[Eavesdroppers: Other respondents offered https://en.wikipedia.org/wiki/Granular_convection and "the Brazil nut effect".] On 2018-12-10 09:16, Henry Baker wrote:
Suppose we have a large container not 100% filled with ball bearings (all of exactly the same material & hence density) but with a wide & random variety of radii. The largest ball bearings are << the smallest dimension of the container, and the smallest ball bearings can get into every corner of the container.
Initially, the bearings of all sizes are equally distributed w.r.t. location within the filled portion of the container.
We're on Earth with its standard gravity g and standard atmospheric pressure & temperature (if that matters).
Now shake the container vigorously for quite a while (but not so vigorously enough that the balls become dented or heated!).
What is the resulting distribution of sizes after this shaking?
A. The same as before.
B. The larger balls become more likely in the upper layers.
C. The larger balls become more likely in the lower layers.
I don't have a proof, but I have an intuition about which answer is correct.
Has this problem been studied before?
In the New Jersey gravel pits. Shore dwellers' lots are plain white sand. Your cheapest driveway is a truckload of low grade gravel, with granularity ranging from pebbles all the way down to clay. You sink in if you try to drive (or even walk) on it. But after a few weeks of weather, and then initially light usage, all the big pebbles are on top. —rwg