The Golden Angle: Why 137.5°

Cut a sunflower head in half and look at how the seeds are arranged. There are no rows, no rings, no obvious grid — and yet nothing is wasted. The florets crowd together so tightly that you would struggle to slide another seed in anywhere, from the dense center to the loosened rim. The same pattern appears in pinecones, pineapples, artichokes, and the spiral of leaves climbing a stem. It is not an accident of biology so much as a consequence of arithmetic, and the arithmetic turns on a single angle: roughly 137.507°. This essay explains where that number comes from, and why it — and almost nothing near it — produces the packing you see.

The accompanying Plate XIV — Phyllotaxis lets you turn the angle by hand and watch the spirals appear, drift, and lock; it is worth keeping open while reading.

A rule for growing a seed head

The generative rule plants observe is astonishingly simple. Seeds are laid down one at a time from the center outward. Each new seed is placed at a fixed divergence angle $\theta$ further around than the previous one, and pushed out to a radius that grows like the square root of its index. If we number the seeds $n = 1, 2, 3, \dots$, then seed $n$ sits at polar coordinates

$r_n = c\sqrt{n}, \qquad \phi_n = n\,\theta.$

The square-root radius is the part that is not about beauty: it is about area. A disk of radius $r$ has area proportional to $r^2$, so if seeds are to occupy area at a constant rate as the head grows, the $n$-th seed must land near radius $\sqrt{n}$. With that settled, every interesting question collapses into one: what should the angle $\theta$ be?

The angle is the only free parameter, and it does all the work.Only the fractional part of $\theta$ relative to a full turn matters; adding any whole number of turns places the seed in the same direction. Two seeds point in nearly the same direction whenever $n\theta$ and $m\theta$ differ by close to a whole number of turns. So the design problem is to choose a turn-fraction $\theta/360°$ whose multiples avoid clustering — for every spacing, at every scale.

Why rational angles fail

Suppose you choose a rational fraction of a turn, say $\theta = \tfrac{1}{5}$ of $360° = 72°$. Then seed $5$ lands in the same direction as seed $0$, seed $6$ aligns with seed $1$, and so on forever. Every seed falls onto one of just five rays emanating from the center. The head grows five long, sparse spokes with gaping wedges of empty space between them. You have packed an entire two-dimensional disk using only five directions.

Any rational $\theta = p/q$ (in lowest terms) does the same thing: the directions repeat with period $q$, so all seeds collapse onto $q$ radial spokes. A large denominator gives more spokes and so looks busier, but the failure is identical in kind — the angle is periodic, and periodicity in the angle means structure that leaves the gaps between spokes permanently unused.

The lesson is that $\theta/360°$ must be irrational, so its multiples never exactly repeat. But irrationality alone is not enough. Numbers very close to a simple fraction behave, over the scales that matter, almost exactly like that fraction. An angle near $\tfrac{1}{3}$ of a turn will produce three thick near-spokes for a long time before the tiny discrepancy finally smears them out. We do not want merely irrational; we want maximally far from every simple rational at once — across all scales, never settling into spokes. That phrase has a precise meaning, and it points to one number.This is a question about how well a number can be approximated by fractions — squarely a topic in the theory of numbers rather than in biology.

The most irrational number

Every real number has a continued fraction expansion, a tower of nested reciprocals,

$x = a_0 + \cfrac{1}{a_1 + \cfrac{1}{a_2 + \cfrac{1}{a_3 + \cdots}}},$

and truncating it gives the best possible rational approximations to $x$ — its convergents. A large term $a_k$ deep in the tower signals that the convergent just before it is unusually accurate: the number is well approximated by a small fraction there. The numbers that resist rational approximation most stubbornly are therefore those whose continued-fraction terms are all as small as possible — all equal to $1$.

That number is the golden ratio,

$\varphi = 1 + \cfrac{1}{1 + \cfrac{1}{1 + \cfrac{1}{1 + \cdots}}} = \frac{1 + \sqrt{5}}{2} \approx 1.6180339887.$

Written as $[1; 1, 1, 1, \dots]$, it has the slowest-converging continued fraction of any number, which is the rigorous sense in which it is the most irrational real number — the hardest to mistake for a fraction at any scale.Hurwitz’s theorem makes this exact: $\varphi$ and its relatives are the worst-case numbers for rational approximation. Its convergents are the ratios of consecutive Fibonacci numbers — $\tfrac{1}{1}, \tfrac{2}{1}, \tfrac{3}{2}, \tfrac{5}{3}, \tfrac{8}{5}, \tfrac{13}{8}, \dots$ — each the best rational approximation to $\varphi$ for its size, and each only barely better than the last.

To turn $\varphi$ into an angle, take the fraction of a turn that is “the golden cut,” the irrational portion left over:

$\theta = \frac{360°}{\varphi^2} = 360° \,(2 - \varphi) \approx 137.50776°.$

The complementary slice, $360°/\varphi \approx 222.49°$, is the same angle measured the other way; by convention we name the smaller, $137.5°$. This is the golden angle, and it is the unique divergence angle whose successive seeds are spread as evenly as a sequence of multiples can ever be.

How Fibonacci numbers reappear as spirals

Set $\theta = 137.5°$ and the seeds settle into the familiar interlocking spirals — the parastichies. Here the convergents return in visible form. Because $\tfrac{8}{13}$ of a turn is an excellent approximation to the golden fraction, seeds whose indices differ by $13$ point in almost the same direction and so appear to lie on a common gentle spiral; counting these gives a family of about $13$ spiral arms. But $\tfrac{13}{21}$ is a slightly better approximation, so seeds differing by $21$ are even more nearly aligned, yielding a tighter family of $21$ arms winding the other way. Which family the eye catches depends on how far out you look, and the two visible counts are always consecutive Fibonacci numbers.

This is the punchline tying the strands together. The radial $\sqrt{n}$ law sets the scale; the golden angle sets the directions; and the Fibonacci convergents of $\varphi$ — the unavoidable mathematical shadow of its continued fraction — determine exactly which spiral counts a viewer perceives. A sunflower with $34$ spirals one way and $55$ the other is not decorated with Fibonacci numbers. It is computing the convergents of the most irrational number, one seed at a time.

Further reading

• The site’s own Library, for the surrounding mathematics of structure and approximation.
• Continued fractions — Encyclopedia of Mathematics, for the approximation theory behind why $\varphi = [1;1,1,\dots]$ is extremal.