Appearance
3.1 — Euclid, and the Invention of Proof
The Egyptians knew that a triangle with sides 3, 4 and 5 has a right angle in it, and used knotted ropes to lay out square corners for buildings. The Babylonians had tables of such triples a thousand years before the Greeks. Both civilisations had geometry that worked.
What neither had was a reason. They knew that it worked, from measurement and long practice. Around 300 BCE in Alexandria, Euclid wrote thirteen books that asked a different question: can we start from a handful of things too obvious to argue about, and derive everything else, with no measurement at all?
The answer was yes, and the result — the Elements — became the second most printed book in Western history after the Bible, was still the standard school geometry text in 1900, and invented the idea that has defined mathematics ever since: you do not know something because you measured it; you know it because you proved it.
1. The structure Euclid invented
Every mathematical subject since has been organised the way Euclid organised geometry, so it is worth naming the parts.
Definitions fix what the words mean. "A point is that which has no part." "A line is breadthless length." These particular definitions are philosophically shaky — "that which has no part" is not much of a definition — and modern treatments simply leave point and line undefined, taking them as primitive terms whose behaviour is fixed by the axioms rather than by description.
Postulates and axioms are the statements assumed without proof. You must start somewhere; if every statement needed a prior justification you would never begin. The honest position is not that these are self-evidently true but that these are the assumptions being made, stated openly so anyone can check what rests on what.
Propositions are everything else, each proved from the definitions, the postulates, and the propositions already established. Nothing is used before it is proved. Chapter 1.1's rule about defining before use is Euclid's rule.
Euclid's five postulates:
- A straight line can be drawn between any two points.
- A straight line can be extended indefinitely.
- A circle can be drawn with any centre and any radius.
- All right angles are equal to one another.
- (The parallel postulate — see Section 4.)
The first three are really a description of what you can do with a straightedge and a compass, which is why "ruler and compass construction" is the standard toolkit of classical geometry. Notice there is no ruler with markings on it: you may draw a line through two points but not measure its length. That restriction is deliberate, and Section 5 shows what it costs.
2. What a proof actually is
A proof is a chain of statements, each one either an assumption or a consequence of earlier statements by a rule of logic, ending at the thing you wanted to show. That is all it is. The mystique around proofs comes from the fact that finding the chain requires creativity, but checking one requires only patience.
Take the simplest proposition worth proving: the angles of a triangle add to 180°.
You could measure a hundred triangles and find roughly 180° each time. That establishes nothing about the hundred-and-first, and your protractor has error in it anyway. Here is the proof.
Draw triangle ABC. Through the vertex A, draw a line parallel to the opposite side BC.
Now use a fact about parallel lines: when a line crosses two parallel lines, the alternate angles it forms are equal. This is proved earlier in Euclid and follows from the parallel postulate. The side AB crosses both parallels, so the angle it makes with the new line at A equals angle B. Similarly, the side AC gives an angle at A equal to angle C.
At the point A we now have three angles sitting side by side on a straight line: one equal to B, the original angle A, and one equal to C. Angles on a straight line add to 180°.
\therefore\; A + B + C = 180°
Done, once, for every triangle that has ever existed or ever will. No measurement, no exceptions, no error bars. That is what proof buys, and it is why mathematics is the only field where results do not expire.
Note what the proof depended on: parallel lines behaving as the fifth postulate says. Change that postulate and the theorem becomes false — which is exactly what happens in Section 4.
3. The pieces of the toolkit
A short vocabulary, since the rest of Part 3 uses it constantly.
Angles. An acute angle is less than 90°, a right angle is exactly 90°, an obtuse angle is between 90° and 180°, a reflex angle is more than 180°. Two angles are complementary if they sum to 90° and supplementary if they sum to 180°. The word "complementary" is worth remembering because Chapter 3.5's cosine is literally the "co-" of complementary: \cos\theta = \sin(90° - \theta).
Triangles by sides. Equilateral (three equal sides, hence three 60° angles), isosceles (two equal sides, hence two equal angles), scalene (all different).
Triangles by angles. Right-angled, acute (all angles acute), obtuse (one obtuse angle).
Quadrilaterals, and the hierarchy matters. A trapezium has at least one pair of parallel sides. A parallelogram has two pairs. A rhombus is a parallelogram with all sides equal. A rectangle is a parallelogram with right angles. A square is both a rhombus and a rectangle. So every square is a rectangle, and every rectangle is a parallelogram, but not the reverse — a classification structure identical to the class hierarchies of Volume I, 9.2.4.
Polygons. A closed figure of straight sides. Its interior angles sum to (n-2) \times 180°, and the reason is one sentence: any n-sided polygon can be cut into n-2 triangles by drawing diagonals from one vertex, and each triangle contributes 180°. So a pentagon gives 540°, and each angle of a regular pentagon is 108°.
4. The fifth postulate, and the crisis it caused
Euclid's fifth postulate is conspicuously unlike the others. In its usual modern phrasing (Playfair's axiom):
Through a point not on a given line, there is exactly one line parallel to the given line.
The first four are one-line statements about what you can draw. This one is long, awkward, and does not feel obvious in the same way. Euclid himself seems uneasy about it — he avoids using it for as long as he possibly can, getting through 28 propositions before he needs it.
For two thousand years, mathematicians tried to prove it from the other four, so it could be demoted from an assumption to a theorem. Every attempt failed, and many failed in the same way: the author would unknowingly assume something equivalent to the postulate and then derive it, which proves nothing.
In the 1820s and 30s, Bolyai in Hungary, Lobachevsky in Russia, and Gauss privately in Germany all reached the same astonishing conclusion. The postulate cannot be proved, because it is not forced. You can replace it with something else and get a geometry that is entirely consistent — different from Euclid's, and not wrong.
Hyperbolic geometry assumes infinitely many parallels through the point. Triangles have angles summing to less than 180°, and the deficit grows with the triangle's area. The surface of a saddle behaves this way.
Elliptic geometry assumes no parallels — every pair of lines meets. Triangles sum to more than 180°. The surface of a sphere behaves this way: take the equator and two lines of longitude, and you get a triangle with three right angles, summing to 270°.
The philosophical shock was enormous. Geometry had been the model of certain, necessary truth about physical space — Kant had built a philosophy on it. Now there were several geometries, all consistent, and which one describes actual space became an experimental question rather than a matter of pure reason.
The answer came in 1915. Einstein's general relativity says mass curves spacetime, so the geometry of the real universe is not Euclidean — it is curved, locally, by whatever mass is nearby. Volume IV, Chapter 6 covers it. The non-Euclidean geometry that looked like an idle exercise in changing an axiom turned out to be the geometry we actually live in.
And there is a homelier example: the Earth's surface is elliptic geometry, which is why the shortest flight from London to Tokyo goes over the Arctic, why that route looks absurdly curved on a flat map, and why no flat map of the world can get areas, angles and distances all right at once.
5. What you can and cannot construct
With straightedge and compass alone, the Greeks could bisect any angle, construct a perpendicular, copy a segment, build an equilateral triangle, a square, a regular pentagon and a regular 15-gon. Three problems defeated them completely:
Doubling the cube — construct a cube with twice the volume of a given one. Needs a length of \sqrt[3]{2}.
Trisecting a general angle — cut any angle into three equal parts.
Squaring the circle — construct a square with the same area as a given circle. Needs a length of \sqrt\pi.
Two thousand years of failure, and then the answers arrived from algebra, not geometry. A construction step can only intersect lines and circles, which means each new length satisfies at most a quadratic equation over the lengths you already have. So every constructible length is algebraic with a degree that is a power of two.
That settles all three. \sqrt[3]{2} has degree 3, which is not a power of two — Wantzel, 1837. Trisecting 60° requires solving a cubic, same argument, same paper. And \pi is transcendental, so \sqrt\pi satisfies no polynomial equation at all — Lindemann, 1882, using the result from Chapter 1.7.
The lesson is one that recurs throughout this book: a question that resists direct attack for centuries is often answered by translating it into a different language. Geometry could not answer these; algebra could, in a page.
6. The Platonic solids: a complete list, and a proof it is complete
A regular polyhedron is a solid whose faces are all the same regular polygon, with the same number meeting at every corner. How many are there?
Exactly five, and the argument fits in a paragraph.
At each corner, at least three faces must meet, and their angles must add to less than 360° — if they summed to exactly 360° the corner would lie flat, and more would not close up.
- Triangles (60° each): three, four or five fit (180°, 240°, 300°). Six give 360° — flat. That is the tetrahedron, octahedron, icosahedron.
- Squares (90°): only three fit (270°). Four give 360°. That is the cube.
- Pentagons (108°): only three fit (324°). That is the dodecahedron.
- Hexagons (120°): three already give 360°. Nothing works, and larger polygons are worse.
Five. Not five that we have found — five that can exist.
Euclid's Elements ends with this result. Thirteen books of build-up, and the last proposition is the proof that there are exactly five. He was making a point about where the whole edifice was heading.
You have held these: a six-sided die is a cube, and the twenty-sided die of tabletop games is an icosahedron. They are used precisely because regularity guarantees fairness — every face is geometrically identical, so no face can be favoured.
7. Where this shows up in your life
Every legal argument, every scientific paper, every code review. The Euclidean structure — state your assumptions, derive consequences, do not use what you have not established — is the template for careful reasoning of any kind. Volume I, 14.7 says the same about design documents.
Every flight path and every world map. Elliptic geometry, and the impossibility of flattening a sphere without distortion.
Every GPS fix. The satellites' clocks are corrected for both special and general relativity, which is non-Euclidean geometry applied to your phone. Volume III, 8.2 covers the mechanism.
Every CAD model and 3D game. Constructions, intersections, and the polygon mathematics of Section 3.
Every dice roll. Five solids, and the reason a fair die has one of those shapes.
Euclid's most famous single result is one he did not discover, and it is the most useful formula in all of geometry. The next chapter proves it — several times over.