Please find below a review on the MW measurement of polysaccharides:

http://www.nottingham.ac.uk/ncmh/documents/papers/paper82.pdf

I hope it will be useful

I was always skeptical about these claims of measuring MW of polymers by light scattering techniques. Both static and dynamic scattering essentially probe the size of a region in space where the polymer presence makes the refractive index differ from pure solvent. There are two possibilities - the solvent is good and the polymer chain is in an expanded random coil shape - or the solvent is poor and the chain is collapsed into a globule (precipitates). 

In the first case, the concentration in solution has to be extremely low (at least an order of magnitude lower than the overlap concentration) to ensure the chain is single. Is this always observed, how would people even know what the overlap concentration should be not yet knowing MW? Then, accepting that this was achieved, we have a spherical volume of the radius of gyration size (N^3/5 for flexible chain, etc) in which the relative volume of polymer is very low (scales as N^-4/5, so e.g. for a chain of N=1000 monomers this is a fraction =4%). How big is the difference in refractive index in such a volume with polymer, compared with the solvent without polymer? Are the errors in light scattering experiment seriously lower than single-% ??  It is always easy to obtain *some* signal, but interpreting it correctly is a real challenge, which I don't see addressed in any of these 'practical guides'.

What about precipitated polymer globules? They need to be floating in solution, not sedimenting as most colloid particles would (hence the word 'precipitation'). May be this is achieved by cleverly adjusting buoyancy of solvent, and then the DLS signal will show the correlation in diffusing colloid particles, where you will extract the size of globule from the Stokes diffusion constant. You must ensure that the polymer globules are made of single chains, because (in poor solvent!) any two chains will of course precipitate together, so back to that very low concentration I mentioned above. Interestingly, because of the high error (arising from inaccuracy of the Stokes drag on this small scale) it would be really hard to distinguish the difference between single chain and 2-3 chains globule. 

Finally, please be aware that all commercial "particle sizers" have the software forcing the fit of the actual DLS signal to a single Gaussian distribution peak. It's amazing, because you will always obtain "a size" in this way, even if the actual signal is bimodal, skewed, or  in some other way reflecting the complex nature of a sample. 

I believe that skepticism is reasonable and is in place in scientific research. Nevertheless MW is and has been measured by a variety of techniques, light scattering is being only one of them. Probably all of them contain questionable assumptions, but they can be tested by each other, therefore they are more or less accepted by the scientific community. Light scattering can also be combined with size exclusion chromatography and other sources of problem, as branching can be assessed by independent methods, therefore a series of cross-check possibilities are present.

I think that GPC technique using adecuate reference substances  such as pullulan standars or GPC coupled to LS detector could be useful technique to determine alginate MW.

I am agree with the comemts of Dr. Eugene, but for obtaining the Mn by DLS analysis you have to determinate the dn/dc values of the evry polymer (form slope of n vs c). Another important alternative is vapour pressure osmometry (It is not dependent of morfology of the polymers) but it is applied to polymers of low Mn (usually < 100.000 ). The method more used is GPC.

Comment to prof. Terentjev's post:

I understand your scepticism, however, all the problems you have mentioned can be overcome. But let me comment point-by-point.

>There are two possibilities - the solvent is good and the polymer chain is in an expanded random coil shape - or the solvent is poor and the chain is collapsed into a globule (precipitates).

Three possibilities, actually, theta region is also there.

>In the first case, the concentration in solution has to be extremely low (at least an order of magnitude lower than the overlap concentration) to ensure the chain is single.

These are usual experimentally achievable polymer solution concentrations. In the first case (good solvent regime) polymers are always single, in other words, polymer solution is a molecular dispersion.

>Is this always observed, how would people even know what the overlap concentration should be not yet knowing MW? 

One does not need to know the overlap concentration. What you have to know, is whether the solution is dilute or not, or, in other words, whether one can subscribe the measured signal to the changes occurred at the level of a single polymer. It can be easily checked experimentally. Dilute the solution twice, measure the optical signal, if it has decreased twice, means you have single molecular (true) solution, or diluted solution regime.

> Are the errors in light scattering experiment seriously lower than single-% ?

Well, there are usually 100 kcps measured. Means, 100 kilo counts per second, with usual measurement length being 3-5 mins, at least 5 such shots averaged, which is 5*3*60*10^5~10^8 measured polymer conformations. Relative error is as usual, $\sqrt(10^8)$^(-1), which is 0.0001. More than enough for any practical applications.

>It is always easy to obtain *some* signal, but interpreting it correctly is a real challenge, which I don't see addressed in any of these 'practical guides'.

I am missing your point here. So, we take a cuvette, put there no polymer, measure DLS signal, no light scattering detected. In the same solvent, we add a tiny amount of polymer, measure, and see scattering. We put this cuvette aside, come tomorrow, measure, same result. We prepare the similar solution several times, measure quantitatively same degree of scattering. What would you attribute the effect to then?

>What about precipitated polymer globules?

Here is our recent paper solving these problems: DOI: 10.1038/srep35450

Flory-Huggins miscibility diagram gives the hint. Increase temperature to go above the border of two-phase region, mix the solution in ultrasound mixer to break agglomerates, measure in DLS at same high T. We have managed to do it for probably the most difficult class of polymers, polysilanes. They have been discovered in 1924, but since almost do not dissolve, are very poorly studied. We have overcome the problem by simply heating the dilute solution to 50 degrees C. DLS actually allowed to detect the temperature value, after which the agglomerates break and do not change in size; also the distribution become very monodisperse. Take a look to the paper, I enjoyed writing it.

> Finally, please be aware that all commercial "particle sizers" have the software forcing the fit of the actual DLS signal to a single Gaussian distribution peak. It's amazing, because you will always obtain "a size" in this way, even if the actual signal is bimodal, skewed, or in some other way reflecting the complex nature of a sample.

I still believe, that the most important tool of the scientist is it's head. No instrument or program can substitute it. There are usual polymer chemistry methods to separate polymers onto fractions. Then the monodisperse sample can be studied.

I have found your post extremely interesting, since you are asking exactly the same questions I had to ask myself four years ago, when started the collaboration with my experimentalist colleagues. It is all doable, no magic.

Thanks!

Artem

Many thanks to all for this valuable information.

You need to know the chemical composition of the polymer and its structure (linear, branched, network...). Need to have a dilute solution and to know the regime of a polymer solution (at given temperature, solvent content and concentration, is polymer in a coil or globular state). If known all the above, DLS will give you the distribution of particle sizes. If the solution is monodisperse, the distribution of sizes is narrow, and one can take the average as the characteristic size. This size is proportional to the power of molecular weight. See, e.g., Soft Matter, 2018, 14, 4735 and Scientific Reports | 6:35450 | DOI: 10.1038/srep35450. Good luck!