Obtaining a Representative Sample

Don't ask me why I am suddenly thinking about this when I ought to be studying real analysis. I don't know.

Basically: it's usually easy to obtain a homogeneous sample from liquids and gases, but what about solids? Suspensions?

In analytical lab classes, the unknowns are usually solid. However, we students don't get very much of it (around 1 g), and it tends to be very fine, so we usually just assume it's homogeneous and use it as-is. It's not that hard to give the unknown a good shake before weighing it out, either. Solutions are even better: shake and use. In analytical lecture, the problem of obtaining a representative sample gets, um, maybe one minute exposure time.

That, of course, is what work terms are for. One of my previous workplaces had evolved some methods to obtain representative solid samples that um, deviated a bit from the norm. I'll not describe them in too much detail (in the unlikely event a coworker reads this), but mention the "usual" ways, and skirt the edges a bit.

Classically, there's the "coning and quartering" technique, where the sample gets dumped onto a suitable surface to form a cone (sometimes "rolling" the corners of the paper to get a more homogeneous mixture), then the cone gets flattened and divided into quarters. The opposite quarters are taken, and they are further coned and quartered until the desired sample size is reached. But this has problems. For example, large particles tend to roll to the bottom edges, smaller ones on top, and it can be difficult to take both large chunks and smaller chunks of sample. Some people cone and take sub-samples (not necessarily quarters) in more creative ways (one scoop here, one scoop there...).

More modern techniques involve things like riffle splitters and rotary splitters. They still require somewhat homogeneous particle sizes, though, and finer particles tend to disappear (well ok, fly all over the place). Also, the samples need to be dry.

Question: how to obtain representative damp samples? Tangentially related question: how to separate really large samples from overenthusiastic samplers? For example, 1 (10, 100, ...) kg* soil sample for moisture analysis. No drying first, for obvious reasons. Coning and quartering large samples is painful, and trying to force the stuff through a riffle or rotary splitter is even more painful; the result isn't really "representative" either.

Liquid suspensions are probably easier to get a representative sample from. Thorough shaking doesn't quite do the trick, though. For example, pipetting a series of river water samples after shaking it really well gives decreasingly representative samples as the solid stuff settles down over time. Stir bars are of limited utility, depending on the sample container shape and size.

I bet there are journals out there that cover just these issues. I just need to, um, find the time and inclination to read them.

*Yes, this is the correct unit. Unfortunately.

Hmm, I have a sense of deja vu. Have I talked about this before?