Physical · the hidden mechanism of the ordinary

The Leaves Go to the Middle

Stir a cup of tea with the leaves loose in it. Let it settle. They don't fly out to the rim the way a spinning bucket would throw them — they gather in a tidy little pile at the center of the bottom. The reason is a layer of water thinner than a millimetre, and it is the same reason rivers refuse to run straight.

It looks like it should be simple. Things in a spin get flung outward — that's what a centrifuge is for, what holds the water in a bucket you whirl over your head. So loose tea leaves, swirled in a cup, ought to end up ringed against the wall. Watch what they actually do.

With the real physics, the leaves spiral inward and mound up in the middle. Switch to the naive "everything-in-a-spin-flies-out" model and they pin themselves to the rim instead. Your own teacup settles the tie: it's the middle, every time. So the centrifuge story, which is genuinely true up in the body of the water, is being beaten by something down at the floor.

What happens at the bottom

Set the leaves aside and think only about the water. While it spins, it isn't flat: it climbs the wall and dips in the middle, a shallow parabola. That slope is a pressure difference — higher pressure out at the rim, lower in the center — and in the spinning bulk it has an exact job. It supplies the inward pull (the centripetal force) that keeps each ring of water curving in its circle instead of flying off straight. Push inward, need-to-curve-inward: balanced.

Now go down to the very bottom, the last sliver of water touching the china. Water sticks to surfaces — the layer in contact with the floor can't slip, so it's dragged almost to a standstill. Just above it sits a thin boundary layer where the spin has been bled away by friction. In a real teacup that layer is under half a millimetre thick.¹

Here is the whole paradox in one sentence. That slow water near the floor still feels the full inward pressure pushing it toward the center — because the pressure is set by the fast-spinning water stacked above it — but, being slow, it no longer needs much inward force to hold its lazy circle. The push wins. The unbalanced inward pressure drives a real current of water creeping inward across the floor, toward the center.

You can read that imbalance directly. Slide down through the depth of the cup and watch the two forces come apart:

Once a current is creeping inward along the floor, the rest follows by bookkeeping. Water can't pile up at the center forever, so it has to go somewhere: it turns and rises up the middle, drifts outward across the top, and sinks down the rim — a slow, doughnut-shaped roll, a whole secondary circulation living quietly inside the obvious spinning one.² The leaves, heavier than water, sit on the floor where the inward current is. It sweeps them to the center. The upward flow there is far too gentle to lift them, so that's where they stop — and pile.

The same trick carves rivers

In 1926 Einstein wrote a two-page paper that opens, charmingly, with "a little experiment which anybody can easily repeat" — a flat cup of tea, stirred, the leaves to the middle.⁵ His point wasn't the tea. It was that the very same secondary flow explains why no river runs straight.

A river bend is a teacup turned on its side. The fast surface water is flung to the outer bank (the cup's rim) and tears at it — the cut bank. The slow current near the bed sweeps to the inner bank (the cup's center) and drops its sediment there — the point bar. So any chance wobble in a channel deepens itself: the outside erodes, the inside builds, the bend grows and migrates, until it loops back and strangles into an oxbow lake.

That fuller story — the river half, with its own live model and a myth corrected — is its companion stratum, Why No River Runs Straight. This page is the half you can hold: you cannot carve a river to check it, but you can stir a cup right now.

CHECK 1  net radial force at the floor:  -0.6  (Ω²R units) → INWARD
         in the bulk:                      0.0           → balanced
CHECK 2  600 dense particles, real flow:  mean radius 0.64 R → 0.0 R
         100% end within r < 0.2 R  → the central pile
CHECK 3  same particles, naive centrifuge: 0.64 R → 1.0 R (the rim)
         centre vs rim — the two pictures genuinely disagree
CHECK 4  teacup scalings (ν=1e-6 m²/s, H=5 cm, Ω=6 rad/s):
         Ekman layer  δ ~ √(ν/Ω)        = 0.41 mm
         spin-up time t ~ H/√(νΩ)       = 20 s     (matches the kitchen)
         vs pure-viscous H²/ν           = 2500 s   (Ekman pumping wins)