@ Saurabh While one can process images (and processing images is NOT the same as generating a reliable and robust particle size distribution) with many programs, including Image J, you have to take into account the following:

• Only a tiny subset (ng) of a system may be seen. The sample needs to be representative and it's easy to calculate a minimum mass
• 10000 particles are needed to calculate the mean to 1% standard error. A distribution implies much more than the mean. So to get a distribution to 1% SE at the x95 or x99 point in the distribution requires 10000 particles in the x95+ or x99+ part of the distribution. I have never seen this number of particle images ever
• In the case of TEM only a thin section is observed - often by microtoming. Thus everything is an artifact. Consider slicing a monodisperse spherical system. One will end with a set of discs with apparent distribution when none was present in the original sample. Consider slicing a cube - all sorts of shapes are possible from triangles to squares to even hexagons

That's why ensemble methods are better for distributions if a statistically significant number of particles are not imaged - which is invariably the case. However, imaging is essential to get qualitative information about the form and nature of the particulate system.

Hello Ranjith I do not reply to PM's other than on the public RG forum. That allows more experts to contribute to your question. Several points:

• DLS provides an intensity-based ensemble distribution based on billions of particles. It is an excellent screening technique and provides great sensitivity to the bigger particles (agglomerates, aggregates) in the system as there is a r^6 dependence on the scattering intensity
• Microscopy based techniques: visualization is essential in order to understand the form (shape) of the particles, the crystalline or amorphous state, the presence of agglomerates and aggregates etc. These are qualitative, not quantitative assessments. Note that distributions are difficult as the number of particles sampled is generally pretty low. To get the standard error of the mean (SEM) to 1% requires 10000 random particles to be measured. To get the x90 to 1% SE requires 10000 particles in the x90+ part of the distribution. Further, one needs to have separated particles (not by software!) in order to deconvoluting aggregates and agglomerates.
• TEM - everything measured by TEM is an artifact of the sample preparation and the selection of particles
• SEM - see the comments with respect to numbers of particles

Note too that it's the soluble form of silver (Ag+ ions) that's toxic to bacteria not silver metal (Ag0). Note too that on the surface (10 - 15 atomic layers) of any silver containing species that the Ag is in the fully oxidized (+1) state and does not exist as Ag metal. See, for example:

September 2nd, 2015 Silver colloids and invisible ink

https://www.brainshark.com/malvern/vu?pi=560497681&b=1&tx=120486&c1=22

Thank you Dr.Alan F Rawle...I guess you did not get my point. As I already know the silver NPs toxicity to bacteria, I would like to know about the characterization of AgNPs especially by SEM or TEM. Is particle size distribution necessary for the identification of NPs size by TEM?

Use DLS as a scanning technique. Use SEM for the things talked about in my first posting above. I do not understand this sentence: ' Is particle size distribution necessary for the identification of NPs size by TEM?'....

I mean to say inaddition to DLS, is it necessary to show particle size distribution rather than showing only particle size that is measured by TEM

TEM doesn't provide size distribution information.

Dear sir,

One can use ImageJ software for particle distribution analysis from TEM and SEM images.

This video link can help you in better understanding.

https://www.youtube.com/watch?v=JQAZ--c9YfI&t=120s

Thank you Dr.Alan and saurabh

@ Saurabh While one can process images (and processing images is NOT the same as generating a reliable and robust particle size distribution) with many programs, including Image J, you have to take into account the following:

• Only a tiny subset (ng) of a system may be seen. The sample needs to be representative and it's easy to calculate a minimum mass
• 10000 particles are needed to calculate the mean to 1% standard error. A distribution implies much more than the mean. So to get a distribution to 1% SE at the x95 or x99 point in the distribution requires 10000 particles in the x95+ or x99+ part of the distribution. I have never seen this number of particle images ever
• In the case of TEM only a thin section is observed - often by microtoming. Thus everything is an artifact. Consider slicing a monodisperse spherical system. One will end with a set of discs with apparent distribution when none was present in the original sample. Consider slicing a cube - all sorts of shapes are possible from triangles to squares to even hexagons

That's why ensemble methods are better for distributions if a statistically significant number of particles are not imaged - which is invariably the case. However, imaging is essential to get qualitative information about the form and nature of the particulate system.

TEM/SEM and DLS are complementary to each other and in my opinion both should be shown in publications. TEM/SEM can give a good estimate of the size distribution when a good number of particles are carefully analyzed by software such as ImageJ Fiji (please consider the criteria mentioned by Alan F Rawle ). It is important to realize that DLS gives the hydrodynamic size distribution, which is the result of an indirect measurement of the motion of the particles. Therefore it's also influenced by the amount of friction the particles experience as they are moving. Certain ligands can influence this friction (for instance, PEG in water forms a lot of hydrogen bridges, leading to lots of friction). Other factors influencing DLS strongly are aggregation, impurities, solvent, concentration, etc, which should be taken into account! In the end you will likely get different numbers from TEM/SEM and DLS, but in my opinion this gives a more complete picture.

Thank you Dr. Sven Askes for the detailed information.

@ Sven Askes I totally agree with your comment that the techniques (microscopy and DLS) are complementary. Visualization is essential as I stated in my last post. However, we need to distinguish between qualitative and quantitative examination of our system. At all times we must be asking ourselves 'Is the image or the result representative of the whole?'. In many cases we are taking a specimen (like grab sampling in mining) and not a sample.

As you indicate (and I concur) we should take care with number and volume/mass and intensity distributions.

I still stand by my comment that everything under TEM is an artifact of sample preparation and particle selection. For scanning a larger area and easier sample preparation, I tend to prefer SEM. TEM I'd only use for looking at atomic spacings and interfaces between surfaces (such as in the semiconductor industry).

@Alan F Rawle: I agree with you sir. Every system has certain limitation and user must have taken that limitation into account.. Guess we might need panaroma feature in TEM and SEM techniques for large area image capturing.

Thank you for enlightening me in this topic.

https://www.researchgate.net/post/how_to_calculate_particle_size_from_TEM_images

Thank you Dr. Gurunathan