How Fibonacci stumbles onto φ
2026-05-08
The Fibonacci sequence is one of the simplest rules in mathematics. Start with two ones. Each new number is the sum of the two before it.
1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233, ...
Now divide each Fibonacci number by the previous one. The numbers are clean integers; the ratios are not. They land at strange, uneven decimals - and they bounce.
—
gap to φ: —
Press play. The first ratio is 1, sitting far to the left of the dashed line. The next is 2, far to the right. After that the dots crowd in, alternating sides, each closer than the last. By the bottom of the stack they have piled on top of one another. The line they are converging on is φ ≈ 1.61803398..., the golden ratio.
Why this number, of all numbers
Call the limit L. The Fibonacci rule is Fn+1 = Fn + Fn-1. Divide both sides by Fn:
Fn+1 / Fn = 1 + Fn-1 / Fn = 1 + 1 / (Fn / Fn-1)
If the ratios on the left approach L, the ratios on the right do too - they are the same ratio, one step earlier. So L has to satisfy:
L = 1 + 1/L
Multiply through by L and you get L² = L + 1. The positive root of that equation is (1 + √5) / 2 = φ. There is nowhere else the ratio could end up.
Nothing in this argument used the starting values being 1, 1. Begin with 2, 5: 2, 5, 7, 12, 19, 31, 50, 81, ... The ratios still chase φ. Begin with anything you like. The rule does the work.
Why they bounce instead of marching
Take a ratio that has fallen a touch short of φ - say 21/13 ≈ 1.615, low by about three thousandths. The next ratio is 1 + 1/1.615 ≈ 1.619: now over φ, by about one thousandth. The map r ↦ 1 + 1/r is a decreasing function. Bigger in, smaller out. So a ratio below φ always produces a ratio above, and vice versa. The bouncing is built into the rule.
Each bounce shrinks. The slope of 1 + 1/r at r = φ is -1/φ² ≈ -0.382: a small error in becomes an even smaller error out, with the sign flipped. After twelve steps an initial gap of 0.6 has been multiplied by roughly 0.382 twelve times - shrunk by a factor of nearly four million.
The same identity, wearing a different hat
In the chessboard-that-gains-a-square puzzle, four pieces of an 8×8 board reassemble into a 5×13 rectangle and a cell appears from nowhere. The trick depends on Cassini's identity:
Fn-1 · Fn+1 − Fn² = (−1)n
Divide through by Fn-1 · Fn and rearrange:
Fn+1 / Fn − Fn / Fn-1 = (−1)n / (Fn-1 · Fn)
Two consecutive ratios differ by a quantity whose magnitude shrinks like one over a Fibonacci-squared, and whose sign alternates. That is the bounce, written algebraically. The same identity that fakes a missing square also tells you exactly how tightly the ratios spiral in on φ.
One rule. Twelve ratios. φ.