The experiment below was first executed in 1963 by the engineer J.C.
Williams to explain what has been dubbed by physicists the Brasil nut effect.
This occurs when multiple grain sizes are randomly placed in the same con-
tainer and then shaken vertically. The grains eventually sort themseves by
size with the largest particles at the top. Williams proposed that as the grains
“jump” when shaken a space is created below each grain. The smaller par-
ticles find their way down into the void left by the large particles. If shaken
enough times all of the particles will sort themeslves. To test this theory, Wil-
liams placed one large grain in the center of a mixture of substantially small-
er grains. The grains immediately to the side of the large central grain were
dyed darker in order to track their movement. As the mixture is shaken the
darker grains reorganize to form a cone beneath
the large grain. Each time the mixture is shaken
the large grain will rise approximately the thick-
ness of the dark cone or d. A computer model o
However, in 1993 Sydney Nagel proved that this theory explains only
this experiment completed in 1992 seemed to
half of the phenomenon. In this experiment glass beads of similar size and
prove Williams theory correct.
one centrally placed larger bead were placed in a cylinder and dyed across
a horziontal axis in order to track movement much like the previous experi-
ment. The cylinder is then shaken vertically. This time the grains exhibit cir-
culating motions like those in a convection cell. As the large central bead is
jacked up towards the top, the smaller beads are initially forced downwards
in the center of the mixture. As the mixture is continuously shaken the beads
on the outer edge move downwards at ta faster rate than those in the center.
Eventually this forces the central beads back up again as the beads along the
walls make their way towards the center. The large bead is not forced down-
wards when it reaches upper wall because it can not fit in the narrow band
of downward moving beads. This causes the size seperation.
When a thin layer of grains and then shaken, several different paterns of peaks emerge along the surface. Michael Fara-
day proposed that a difference in air pressure caused by the air pockets between the grain mixture. When the grain is shaken
enough the bottom layers jump from the bottom creating a negative difference in air pressure which then sucks adjacent grains
under the pervious layer. These grains can develop into striped, triangular, square, or hexaonal wave patterns. The patterns can
change based on the shaking frequency. By changing the frequency or amplitude of the shake the same set of grains will tranis-
tion through the different possible configurations seen in the graph to the left. If the amplitude exceeds a certain threshold, dis-
order will emerge regardless of the shaking frequency.
At low vibration amplitudes the entire mixture rises and falls at the same rate, but as the first critical amplitude is reached a
bifurcation emerges in the falt layer resulting in striped patterns at high frequencies and squared patterns at low frequencies.
At a second critical amplitude yet another bifurcation occurs which results in hexagonal patterning. This doubled up oscilation
results in a pattern transition which switches between peaks and cells.
Troy Shinbrot noticed that the rate which the grains fall at is that of gracity adn therefore remains constant and resolved to
look deeper into the horizontal components of the grains’ motions. Since the patterns are horizontal, he proposed and tested
weather or not simply horizontal movement would produce these patterns. In a set of 100 shakes, complex patterns emerged
based on the strength of what is known as the randomizing effect. This occurs from the loss of energy in heat form with each
consecutive collision. These sets of patterns were more extensive than those of the vertical shake ex-
periments. Although the experiment is a gross simplification of all of the forces at work, the horizonal
motion model proves that gravity does not need to be taken into account.