Piles That Never Empty

It starts with a small, almost childish question: why can't the square root of certain whole numbers ever be a whole number? Take √2. It is not 1 and not 2, and — as the Greeks found — it is not any fraction either. Its decimal runs forever with no repeating block. You cannot pin it down; you can only close in on it. That refusal to arrive is the seed of everything that follows.

What grew out of it, over one long evening, was not a new theorem. Every fact below is old — Zeno, Archimedes, Euler, Mertens, Lagrange, Torricelli. What was new, at least to me, was a single road that runs through all of them: the image of a pile you keep chipping at, and the one question that turns out to organize the whole landscape — does the pile ever empty?

A machine that removes forever

Imagine feeding pieces into a chipper. Each piece is a reciprocal of a prime: ½, ⅓, ⅕, ⅐ … and you keep adding them to a running total. Every 1/p has a repeating decimal — 1/7 = 0.142857 142857 … — so each one is a small, exact, endlessly cycling clock. The chipper runs by these clocks.

Now the question that decides everything: as you add reciprocal after reciprocal, does the total settle on a finite number, or does it climb forever? For the primes, Euler answered it in 1737 — the sum Σ1/p diverges. It never settles. The pile never empties. That is the whole character of the woodchipper: an exact, orderly, unstoppable process that removes without end.

Empties versus never-empties

Not every such pile refuses to close. The dividing line is a growth rate. Line up your reciprocals; if the terms shrink fast enough, the total lands on a finite ceiling and the pile empties. If they shrink too slowly, it climbs forever. The wall between the two sits, almost exactly, at 1/n — the harmonic series, the borderline that just barely diverges.

Turn each machine inside out — replace 1/p by p, 1/√n by √n — and you can see the answer. Stand a pillar at each value and let it rise to pierce a curve. The machines that never empty fill the curve; the machines that empty abandon it, their pillars racing off the top.

The diagonal decides

Once you order the machines by how fast they grow — √n, then n, then n², 2ⁿ, n! — the whole night collapses onto one axis. And the mark on that axis where never-empty flips to empty is the line y = x: plain linear growth. Slower than the diagonal, the pile never empties. Faster, it empties. The identity line — which is its own reciprocal 1/x — is literally the border, because x and 1/x meet only there.

And the boundary is not a razor. The primes grow like n·ln n — a hair faster than linear — yet their sum still diverges. They live above the diagonal and remain never-empty, in a fuzzy collar with no sharp far edge, an infinite hierarchy of log-upon-log rates all still refusing to close. The primes were the right machine to build on precisely because they grow as fast as anything can while still never emptying. They ride the very lip of the zone.

Settling into an infinite pattern

Return to √2, the number that refused to arrive. Its refusal is orderly. Written as a continued fraction it is [1; 2, 2, 2, 2, …] — the 2 repeating with total regularity (Lagrange: a continued fraction repeats exactly when its number is a quadratic irrational). It does not fail to settle. It settles into an infinite pattern. Equilibrium reached at once; the equilibrium is a cycle, not a stop.

That pattern runs a clock. The best whole-number approximations to √2 are the Pell numbers — 1, 2, 5, 12, 29, 70, 169, … — each built from the two before by Pₙ = 2·Pₙ₋₁ + Pₙ₋₂, the repeating 2 as the gear. The pile of Pell reciprocals shrinks fast, so 1/Pₙ is an emptying machine — the mirror of the woodchipper across the wall. And the Pell numbers that happen to be prime — 2, 5, 29, 5741, 33461 — are the record-holders, each closing on √2 more tightly than every prime before it, the gap shrinking by the same factor forever and landing on zero never. Approaching a place with no rational address, along a perfectly lawful road.

The body of an emptying machine

Give a machine a shape and it becomes a spiral. The Fibonacci spiral is the clean case: a logarithmic curve whose radius scales by a fixed factor each quarter-turn. It answers the whole night's question — does the infinite process arrive? — by direction, both answers in one curve.

Wind inward and the spiral reaches its centre — the eye — in finite total length, though it turns infinitely many times. That finite total is Zeno's paradox resolved as geometry: infinitely many shrinking steps, a finite whole, and it arrives. Wind outward and it races to infinity and never arrives. The type of spiral even names the zone: a logarithmic spiral is the emptying, red case; the Archimedean spiral, whose radius grows only linearly, is the y = x boundary itself.

One honesty, so the joint holds: this explains why the Fibonacci spiral has an eye and finite inward length — because it is geometric, emptying — but it does not explain what makes the golden ratio special. Any exponential base gives an eye. The golden-ness lives on a different axis: the continued fraction [1; 1, 1, 1, …], the slowest of all, the most stubbornly irrational number there is. The ruler tells you the spiral arrives; it does not tell you why φ.

Two knobs, one map

The last picture ties the two threads together. Take a polygon and let it turn a fixed angle each step while rescaling its radius — a camera's iris, an aperture. It has two independent dials. The number of sides sets resolution: three sides is a coarse triangle, infinitely many is a smooth circle — the coarse-versus-fine circle from the very first question, the difference between a rough rational bracket and the exact curve a square root draws. The scale factor sets fate: below one it shrinks to a point and empties; exactly one it closes into the polygon; above one it opens forever and never empties.

Resolution runs across; emptiness runs down. Smoothness and shrinking are separate hands on separate dials — one makes the aperture round, the other makes it vanish or grow. Every object of the evening is somewhere on that map.

Everything here is an exact, orderly, unstoppable process converging on something it never touches — because the target is irrational and the process is made of integers. That is not many phenomena. It is one, seen from many sides.

The square root that cannot be whole, the woodchipper pile that never empties, the arc the primes outline but never fill, the nested triangle that swells to 99.99998% and never to 100, the gaps that follow a perfect law straight toward a place they cannot reach, the spiral that empties into an eye it takes forever to enter — and, at the last, the boundary between empty and never-empty, which itself frays into an endless layering rather than resolving to a point. Even the divider will not settle. There is no bottom to it.

None of it is new to mathematics. All of it was new to me for an evening — which is its own kind of arrival, the kind that counts. The primes float in that unsettled band, on the diagonal's edge, keeping perfect time to a rhythm laid down by a number that will never be a fraction. That is where they live, and now I know the address.