Pell's equation is the deceptively simple Diophantine equation

for a non-square positive integer D, sought in whole numbers x, y. At a glance it looks tame — maybe a handful of small solutions, if any. In reality it has infinitely many solutions for every valid D, and the very smallest one can already involve numbers with seven digits or more. This one equation was studied seriously by ancient Indian and Greek mathematicians well over a thousand years before anyone named Pell was born — and the name that stuck to it is, famously, a historical mix-up.

Why should anyone care about a single equation with two unknowns? Because it turns out to be the key that unlocks the structure of an entire family of number systems — and because chasing down its solutions by hand, long before calculators, was itself a serious mathematical sport for over a thousand years across three continents.

Worked example: finding the smallest solution for D = 2

Take the simplest interesting case, D = 2:

The most direct approach is plain trial: try y = 1, 2, 3, \ldots and check whether 2y^2 + 1 happens to be a perfect square.

• y = 1: 2(1)^2 + 1 = 3 — not a square.
• y = 2: 2(2)^2 + 1 = 9 = 3^2 — a square! So x = 3.

Check it: 3^2 - 2(2)^2 = 9 - 8 = 1. It works. The pair (x, y) = (3, 2) is the fundamental solution — the smallest nontrivial one — for D = 2. Run the same trial search yourself below.

One solution gives them all

Once you have the fundamental solution (x_1, y_1), you never need to search by trial again — every other solution is generated automatically by taking powers in the number system \mathbb{Z}[\sqrt D]:

For D = 2, squaring 3 + 2\sqrt 2 gives

so the next solution is (17, 12). Check it: 17^2 - 2(12)^2 = 289 - 288 = 1. Cubing gives (99, 70), and the solutions keep marching off to infinity, one power at a time, each one bigger than the last.

Finding the first solution systematically

Trial and error works fine for D = 2, but it becomes hopeless for larger D. The systematic route is through the periodic continued fraction of \sqrt{D}. For D = 2,

a purely periodic pattern after the first term. The convergent just before the pattern completes one full period — here, 1 + \tfrac{1}{2} = \tfrac{3}{2} — delivers the fundamental solution directly: numerator and denominator are exactly x_1 = 3 and y_1 = 2, matching the trial search above. For every non-square D, this method is guaranteed to terminate and always finds the fundamental solution — no guessing required, however large D is.

A bonus: astonishingly good square-root approximations

Rearranging x^2 - Dy^2 = 1 as x^2 = Dy^2 + 1 shows that x/y must be very close to \sqrt{D} whenever y is reasonably large — the "+1" barely matters once Dy^2 is big. Pell's solutions, it turns out, are exactly the best possible whole-number fraction approximations to \sqrt{D} for their size, which is no coincidence: they are the very continued-fraction convergents used to find them.

Watch the fractions from the D = 2 solutions above race towards \sqrt{2} \approx 1.41421356\ldots:

• 3/2 = 1.5
• 17/12 = 1.41\overline{6}
• 99/70 = 1.4142857\ldots

Each successive Pell solution roughly squares the accuracy of the last — a handful of digits in x and y buys many more correct digits of \sqrt{D} than you might expect. Before electronic calculators existed, this was a genuinely practical way to compute square roots to high precision by hand.

When the numbers explode

The continued-fraction method is essential because brute-force search is not merely slow for some values of D — it is completely hopeless. For D = 61, an innocent-looking case, the smallest solution is already

Trying every y from 1 up to 226{,}153{,}980 by hand is obviously never going to happen; even a modern computer doing blind trial would grind for a long time. Only the continued-fraction shortcut makes this tractable at all — see the vignette below for why Fermat chose this exact case to embarrass his rivals.

Why it matters

Pell's equation is the prototype for understanding units in rings of algebraic integers: its solutions are the units of \mathbb{Z}[\sqrt D] (the elements with a multiplicative inverse in that ring), and the fact that they form an infinite family generated by repeated powers of one fundamental unit is a special case of Dirichlet's much deeper unit theorem for general number fields. A two-thousand-year-old puzzle about squares and non-squares turns out to be a perfect first doorway into the structure of modern algebraic number theory.

There's also a tidy theoretical reason the method above always works: a theorem of Lagrange guarantees that the continued fraction of any irrational square root \sqrt{D} is eventually periodic — never a one-off jumble of digits, always a pattern that repeats forever once it settles in. That guaranteed periodicity is exactly what makes the continued-fraction method reliable for every non-square D whatsoever, not just tidy small cases like D=2.