Draw a square one unit on each side. Its area is one square unit. Now ask for a square with twice that area. Doubling the side seems an inviting first move, until the result occupies four square units. The extra area has arrived faster than expected.
The required side is the original square's diagonal. That length can be constructed exactly in Euclidean geometry, even though no fraction of whole numbers expresses its ratio to the original side. A modest drawing has opened two questions: how do we establish a shape's properties, and what kinds of numbers do those properties require?
First-principles reasoning begins by making the starting commitments explicit. Definitions identify the objects; assumptions specify the relationships available; proof carries those relationships into a conclusion. The doubling square lets us follow the whole exchange without mistaking the appearance of a diagram for its authority.
Naming a square does not construct one
Euclid's Book I defines a square through equal sides and right angles. That tells us what qualifies. Proposition 46 then supplies a construction on a given segment and argues that the resulting figure has those properties. The definition and the construction have different jobs: one states the requirements; the other shows how to meet them within the geometric framework. Euclid, definition 22 and proposition 46.
We can make a modern coordinate version of the distinction. In a flat Euclidean plane, choose four points: (0, 0), (1, 0), (1, 1) and (0, 1). Join them in that order. The horizontal sides differ by one in their first coordinate and zero in their second; the vertical sides do the reverse. All four lengths are one. The neighboring horizontal and vertical directions are perpendicular. The coordinates therefore describe a square.
We have used assumptions along the way. The coordinates belong to a Euclidean plane with the usual distance rule and perpendicular axes. Merely calling the axes x and y would not establish those relationships. Stating the framework makes the conclusion inspectable: another reader can follow which property licenses each step.
Suppose we instead use (0, 0), (2, 0), (2, 1) and (0, 1). The angles remain right, but two sides have length two and two have length one. This is a rectangle, not a square. The counterexample exposes the missing condition in the claim that four right angles suffice. A definition is useful partly because it tells us which tempting shortcuts fail.
Doubling area without doubling every length
For a square of side s, area is s × s, or s². If we multiply the side by a factor k, area becomes (ks)² = k²s². This original calculation explains why doubling the side quadrupled the area: the same factor enters twice.
To double area, we need k² = 2. The positive solution is √2, the number whose square is two. For the unit square, the distance between (0, 0) and (1, 1) is √(1² + 1²) = √2. A square built on that diagonal has area (√2)² = 2.
The construction does not require a decimal approximation. A calculator's 1.41421356 is useful for numerical work, but the geometric relationship was settled before those digits appeared. A decimal rounded to eight places and the exact expression √2 are different descriptions with different guarantees.
There is another way to see the area relationship. Join the midpoints of the four sides of a square. The new figure has four equal sides and right angles, and the four corner triangles together occupy half the original square. The inner square therefore occupies the other half. Reverse the relationship: a square whose side is the inner square's diagonal has twice the inner square's area. The diagram helps organize the pieces; the equal lengths, angles and areas justify the result.
Why no fraction finishes the diagonal
Could √2 simply be a fraction we have not found yet? We can answer with a complete contradiction argument. Here the numbers are positive integers, and the ratio is expressed in lowest terms—numerator and denominator share no factor greater than one.
Assume √2 = p/q for such integers p and q. Squaring gives 2 = p²/q², so p² = 2q². Thus p² is even.
An odd integer has the form 2n + 1. Its square is 4n² + 4n + 1, which is odd. An integer with an even square must therefore itself be even. Write p = 2m.
Substituting yields 4m² = 2q², or q² = 2m². The same argument makes q even. Both p and q are divisible by two, contradicting the starting requirement that the fraction be in lowest terms. No such fraction exists. The diagonal-to-side ratio is irrational.
This proof uses familiar facts about integers as well as the geometric relationship. It is a modern explanation presented here, not a claim to reproduce a particular ancient discovery or Euclid's construction proof. No story about the discoverer's fate is needed to make the result remarkable.
Fractions can still approximate the length. Seven-fifths is 1.4; its square is 1.96, slightly below two. Ninety-nine-seventieths is about 1.414286; its square is 9801/4900, slightly above two. Closer approximations improve a calculation without becoming an exact ratio. Searching through a large list of fractions could illustrate this behavior, but that finite search would not prove that every possible fraction fails. The contradiction argument reaches all positive integer pairs at once.
What follows when the starting point changes
A proof makes a conditional commitment: given these premises and valid steps, this conclusion follows. The unit square's diagonal tells us something exact inside the chosen framework. Whether a physical object matches that framework well enough is another question, developed in Points, Lines and the Work of Measurement.
Even within mathematics, changing the question changes what follows. A rectangle of sides one and two has diagonal √5. Its diagonal would produce a square of area five, while the rectangle's own area is two. The doubling result depended on equal sides. Removing that premise does not weaken the original proof; it reveals its proper scope.
The diagonal remains available as an exact construction while escaping every whole-number fraction. That is the reward of following a small problem carefully. First principles can force us to enlarge the language in which an answer is expressed.
Loading comments…