Jorik Waeterschoot

The z-average and PDI are defined in, and a direct consequence of using, the cumulants analysis (CA). They come from the formula used. Now the CA only uses the first 90% of the correlogram so I suspect that you have far field instability meaning large and maybe sedimenting particles. And consequently a high PDI (1.000). Can you show your report and the associated error report that I suspect you may have. So, if you have big material is it a consequence of your sample preparation and debris present? How are your samples diluted? Should you be filtering at 20 nm (0.02 um?). Or if you're expecting you particles to be ~ 1000 nm, then I'd not be using DLS but laser diffraction where you have a model-independent analysis fitting of the multi-angle data. Here your width would be defined by the span - (x90 - x10)/x50. And of course there are a number of spans - another one is (x84 - x16)/x50. I'm sure you'll know why some prefer the latter...

If you convert your intensity result to volume do you get the main peak at ~ 150 nm?

There must be a reason for the "wrong" average size that you get with the Cumulant method. As pointed out by Alan F Rawle, you are probably having problems with dirt. You could try to use more aggressive chemicals to clean your cuvette, then use a lot of purified water (e.g. MilliQ) to flush the cuvette. Fill the sample into the cuvette in a laminar flow box to avoid dust contamination. There can (and will) still be dust in your sample. You can point a laser beam through the cuvette and look carefully if you can spot little blinking points in the sample. As a last resort, you may put the cuvette in a centrifuge before putting it in the DLS. Or let it sediment for a while, before starting the measurement.

To your question: When you use a stretched exponential, you can estimate the Polydispersity via the second moment of the distribution of relaxation times. It is determined by the stretching exponent "beta" and by the relaxation time t_k (see eg https://en.wikipedia.org/wiki/Stretched_exponential_function or http://www.moreisdifferent.com/wp-content/uploads/2015/07/stretched.pdf)

(Note that the average relaxation time is given with the first moment of the distribution, it is not the relaxation time that is in the stretched exponential funktion ( is not t_K). )

Hi Jorik,

Could you share with us your autocorrelation curve and the histogram associated with % intensity %mass, %polydispersity and hydrodyanamic radius for each population.

Also, what kind of size do you expect? Do you use other approches to monitor the size of your lipid particles?

Best,

Sébastien

Sébastien Brûlé in attachment you can find the raw data. I do not have the other histograms since I am not using a commercial software package

Moritz Schloetter I agree that there should be a reason for the wrong size of the cumulant, but is it really a problem if the stretched exponential gives the right fit?

Alan F Rawle I have no report, only the raw data since I am not using a commercial software packet. (you can find the data in attachment). With the stretched exponential I get the result I expect, but not with the cumulant method. The samples are produced by extruding them through a 100nm membrane. They are simply diluted by adding buffer.

Jorik Waeterschoot

OK, then are you using the cumulants method correctly? As per ISO 22412 (and the much earlier, now withdrawn, ISO 13321)?

Jorik Waeterschoot

You state 'They are simply diluted by adding buffer'... Is this identical to the starting continuous phase? Your state your filtration regime is 200 nm. What type of filter? Are the buffers filtered to 20 nm? Again, at these stated size of 1000 nm I wouldn't be using DLS - use the orthogonal laser diffraction. You appear to have a goniometer set up, so use static light scattering to deal with the multi-angle dependence you're seeing.

Alan F Rawle What is the exact differences between these two ISO's.

The sample is diluted with the starting continues phase. I am using a polycabonate filter, a standard setup to make small lipid vesicles.

What do you mean with a these stated size of 1000nm? the lipid vesicles produced have a radius of 100 to 200nm.

Jorik Waeterschoot ISO13321:1990 is withdrawn in favor of the later 22412 standards (2008, 2017). However, many people will be more familiar with 13321 which is why I mentioned it. But both these norms show how to implement cumulants correctly.

A mean or median diameter of the order of 400 nm (2 X 200 nm - your radii above) is huge for DLS. It wouldn't surprise me that you have 1000 nm material in your system. If you have issues fitting then that tells me that your polydispersity must be high and that you don't have a narrow distribution. If you have a narrow distribution then almost any model will provide correct results.

Distributions are the bane of all particle size techniques... But even more so for DLS which is a low resolution technique.

Jorik Waeterschoot

I'm very impressed with your 'home built' system and its apparent ability to be able to provide 4 places of decimals after the nm decimal point. A discrimination level to 0.0001 nm is really impressive!