You have been told your whole school life that \pi is irrational. Its decimals never terminate and never repeat. But how does anyone actually know that? A decimal expansion can only be checked one digit at a time — you could compute a trillion digits and still, in principle, worry that the trillion-and-first digit starts a repeating block. Irrationality is a statement about all the digits, forever. You cannot verify it by computation. You have to prove it.
The short answer: somebody did — and the cleanest modern proof fits on a single page. This article walks you through what "prove \pi is irrational" actually means, who did it first, and the strategy of that proof.
What "prove \pi is irrational" actually asks for
Recall the reflex from Proof-by-Contradiction for Irrationality. To prove a number x is irrational, you do not try to examine its decimal expansion. You assume the opposite — that x = p/q for some integers — and derive a contradiction. Then you conclude that no such p, q can exist, so x is not rational.
For \sqrt{2}, the contradiction came out in a few lines using divisibility. For \pi, divisibility alone does not work — \pi is not the root of any integer polynomial, so the "square both sides" trick has nothing to square. The machinery turns out to be calculus: integration by parts, and a carefully chosen polynomial whose integral you can bound two different ways. The two bounds contradict each other if \pi were rational. That is the whole game.
The history: Lambert 1761, Niven 1947
Johann Heinrich Lambert proved \pi irrational in 1761 using continued fractions. He showed that the continued-fraction expansion of \tan(x), when x is a nonzero rational, is infinite and never terminates — and therefore \tan(x) is irrational for every nonzero rational x. Since \tan(\pi/4) = 1, which is rational, \pi/4 cannot be rational — so \pi cannot be rational either. Lambert's proof is beautiful but technical; it requires a detour into the theory of continued fractions that most school students never see.
Almost two centuries later, in 1947, Ivan Niven published a proof that is famously short — about one page in the Bulletin of the American Mathematical Society (Niven 1947). It uses only polynomial algebra, integration by parts, and the idea that a positive integer cannot lie strictly between 0 and 1. No continued fractions, no complex analysis. This is the proof you usually see quoted today.
Niven's strategy, in plain words
The whole proof rests on a single clever move: Niven defines a specific polynomial, picks a specific integral, and shows that if \pi were rational the integral would have to be both (a) a positive integer and (b) strictly less than 1. Since there is no positive integer less than 1, the assumption must have been wrong.
Here is the setup. Suppose, for contradiction, that \pi = p/q where p and q are positive integers. For each positive integer n, define the polynomial
This polynomial looks strange, but it is engineered to have two properties that matter:
-
f has a "symmetric" shape on the interval [0, \pi]. Since \pi = p/q, you have p - q\pi = 0, so f(\pi) = 0. Also f(0) = 0. Between 0 and \pi, f is a small positive bump, and the factor of 1/n! makes it very small for large n.
-
f and all its derivatives take integer values at x = 0 and x = \pi. This is not obvious — it is the heart of the proof. Expanding (p - qx)^n and tracking what happens when you differentiate f(x) k times and evaluate at 0, you find that every such value f^{(k)}(0) is an integer (the n! in the denominator gets cancelled out). The same holds at x = \pi by the symmetry f(\pi - x) = f(x).
Now Niven considers the integral
Bound 1: I_n is a positive integer. Using repeated integration by parts (integrating \sin x and differentiating f), you pick up boundary terms at x = 0 and x = \pi of the form f^{(k)}(0) and f^{(k)}(\pi). By property (2) above, every boundary term is an integer. After enough integrations by parts, the remaining integral has an even-order derivative of f multiplied by \pm \cos x or \pm \sin x, and the f^{(k)} eventually hits zero because f is a polynomial of finite degree. The total is therefore a finite integer sum: I_n \in \mathbb{Z}. And I_n > 0 because f(x) \sin x > 0 strictly for 0 < x < \pi. So I_n is a positive integer — meaning I_n \ge 1.
Bound 2: I_n < 1 for large n. Bound f crudely on [0, \pi]. On that interval, x \le \pi and |p - qx| \le p, so |f(x)| \le \pi^n p^n / n!. Since |\sin x| \le 1, you get
The expression \pi^{n+1} p^n / n! goes to 0 as n \to \infty — factorials grow faster than any exponential. So for n large enough, I_n < 1.
Contradiction. For that large n, I_n is simultaneously a positive integer (so I_n \ge 1) and strictly less than 1. No such number exists. Therefore the original assumption \pi = p/q must have been wrong. \pi is irrational. \blacksquare
Why the polynomial f(x) = x^n(p - qx)^n / n!? The x^n makes f and its first n-1 derivatives vanish at x = 0. The (p - qx)^n makes f vanish at x = \pi = p/q. The n! cancels the coefficients that come out when you differentiate, leaving integers. It is not a polynomial anyone would stumble on by accident — it was reverse-engineered to make the integration-by-parts bookkeeping come out clean.
What you are taking on trust
The full write-up asks you to verify three technical steps: every derivative of f at 0 is an integer, the integration-by-parts of I_n produces only integer terms, and \pi^{n+1} p^n / n! \to 0. The first two are tedious bookkeeping; the third follows from the fact that n! outgrows any exponential. None of it needs more than Class 12 calculus, and Niven's original 1947 paper is freely available — see the references below.
A stronger fact: \pi is transcendental
Irrational means: \pi is not a ratio of integers. Transcendental means something stronger: \pi is not the root of any polynomial with integer coefficients. That is, there is no polynomial
with integer a_i (not all zero) that has \pi as a solution. Every rational number is a root of a linear integer polynomial (qx - p = 0 for p/q). Some irrationals, like \sqrt{2}, are roots of integer polynomials — \sqrt{2} satisfies x^2 - 2 = 0. These irrationals are called algebraic. Irrationals that are not algebraic — like \pi and e — are called transcendental.
Ferdinand von Lindemann proved in 1882 that \pi is transcendental (Lindemann 1882). His proof is considerably harder than Niven's, building on Hermite's 1873 proof that e is transcendental. Lindemann's result settles the ancient problem of squaring the circle: you cannot construct, with compass and straightedge alone, a square whose area equals that of a given circle, because compass-and-straightedge constructions produce only numbers algebraic over the rationals. Transcendence implies irrationality, but not the other way around — so Lindemann's theorem is a strictly stronger version of Lambert's, and a much harder one to prove.
The takeaway
- \pi is irrational. This was proved by Lambert in 1761 using continued fractions, and re-proved more accessibly by Niven in 1947 using a one-page integration-by-parts argument.
- The proof does not work by examining decimals. It works by assuming \pi = p/q, building a specific polynomial and integral, and showing the integral is forced to be both a positive integer and less than 1 — impossible.
- The strategy generalises. The same polynomial trick proves e^r is irrational for every nonzero rational r, and underlies the transcendence proofs of e and \pi.
- \pi is also transcendental (Lindemann, 1882) — a stronger statement that settles the impossibility of squaring the circle with compass and straightedge.
You do not need to memorise the proof. You need to know it exists, that it is short once you have first-year calculus, and that it rests on a very specific kind of cleverness — not on any argument about digits.
This satellite sits inside Real Numbers — Properties.
References
- Niven, I. (1947). A simple proof that \pi is irrational. Bulletin of the American Mathematical Society, 53(6), 509. PDF
- Lindemann, F. (1882). Über die Zahl \pi. Mathematische Annalen, 20, 213–225. Wikipedia summary
- Lambert, J. H. (1761). Mémoire sur quelques propriétés remarquables des quantités transcendantes circulaires et logarithmiques. Mémoires de l'Académie des sciences de Berlin. Wikipedia: Proof that π is irrational