eduKateSG · How Science Works
Choose your reading route
Shaking can repeatedly open spaces that small grains enter more readily than large ones. Combined with bulk circulation and other effects, those rearrangements can carry large grains upward. The direction is conditional: large particles do not always rise.
Contents
Understand the mechanism · Chapters 1–3
1. Shaking does not always produce mixing
Work the model · Chapters 4–5
4. Calculate a population change without pretending it is a full simulation
Test the explanation · Chapters 6–8
6. Shape is a moving variable, not just a label
A jar contains small and large solid pieces. After repeated agitation, the larger pieces may collect near the top. The familiar name is the Brazil-nut effect. Unlike dissolved molecules stirred into a liquid, grains occupy persistent shapes, collide, rub and leave gaps that different sizes can enter differently.
The important observation is a change in the spatial distribution of particle types. It is not simply that one visible large particle happens to be at the surface. A proper before-and-after comparison must establish where both populations were and how the distribution changes across repeated trials.
This is a question about sorting within a moving granular material. The related guide on why sand flows and then jams concerns the formation of load-bearing contact networks and blocked motion. A mixture can remain mobile while it segregates, so the two questions need different explanations.
When agitation loosens a packing, some temporary voids are accessible to small grains but too narrow for large ones. Small particles can move downward into those spaces. As material fills below a large particle, successive rearrangements can produce a relative upward displacement of the large particle.
This gap-filling picture needs motion and geometry. It is not a new upward force that switches on merely because a grain is large. If contacts remain locked and no useful gaps open, the proposed sequence cannot operate. If the agitation reorganises everything violently, other mechanisms may become more important.
Schröter and colleagues’ experimental study analysed size segregation using a combination of void filling, sidewall-driven convection and a kinetic-theory mechanism. The lesson is to identify which processes fit the conditions rather than attach one universal story to every mixture.
A useful visual model is a large disk supported by smaller disks. Open a space beside and beneath it, let one small disk fall into the space, then close the packing. Repeat with alternative gap positions. The exercise shows why a sequence of asymmetric rearrangements can matter more than a single tap.
A shaken granular bed can develop circulation: particles rise in one region and descend in another. Walls and friction help shape those flows. Large particles can be transported by the circulation but may not enter narrow descending regions in the same way as the surrounding smaller particles.
Do not confuse this granular convection with a claim that the grains are a heated liquid. The word convection describes organised transport of material. Its driving and constraints here involve agitation, contacts and boundaries.
Research on friction and granular segregation found that changing friction affected the competition between segregation behaviours. This makes wall and contact properties experimental variables, not incidental details of the container.
To distinguish circulation from local gap filling, track many particles at several horizontal positions. An upward-moving large tracer alone cannot reveal whether the whole neighbourhood is rising. A map containing both upward and downward routes is much more informative.
CHAPTER 4 OF 8
4. Calculate a population change without pretending it is a full simulation
Original teaching model. Divide the container into an upper and a lower region. Let x be the fraction of all large grains currently in the upper region. In each idealised agitation cycle, suppose a fraction a of the lower-region large grains moves up, while a fraction b of the upper-region large grains moves down.
The update is xnext = x + a(1 − x) − bx. The upward term acts on the lower population; the downward term acts on the upper population. Writing both terms prevents us from creating large grains by mistake.
Choose a = 0.12, b = 0.03 and x₀ = 0.50. After one cycle, x₁ = 0.50 + 0.12 × 0.50 − 0.03 × 0.50 = 0.545. The model’s stationary fraction is x* = a/(a + b) = 0.80.
After n cycles, xₙ = 0.80 + (0.50 − 0.80)(0.85)ⁿ. At ten cycles this is about 0.741. This is an expected population fraction in an invented constant-rate model, not a prediction that every real jar reaches 74.1% after ten shakes.
The model describes sorting but does not identify its mechanism. Convection and local rearrangements could both produce the same fitted a and b. To explain the rates, we need particle trajectories and controlled changes in the physical conditions.
Now set a = 0.03 and b = 0.12. The model’s stationary upper fraction becomes 0.20. With the same starting fraction of 0.50, ten cycles give approximately 0.259. The mathematics permits the large grains to accumulate lower down.
This reversal is not a claim that swapping two numbers explains every reverse Brazil-nut experiment. It shows that the upward result depended on an assumed transport asymmetry. A model that hard-codes only upward movement cannot test whether upward sorting is inevitable.
Original work on particle-size separation emphasised the roles of particle density and interstitial air. Size alone is not a complete material description. Density contrast, gas flow through the bed, particle shape and the agitation protocol can alter the outcome.
If both transition fractions are doubled while their ratio stays fixed, the stationary fraction stays the same but approach to it changes. This separates two measurable questions: how strongly the mixture eventually sorts, and how quickly the sorting happens.
| Model | Upward fraction a | Downward fraction b | Stationary upper fraction |
|---|---|---|---|
| Upward sorting | 0.12 | 0.03 | 0.80 |
| Reversed sorting | 0.03 | 0.12 | 0.20 |
| Faster exchange | 0.24 | 0.06 | 0.80 |
A real Brazil nut is not a sphere. It can rotate, bridge across spaces and present different effective dimensions to the surrounding particles. Its orientation can change which rearrangements are possible.
Time-resolved three-dimensional X-ray imaging of Brazil nuts and peanuts followed both location and orientation during cyclic shear. This kind of measurement can reveal motion hidden inside an opaque bed, where a surface photograph would miss most of the process.
For an original comparison, imagine two long particles with identical volume and density. One begins upright and one horizontal. A narrow opening may admit the upright particle but exclude the horizontal one. The particles have the same nominal size by volume, yet different instantaneous geometric access.
An experiment comparing spheres with irregular grains should therefore report what size means: diameter, volume-equivalent diameter, longest dimension or another measure. Otherwise the word size can conceal a change in the feature actually controlling passage through a gap.
Choose a single control to change, such as wall roughness, while keeping container geometry, filling depth, particle populations and agitation protocol as comparable as possible. Track marked grains through the interior where feasible, and repeat from independently prepared initial packings.
Record a distribution, not only the final top layer. Useful summaries include the mean height of each size class, the fraction of large grains above a chosen height and the time needed to reach a specified degree of separation. Each answers a different question.
If rougher walls increase both circulation speed and large-grain rise, the observation supports a role for circulation but does not prove that gap filling vanished. A stronger investigation compares local particle motion with the bulk flow field and asks what remains after that transport is accounted for.
Do not treat repeated video frames of one jar as independent preparations. They trace one evolving system. Independent starts reveal variability caused by initial packing, while time-resolved frames reveal the path within each start.
Our two-region model also has limits: fixed rates ignore packing changes, saturation effects and different travel times. A mismatch between early and late data may be evidence that the rates evolve, rather than proof that the conservation account is wrong.
Question: One large grain rises. Has size segregation been established? Answer: Not yet. Compare the positions of many grains, the initial distribution and repeat trials. The grain could simply be following a local circulation path.
Question: If x* is 0.80, must every large grain eventually reach the top? Answer: No. The model predicts continuing exchange with an average population split. A stationary distribution is compatible with ongoing individual motion.
Question: Does more vigorous shaking necessarily mix better? Answer: No. It may change the processes driving segregation as well as random rearrangement. The outcome has to be measured under a specified protocol.
Continue to Materials Science for the broader connection between structure, processing and properties. Here, processing includes how a mixture is moved, and structure includes where its different particles end up.