Dear Son,

the 1st  coordination shell  of the above mentioned oxides consits of O atoms. Their backscatter amplitude is a strongly decreasing function of k.

In contrast to that  the 1st shell of the metals are the metal atoms themselves. For high atomic number  scatterers (such as Pt or Pd ) the backscatter amplitude extends over a quite large k-range having significant scatter strength.

So I fear that for the oxide case  the first peak  in the radial distribution function of EXAFS might be too small in order to determine  a coordination number precisely enough to have a reliable  estimate of the particle size. I think this also holds when EXAFS evaluation is done by fitting as you mentioned above.

But it is only a guess.

May be other colleagues comment on that more positiv.

P.S. You might request full-text of the attached paper at G. Tolkiehn or me.

Chapter Experimental Investigation of the Systematic k- and Z-Depend...

Dear Son,

the 1st  coordination shell  of the above mentioned oxides consits of O atoms. Their backscatter amplitude is a strongly decreasing function of k.

In contrast to that  the 1st shell of the metals are the metal atoms themselves. For high atomic number  scatterers (such as Pt or Pd ) the backscatter amplitude extends over a quite large k-range having significant scatter strength.

So I fear that for the oxide case  the first peak  in the radial distribution function of EXAFS might be too small in order to determine  a coordination number precisely enough to have a reliable  estimate of the particle size. I think this also holds when EXAFS evaluation is done by fitting as you mentioned above.

But it is only a guess.

May be other colleagues comment on that more positiv.

P.S. You might request full-text of the attached paper at G. Tolkiehn or me.

Chapter Experimental Investigation of the Systematic k- and Z-Depend...

Dear Gerhard,

Thank you for your useful feedback. However, what I observed with my PdO (supposed to be nanoparticles) supported on H-ZSM5 samples is that the first peak in the chi(R) function is not small at all. Please take a look at the attached picture. The first peak between 1-2 Angstrom is due to the Pd-O path and it is significant. The fitting result also gives a coordination number of 4.72+/- 0.32 so it can be determined with good enough precision. The point is I have a hard time finding literature that correlate that coordination number to particle size of an oxide while quite a lot have done that for a metal and I have no idea why oxide can not be done the same way.

Dear Artur,

I am aware of the limitation of the method of determining the particle size using EXAFS (i.e. the particle should definitely be nanoparticles and the size should be smaller than 3 nm to be used reliably). However, let's assume that my particle size is in the application limit of the method. The problem is I can not find a paper which studies the correlation between CN and particle size of oxide nanoparticles while there are quite some studies for metal nanoparticles and I don't understand why.

Dear Son,

if you can trust your fitted coordination number there is no reason why Oxygen as first coordination shell scatterer cannot be used to evaluate nano particle size compared to other  'high z' atoms.

I havn't seen yet seen any paper(s) dealing with that subject on oxides.

However, I think in general it is brave to use mean coordination number for particle size estimation. These  particles have to have a maximum size of a very small number of crystallographic cells. You mentioned 'smaller than 3nm' already.

Dear Marten,

I wonder for my oxides instead of using the first Pd-O coordination number, is it a good idea to use the first Pd-Pd coordination number instead? However, as I already stated I could not find any proven correlation in literature re. this. For the time being, I am thinking of using the first Pd-Pd coordination number as a "qualitative" index of the particle size (i.e. smaller coordination number means smaller particle size) and use this index to compare the change in Pd's particle size between different samples of mine.

It might work.

But let me point to another aspect.

The coordination number can be extracted from EXAFS if a) sample preparation , b) EXAFS measurement, c) data pre-processing and d) evaluation of coordination number is done correctly.

But ..... is the mean coordination number really a measure of the particle size ?

For  samples consisting of one atom species one would say : Yes: the outer atoms sitting at the surface only see a fraction of the first and 2nd coordination sphere due to the precence of the particle surface and thus the mean coordination number for first and even second shell is reduced.

But what is about particles containing of two different types of atoms, such as in your case PdO and PtO2?

Do these type of particles have an atom  distribution at its surface which is according to its stoichiometry? e.g. for PdO will have Pd:O = 1:1 or for PtO2 will have Pt:O = 1:2?  .... If so, in this case I think one can correlate mean coordination number to particle size.

But is this really the case from the crystallographic/chemical bond point of view?

There may be two extreme cases:

a) only O atoms are sitting at the surface of the nanoparticle or

b) only Pt or Pd atoms are at the surface of the NPs.

In these cases you may also relate the mean coordination number to the particle size, but you have to take into account the relative atom distribution at the surface.

 I think for oxides one does not really know the atom distribution at the particle surface; so it will make no sense to take the mean EXAFS coordination number from the Oxygen 1st shell.

For the case of the second shell, as you asked above, I think the situation could be  relaxed, even for a non stoichiometric atom distribution at the particle surface. But at the moment it is not clear to me whether  the second shell ( Pt-Pt or Pd-Pd) will really solve your challenge.

May be other colleagues will comment on this.

Dear Gerhard,

Excellent insight !!! Thank you for that. I think this is because of this not-known (and perhaps uneven?) distribution of Metal and Oxygen atoms on the surface which leads to the lack of literature works on this. Because you don't know the change in mean coordination number from one sample to another is caused by the change in atom distribution on the surface or the change in particle size. I am waiting for comments from other colleagues to see if the "qualitative" approach that I mentioned earlier is applicable.

I came on this idea while harvesting blue-berries and black-berries into one pot this afternoon.

During harvesting the relative berry distribution at the surface changes all the time.

Additional remark:

The extreme situations a) only O-atoms at the surface or b) only metal atoms at the particle surface will be very unlikely. The real situation will be close to the stoichiometric distribution due to the minimizing principle of the surface energy while forming the NPs; but I do not know how close.

But in addition Oxygen from the air may adhere to the particle surface. These  bond lengths may be somewhat larger than that of the 1st Oxygen bulk shell but I assume some contribution to the 1st Oxygen peak via slight increase in  peak height and increase in peak width.

Late, but nevertheless additional informations with respect to particle size and coordination number(s):

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