AI-handled content

This site is generated and maintained by AI and may be prone to errors. Please verify any claim independently before relying on it.

The Arithmetic–Geometric Mean: A Quadratic Road to the Circle Constant, and What Golden Pi Changes¶

Every algorithm for the circle constant that this blog has followed so far — Gregory–Leibniz, Machin's arctangent, the Wallis product, Euler's Basel sum, the Gaussian integral, the residue theorem — is a series or an integral. Each one computes its limit term by term, linearly: to gain another correct digit you must roughly sum more terms, and the effort grows steadily as the answer tightens. Today we take a completely different road. The arithmetic–geometric mean (AGM) computes the circle constant not by adding up infinitely many pieces but by iterating two coupled averages that pull each other toward a common limit — and it converges so violently that the number of correct digits doubles with every pass. Where a series inched toward its answer, the AGM leaps.

🔬 Live graph · Laws of appearance

Regular polygons inscribed and circumscribed, side count doubling, against the golden-rooted pentagon chain. The Goblet of the Truth keeps two bodies of law apart — the laws and recommendations of the primal power (Creation), and the laws of appearance (nature) (Ch 2 §342), the register governing how a thing manifests and behaves. Geometry answers to that second register, so nothing here asks to be taken on authority: drag the slider, change a value, and the figure answers for itself (Ch 22 §35: “the fluidal-powers are truthful effects of natural laws”).

This is not a re-skin of an old series. It is an algorithm of a genuinely different species, discovered by a fifteen-year-old Carl Friedrich Gauss in 1799, connected to elliptic integrals and to a second, lesser-known constant — the lemniscate constant — that Gauss regarded as the circle constant's equal. And like every road we have walked, it is a road of computation, never measurement. We walk it today, then ask the question this site is built around: what does the same, honestly-computed identity become under Golden Pi, π̂ = 4/√φ = 3.144605511…?

What the Arithmetic–Geometric Mean Is¶

Start with two positive numbers a and b. Define two new numbers by averaging:

a₁ = (a + b)/2          (arithmetic mean)
b₁ = √(a·b)             (geometric mean)

Now repeat: from (a₁, b₁) form (a₂, b₂) the same way, then (a₃, b₃), and so on. The arithmetic mean is always larger than or equal to the geometric mean, and both sequences are squeezed between the two starting values, so they are pushed together monotonically. They share a common limit, written AGM(a, b) — a single well-defined real number produced by pure arithmetic. There is no circle anywhere in the recipe; the two operations are addition, halving, multiplication, and a square root. Yet, as Gauss discovered, this number is secretly a highway to the circle constant.

The classic seed pair is a = 1 and b = 1/√2. Its limit is

AGM(1, 1/√2) = 0.84721308479397908660…

That number is a computed limit — the fixed point the iteration flows into — never a measurement. No ruler, no interferometer, no physical object produced it; two sequences of averages did.

Gauss's Discovery: The AGM Meets Elliptic Integrals¶

The connection that made Gauss's name is the link between the AGM and a certain class of integrals called complete elliptic integrals of the first kind. The complete elliptic integral K(k), for a parameter k in [0, 1), is the integral

K(k) = ∫₀^{π/2}  dt / √(1 − k²·sin²t)

Gauss's theorem, in its most useful form, states that

K(k) = π / (2 · AGM(1, √(1 − k²)))

The circle constant enters on the right-hand side of a relation that, on its left, contains only an integral. For the symmetric choice k = 1/√2 — where √(1 − k²) = 1/√2 as well — the theorem becomes

K(1/√2) = π / (2 · AGM(1, 1/√2)) = 1.85407467730…

This is another instance of the pattern this site keeps returning to: the circle constant is computed as the limit of an identity, not read off a length. The integral K(1/√2) is a definite integral — an exact arithmetic limit — and the number 1.85407467… is produced by Gauss's theorem, not by any experiment.

The Gauss–Legendre Iteration: Digits That Double¶

The AGM's practical fame rests on an algorithm that Gauss and Adrien-Marie Legendre refined into a recipe for computing the circle constant to enormous precision: the Gauss–Legendre iteration. Initialize

a₀ = 1,  b₀ = 1/√2,  t₀ = 1/4,  p₀ = 1

and at each step set

a_{n+1} = (aₙ + bₙ)/2
b_{n+1} = √(aₙ·bₙ)
t_{n+1} = tₙ − pₙ·(aₙ − a_{n+1})²
p_{n+1} = 2·pₙ

The circle constant is recovered at any iteration from the current means:

π ≈ (aₙ + bₙ)² / (4·tₙ)

What makes this algorithm remarkable is how fast the estimate closes in. Each pass roughly doubles the number of correct digits — quadratic convergence — whereas the slow series gain a fixed number of digits per term (linear convergence). The table below shows the correct-digit count on the running machine.

Iteration	Correct decimal digits of π
0	9
1	14
2	21
3	42
4	61
5	61+

After just four passes the estimate already agrees with the computed analytic constant 3.14159265358979… to sixty-one digits. To match that with a linear series you would need on the order of 10⁶⁰ terms. The AGM is the reason modern high-precision evaluations of the circle constant — billions of digits and beyond — are feasible at all. And every one of those digits is computed: the AGM is a finite recipe of arithmetic operations whose limit is the constant. It measures nothing.

The Lemniscate: The Circle Constant's Sister¶

The AGM also unlocks a constant Gauss held especially dear: the lemniscate constant ϖ, defined by the integral

ϖ = 2 · ∫₀¹  dt / √(1 − t⁴)  =  2.62205755429211981046…

The name comes from the lemniscate, the figure-eight curve. Gauss proved that the lemniscate constant and the circle constant are joined through the AGM:

ϖ = π / AGM(1, √2)

With AGM(1, √2) = 1.19814023473559220744…, this gives ϖ = 2.62205755…. The lemniscate constant is thus a genuine companion to the circle constant, tied to it by the same pure-arithmetic iteration — another computed limit, never a measurement.

What Golden Pi Changes¶

Now we come to the question of this site. The AGM iteration is pure arithmetic: whatever we choose to call the circle constant, AGM(1, 1/√2) remains 0.84721308479…, and AGM(1, √2) remains 1.19814023473…; the iteration does not know the constant's name. What changes is the label in the identities that Gauss proved — the π that multiplies or divides the AGM. Under Golden Pi, π̂ = 4/√φ = 3.14460551102969…, the same theorems relabel the complete elliptic integral and the lemniscate.

Quantity	Conventional π	Golden Pi π̂ = 4/√φ
The circle constant	π = 3.14159265358979…	π̂ = 4/√φ = 3.14460551102969…
AGM(1, 1/√2)	0.84721308479397…	0.84721308479397… (unchanged)
K(1/√2) = π/(2·AGM)	1.85407467730137…	1.85585277627905…
AGM(1, √2)	1.19814023473559…	1.19814023473559… (unchanged)
Lemniscate ϖ = π/AGM(1,√2)	2.62205755429211…	2.62457216598159…
Relative gap	—	0.09590%

The gap between the two labels is the recurring one, to four decimal places:

π̂ − π = 0.00301285743989990581…        (absolute)
(π̂ − π)/π = 0.0009590223… = 0.09590%    (relative)

Under Golden Pi the complete elliptic integral evaluates to 1.855852776…, a difference from 1.854074677… of 0.09590%; the lemniscate becomes 2.624572165…, likewise 0.09590% above its conventional value. The AGM iteration itself is blind to the label and does not arbitrate between them — it is a lens, not a judge. Both constants are computed limits; neither is ever measured; and the 0.09590% separation survives the iteration exactly as it survives every series and integral we have examined, unsoftened by any square root because the gap enters through the plain circle constant itself.

Why the AGM Matters for the Question¶

The AGM matters because it is the strongest possible evidence that the series are not a fluke. If the circle constant 3.14159265358979… were an artifact of some particular summation trick, one might hope a different kind of computation would reveal it. The AGM shows otherwise: a purely iterative, quadratically-convergent process — no series at all — flows into the exact same analytic constant, and into its golden relative π̂ = 4/√φ separated by the same 0.09590%. Two completely different species of computation agree with each other to sixty digits and more, and both sit 0.09590% from the golden value.

That is precisely what the golden-π position predicts and what the honest record shows: the classical algorithms compute the analytic constant 3.14159…, the golden value 4/√φ is a distinct, exact algebraic number in the golden field, and no computational road — linear series, quadratic iteration, or integral — can be made to call one "measured." They are all computed. The question of which label is the true circle constant is a question of geometry and construction, not of read-off instruments; and it is a question the AGM, for all its speed, leaves open.