The standard proof that \sqrt{2} is irrational ends with a contradiction: "p and q are both even, but we assumed p/q was in lowest terms." That is the usual conclusion. But hidden inside the algebra is a deeper, more dramatic structure — an infinite descent. The same argument that produces one factor of 2 can be rerun to produce another, and another, forever. The factor of 2 keeps coming out and never stops. Positive integers simply cannot sustain that; there is a floor they hit.
This animation exposes the descent directly. You assume \sqrt{2} = p/q for some positive integers p and q — and you just watch them shrink by factors of 2 indefinitely. The impossibility is visual, not just algebraic.
The setup
Assume \sqrt{2} is rational. Then there exist positive integers p, q with \sqrt{2} = p/q, i.e., p^2 = 2q^2.
Crucially: don't bother assuming lowest terms. The descent argument will build that contradiction for you from scratch. You just need some positive integer pair (p, q) satisfying p^2 = 2q^2.
Now run the same argument from the classical proof, but instead of stopping at "p and q are both even," divide them both by 2 and repeat.
The descent, step by step
Round 1. Start with (p_0, q_0) positive integers with p_0^2 = 2q_0^2. From p_0^2 = 2q_0^2, p_0^2 is even, so p_0 is even: p_0 = 2p_1. Substitute: 4p_1^2 = 2q_0^2 \Rightarrow q_0^2 = 2p_1^2, so q_0 is even: q_0 = 2q_1. New pair: (p_1, q_1) = (p_0/2, q_0/2), both positive integers, with p_1^2 = 2q_1^2.
Round 2. Same argument on (p_1, q_1): p_1 = 2p_2 and q_1 = 2q_2, and (p_2, q_2) = (p_1/2, q_1/2) is a smaller positive-integer pair satisfying the same equation.
Round n. After n rounds, (p_n, q_n) = (p_0/2^n, q_0/2^n), and both are still positive integers satisfying p_n^2 = 2q_n^2.
But p_0 is a fixed positive integer. You cannot divide a positive integer by 2 infinitely many times and still have a positive integer — after \log_2 p_0 divisions, you fall below 1. Yet the descent says you can. Contradiction.
The interactive descent
Hit Descend to step through one round at a time. On each click, the current pair (p, q) shrinks — both values halve — and you see the shrinking pair animated down the "staircase." Each staircase row is one round; the dot representing (p,q) literally walks down and gets smaller. When a value would fall below 1 it turns red and locks — that is the impossibility made visible.
Why descent is the sharper form of the proof: the "lowest terms" version of the proof uses a single contradiction at the end. The descent version removes the lowest-terms assumption and instead uses the structural impossibility of infinitely many divisions — every positive integer has a finite number of factors of 2. This form is due in spirit to Fermat, and it is one of the cleanest kinds of impossibility argument in number theory.
Why the descent is sharper than "lowest terms"
The lowest-terms proof has a moment where it imports an extra assumption: "we may assume p/q is in lowest terms." This is fine — any fraction can be reduced — but it feels like a trick the proof is using. Students sometimes ask "but what if I do not reduce?" The descent form answers cleanly: you do not need to reduce. The impossibility is intrinsic to the equation p^2 = 2q^2 in positive integers.
Pictorially, "lowest terms" is one stop sign planted in the road. Infinite descent is an endless road with no stop sign — the traveller never arrives, which is the impossibility. Same destination, different argument.
Why both forms are valid: the lowest-terms argument extracts the contradiction at the first halving (p and q both even while supposedly coprime). The descent argument extracts it as a structural limit (you cannot halve positive integers forever). Either argument refutes the assumption. Fermat and many later number theorists preferred the descent form because it generalises cleanly to other diophantine equations (including Fermat's own proof of the n = 4 case of his last theorem).
The Indian connection
The root of 2 — and other surds — were studied extensively in the Sulba Sutras (roughly 800 BCE), ancient Indian geometric treatises. Baudhayana gave an astonishingly accurate rational approximation: \sqrt{2} \approx 1 + \frac{1}{3} + \frac{1}{3 \cdot 4} - \frac{1}{3 \cdot 4 \cdot 34} \approx 1.41421568\ldots — correct to five decimal places. The Sulba mathematicians did not prove irrationality in the modern sense, but they knew the approximation never quite hit the mark, no matter how many terms you took. The descent argument, centuries later, is the formal statement of what they already felt: the diagonal of a square does not share a measuring unit with its side, however small you take the unit.
A subtle point about the starting pair
In the animation, we start with p_0 some power of 2 and q_0 = p_0/\sqrt{2} (which is of course not an integer) — that is just for illustration. The real descent argument says: assume there exists any positive-integer pair satisfying p^2 = 2q^2. Call it (p_0, q_0). Whatever it is, the descent halves it each round. Starting values do not matter — only the existence does. Non-existence of a descent terminus is the non-existence of a starting pair is the non-existence of a rational \sqrt{2}.
This is a feature of descent proofs generally: they are often proofs by contradiction over a hypothetical starting point, with the contradiction built from the structure of positive integers rather than a specific algebraic clash.
The takeaway
Infinite descent makes the impossibility of \sqrt{2} being rational tangible: if it were rational, you would have an infinite halving chain of positive integers, and there is no such thing. The standard "lowest terms" proof is one snapshot of this chain — the first halving already exposes the contradiction when you have committed to lowest terms. Both proofs deliver the same verdict; the descent version simply shows you more of the scaffolding that makes the verdict inevitable.
Watching the descent also builds a general proof instinct: "if this equation had a solution, you could reduce the solution to a smaller one with the same structure — but positive integers run out." That instinct opens the door to an entire class of contradiction proofs.
Related: Proof by Contradiction · Proof by Contradiction for Irrationality · Tennenbaum's Visual Proof of Root-2 Irrationality · Two-World Split Screen — Contradiction Collapse