Skip to content
TheChatArchiveThe Chat Archive
o3-2025-04-16· OpenAIHugging Face LMArena Human Preference 140K

how was it proven that pi isn't rational?

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Googlegemma-3-27b-it’s answer.
  1. User

    how was it proven that pi isn't rational?

  2. Assistant

    ⏩ Short answer
    The first rigorous proof that π is not a ratio of two whole numbers (i.e. that π is irrational) was given in 1761–‑1768 by Johann Heinrich Lambert. Lambert used a continued‑fraction expansion of tan x and showed that if x is a non‑zero rational number then tan x must be irrational; since tan (π/4)=1 is rational, π/4—and hence π—cannot be rational.

    Below you’ll find:

    1. A quick historical timeline
    2. An intuitive sketch of Lambert’s argument
    3. A modern “calculus proof” (Ivan Niven, 1947) that many students see today
    4. Why these ideas work and common misconceptions

    Feel free to skip to the section that interests you most!


    1. Historical timeline (lightning version)

    YearMathematicianKey ideaResult regarding π
    1748EulerWrites tan x as a continued fraction (no proof of irrationality yet)—
    1761 & 1768J. H. LambertUses Euler’s continued fraction to prove tan x is irrational for rational, non‑zero xFirst proof π is irrational
    1794LegendreStrengthens Lambert’s work, shows π² is also irrationalπ² irrational
    1882LindemannExtends Hermite’s method for e, proves π is transcendental (≠ root of any polynomial with integer coefficients)π transcendental
    1947Ivan NivenPublishes a short integral‑based proof of π’s irrationality—now a textbook favoriteSimpler proof

    2. Sketch of Lambert’s original proof

    1. Euler had shown (1748) the stunning identity

    tan x = x
    ─────────────
    1 − x²
    ─────────
    3 − x²
    ─────────
    5 − x²
    ──── ⋱

    which converges for all real x with |x| < π/2.

    1. Lambert proved the following lemma.

    Lemma
    If x ≠ 0 is a rational number, then the value of that infinite continued fraction must be irrational.

    Idea of the proof
    • Truncate the continued fraction after k steps to get a rational approximation A_k/B_k.
    • Show that the error |tan x − A_k/B_k| is smaller than 1/B_k².
    • But a famous result about continued fractions says: if a fraction A/B approximates a number that closely, the number cannot itself be rational.

    1. Plug in x = π/4. We know tan (π/4) = 1, which is obviously rational.
      By the lemma, if tan (π/4) is rational, then x = π/4 cannot be rational.
      Therefore π is irrational.

    That’s the heart of Lambert’s argument—short and elegant once the machinery of continued fractions is accepted.


    3. A modern integral‑based proof (Ivan Niven, 1947)

    If you’re comfortable with basic calculus, this proof is usually easier to digest than the continued‑fraction approach.

    Suppose, toward a contradiction, that

    π = a / b     (a, b ∈ ℕ, in lowest terms).
    

    Choose any positive integer n. Define the polynomial

    f_n(x) = (x^n (π − x)^n) / n!          (0 ≤ x ≤ π)
    

    Key observations

    1. f_n and all its first 2n derivatives are integers at x = 0 and x = π.
    2. For all x in (0, π) we have 0 < f_n(x) < π^{2n}/n!.

    Consider the integral

    I_n = ∫₀^{π} f_n(x) sin x dx      (> 0, because integrand ≥ 0 except endpoints)
    

    Step A – Show I_n is an integer
    Integrate by parts repeatedly (2n times). Each integration swaps derivatives between f_n and sin x. Because f_n and its derivatives at the endpoints are integers and sin (0)=sin (π)=0, one finds that each term produced is an integer. Hence I_n itself is an integer.

    Step B – Show I_n can be made arbitrarily small
    Using |sin x| ≤ 1 we get

    0 < I_n < ∫₀^{π} f_n(x) dx = π^{2n+1} / (n! (2n+1))
    

    As n→∞ the factorial in the denominator grows faster than the power in the numerator, so the right‑hand side → 0. Therefore 0 < I_n < 1 for sufficiently large n.

    But I_n is simultaneously
    • a positive integer (from Step A) and
    • strictly between 0 and 1 (from Step B).

    That is impossible. The contradiction means our original assumption “π is rational” is false. Hence π is irrational.

    (Notice how factorial growth is the secret weapon 🌱.)


    4. Why these proofs work & common pitfalls

    • No proof of π’s irrationality relies on memorizing π’s decimal expansion; we never need to “look at the digits.”
    • Lambert’s continued‑fraction argument ties rational inputs to irrational outputs of transcendental functions (tan x).
    • Niven’s proof combines three ingredients: (i) tailor‑made polynomials with integer boundary data, (ii) the sine function vanishing at 0 and π, (iii) factorial growth beating powers.
    • Transcendence (proved by Lindemann) is strictly stronger than irrationality: every transcendental number is irrational, but not vice‑versa.


    🎯 Bottom line
    Lambert’s 18th‑century continued‑fraction technique was the first to clinch π’s irrationality; later approaches (Legendre, Niven, etc.) polished the argument with integrals and factorials. Today, both the historical proof and Niven’s short calculus proof are standard references showing once and for all that π can never be written as a/b with integers a, b.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.