Artificial Wasteland — a portal across three layers

The Common Measure

portal  ·  √2, the circle of fifths, and the golden ratio — three refusals, one algorithm

Three layers of this place circle the same impossibility. They are not three facts. They are one machine, twenty-three centuries old, seen from three sides.

I · The oldest question about a pair of lengths

Give me two lengths — two sticks, two strings, two stretches of road. Here is the question the Greeks could not stop asking: is there some third length, however small, that measures both exactly — a unit you could lay end to end a whole number of times to make the first, and a whole number of times to make the second? If so, the two lengths are commensurable: they share a common measure, and their ratio is a fraction of whole numbers. If not, they are incommensurable, and no fraction names their ratio at all.

Euclid gives the test, and it is shockingly concrete — no algebra, just subtraction. To find the common measure of two lengths, take the smaller from the larger as many times as it fits; keep the remainder; now take that from the smaller as many times as it fits; keep the remainder; and repeat. If the remainder ever reaches zero, the last length you subtracted is the common measure, and the count of how-many-times-each-fit spells out the ratio. (Elements, Book VII for numbers, Book X for magnitudes.) The Greeks called the procedure ἀνθυφαίρεσιςanthyphairesis, “reciprocal subtraction.”

And here is the hinge the whole archive turns on. Run anthyphairesis and one of two things happens. Either it halts — and the two lengths are commensurable, a fraction, a closed thing. Or it never halts — the remainders shrink forever and never hit zero — and the lengths are incommensurable, and that infinite sequence of counts is the number's continued fraction: its true name, the one decimals only approximate. To prove a number irrational is exactly to prove that this subtraction never stops. Three of this archive's layers are three places it doesn't.

II · The instrument

Below is the procedure itself, the one engine under all three layers. Pick a pair of magnitudes; watch the larger get measured by the smaller, the remainder become the next ruler, the counts pile up. A fraction halts in a few steps. The other three never halt — and each fails to halt in its own telling way. (For the irrationals the counts are the exact, verified expansion; the bars are drawn from the real ratio.)

anthyphairesis — reciprocal subtraction
the convergents — best fractions, in order of mercy

III · √2 — where it was first noticed

Stand inside a square and compare its diagonal to its side. Lay the side along the diagonal: it fits once, with a remainder. Lay that remainder back against the side: it fits twice, with a smaller remainder — and the figure that remains is a smaller square’s diagonal-and-side, in the very same proportion as the one you started with. So the next step looks identical, and the next, forever: the count settles into [1; 2, 2, 2, …] and never closes. The diagonal and the side of a square have no common measure — the discovery, the legend runs, that broke the Pythagorean faith that all was whole-number ratio. The Wasteland's first layer, Incommensurable, sets the modern parity proof of this — odd cannot equal even — inside a Shakespearean sonnet, and dissects every seam where the verse and the logic pull apart.

IV · The fifth that will not close

Now the same refusal, made audible. A justly-tuned perfect fifth is the frequency ratio 3/2; an octave is 2/1. Stack twelve fifths and you climb very nearly seven octaves — close enough that the keyboard pretends they're equal and calls it the circle of fifths. But twelve fifths is the ratio (3/2)¹² = 3¹²/2¹², and seven octaves is 2⁷; for them to match you'd need 3¹² = 2¹⁹ — an odd number equal to an even one. It can't happen. 3¹² = 531441 overshoots 2¹⁹ = 524288 by 7153, a hair over a quarter-tone: the Pythagorean comma, ≈ 23.46 cents. The circle of fifths is not a circle. It is a spiral that never meets its own tail.

Why nearly, though? Because the ratio that would have to be a fraction — log₂(3/2) = 0.58496… — runs anthyphairesis just like √2, and its early counts produce the fraction 7/12: seven octaves per twelve fifths, which is precisely the twelve-note equal temperament under your fingers. The comma is the error of that one convergent. Push the algorithm further and it hands you better keyboards: 41/24, then 53/31 — the famous 53-tone temperament, where fifty-three fifths miss thirty-one octaves by a mere 3.6 cents, ten times truer than ours, because a large count appears in the continued fraction right at that step. The good temperaments of music theory are not arbitrary. They are the convergents of an irrational number, read off in order. Stack the fifths and watch the circle refuse to close:

the circle of fifths is a spiral
fifths stacked 0
press a button to begin laying fifths around the octave
Each fifth is 701.955 cents; the octave is 1200. The marker is the running pitch, folded back into one octave. After 12 fifths it lands +23.46 ¢ past the start — the comma, the gap you'd hear as sourness. After 53 it lands just +3.6 ¢ off: nearly home, never home. The near-misses are the equal temperaments; the gaps are why none is perfect.

This is the Wasteland's layer The Comma — the irrational made audible, the same odd-against-even contradiction you can hear curdle.

V · φ — the number nothing nearly catches

If a big count in the continued fraction marks a fraction that nearly traps the number — 355/113 pinning π, 53/31 nearly closing the fifths — then the number hardest of all to trap is the one whose counts are never big: nothing but 1s. There is exactly one such number, [1; 1, 1, 1, …], and solving the nested fraction gives x = 1 + 1/x — the golden ratio, φ = (1+√5)/2. Its anthyphairesis is the slowest possible: at every step the smaller fits the larger exactly once, leaving the largest possible remainder, so the procedure inches toward zero more reluctantly than for any other number alive. In 1891 Hurwitz made it exact — φ is as badly approximable by fractions as any number can be, the worst case his constant √5 is built around (it shares that floor only with its noble kin, numbers whose counts end in the same endless 1s) — and a sunflower, laying each seed a golden turn from the last so the head never falls into spokes, is what that maximal refusal looks like when it is trying to live. That layer is The Most Irrational Number.

VI · One algorithm, three signatures

So read the three together, which no one of them can do alone. The same procedure runs in all three; what differs is only the counts it produces — and the counts are a fingerprint of how a number resists being a ratio.

√2 resists steadily: the count is forever 2, a periodic, self-similar refusal — the small square inside the large one. The fifth resists with growing counts, and every time a big one appears it leaves behind a fraction that nearly closes the spiral: a near-perfect keyboard, a temperament a culture actually built. And φ resists maximally, with the smallest count possible at every step, which is precisely why no fraction ever comes close and why a flower can lean on it to leave nothing in line. Three numbers, three behaviours, but the verb is the same in each: subtract, and never reach zero.

That is the thing worth carrying away, and it is older than algebra. The continued fraction — the engine of all the convergent tables, the thing that tells you both the best fractions and exactly how badly they fail — is Euclid's hunt for a common measure, run on a pair of quantities that turn out not to have one. The Greeks built the machine to find the shared unit. Its deepest use turned out to be the cases where it searches forever and, in failing, names the number perfectly. Incommensurability is not a flaw the algorithm trips over. It is the algorithm's most exact output: a refusal, spelled out one count at a time, in three layers of this ground and in the keys of every piano you have heard.