idea of most of these parameters but one needs to take into consideration that the sampling volume is very small. What you see may not be representative for the batch of particles prepared. Also, AFM and STM can be used but the problem of sample size is even more severe and also tip convolution can lead to overestimating the size significantly. For monodisperse particles in aqueous dispersion, DLS is convenient, fast and generally reliable but it is very sensitive to "contaminations" with dust or larger particles. Also, it will give an average based on particle mass and not on number. It is thus possible, that only few particles are actually of the size the instrument gives you. For really monodisperse preparations this is not an issue since mass and number average will be the same. In our lab, we have used analytical centrifugation techniques with good success. They separate polydisperse preparations to some extent and give good estimates for size distributions. These technique go down to about 5 nm (for gold particles). Smaller ones take too long to sediment and it becomes impractical. It can be extremely sensitive to tiny changes in particle size, for example by chemical surface modification. I believe it is perhaps the single most powerful technique to monitor such subtle changes. For dry powders, if you want to estimate the size of crystalline domains, XRD line broadening analysis can be used applying the Scherrer equation. This often underestimates significantly the particle size if the sample is poorly crystalline. There are several other methods that can be very good for certain samples. One method traded as "Nano Sight" optically tracks individual particles using light scattering and then calculates the diffusion coefficient. This is quite attractive since the instrument images the particles and you can "see" them on the screen but the method is really conceptually equivalent to DLS and has the same limitations. Finally, a very cheap and often quite convenient way is simply optical spectroscopy. Semiconductor quantum dots have a characteristic size dependent colour, and also gold nanoparticles between about 3 and 30 nm can be estimated quite well simply from the appearance of the optical spectra. There is a paper by Haiss W. et al. in Analytical Chemistry that gives a simple formula to do this.

Light scattering (laser diffraction and DLS) are the most common techniques but others find use too depending on the width of the distribution. Visualization with some form of electron microscopy is essential.

For nano you can use Light Scattering (DLS), the range is 3-1000 nm, it´s good for measurements arround 5-400 nm.

For micro, I used a Coulter (laser diffraction) with a small volumen module. From 50 nm to 5 micras.

As others mentioned, DLS for particles under 1 micron but a coulter counter for particles above 1 micron.

There is not a "best" method. The characterisation depends on the type of material, the size, the size distribution, shape, stability on so on. Electron microscopy (both TEM and SEM) is generally a good way to get an idea of most of these parameters but one needs to take into consideration that the sampling volume is very small. What you see may not be representative for the batch of particles prepared. Also, AFM and STM can be used but the problem of sample size is even more severe and also tip convolution can lead to overestimating the size significantly. For monodisperse particles in aqueous dispersion, DLS is convenient, fast and generally reliable but it is very sensitive to "contaminations" with dust or larger particles. Also, it will give an average based on particle mass and not on number. It is thus possible, that only few particles are actually of the size the instrument gives you. For really monodisperse preparations this is not an issue since mass and number average will be the same. In our lab, we have used analytical centrifugation techniques with good success. They separate polydisperse preparations to some extent and give good estimates for size distributions. These technique go down to about 5 nm (for gold particles). Smaller ones take too long to sediment and it becomes impractical. It can be extremely sensitive to tiny changes in particle size, for example by chemical surface modification. I believe it is perhaps the single most powerful technique to monitor such subtle changes. For dry powders, if you want to estimate the size of crystalline domains, XRD line broadening analysis can be used applying the Scherrer equation. This often underestimates significantly the particle size if the sample is poorly crystalline. There are several other methods that can be very good for certain samples. One method traded as "Nano Sight" optically tracks individual particles using light scattering and then calculates the diffusion coefficient. This is quite attractive since the instrument images the particles and you can "see" them on the screen but the method is really conceptually equivalent to DLS and has the same limitations. Finally, a very cheap and often quite convenient way is simply optical spectroscopy. Semiconductor quantum dots have a characteristic size dependent colour, and also gold nanoparticles between about 3 and 30 nm can be estimated quite well simply from the appearance of the optical spectra. There is a paper by Haiss W. et al. in Analytical Chemistry that gives a simple formula to do this.

Thanks for the detail answer Mathias Brust

The best techniques to be used for particle size distribution analysis for nano- to micro particles are:

1/ TEM and SEM spectroscopy

2/ DLS or laser granulometry