17  Euclidean Geometry

You compute in \(\mathbb{R}^n\) effortlessly — distances, dot products, rotations. “Euclidean geometry” historically meant something structurally different: a system built from a handful of statements about points, lines, and circles, with no coordinates and no numbers anywhere in sight. This chapter is about both versions, and the surprisingly modern idea that makes them equivalent.

17.1 The axiomatic version, and the postulate that took two thousand years

Euclid’s Elements (Alexandria, roughly 300 BCE) builds geometry from five postulates: essentially, that a line can be drawn between any two points, that any line segment can be extended indefinitely, that a circle can be drawn with any center and radius, that all right angles are equal, and — the awkward one — that through a point not on a line, exactly one line can be drawn parallel to it. The first four are simple and uncontroversial; the fifth is noticeably more complicated to even state, and for roughly two thousand years mathematicians tried to prove it from the other four, convinced it had to be a consequence rather than an independent assumption.

It doesn’t. In the 1820s–30s, János Bolyai and Nikolai Lobachevsky (independently) showed that replacing the parallel postulate with its negation produces an entirely consistent geometry — hyperbolic geometry — where triangles’ angles sum to less than 180°. (Elliptic geometries, where the postulate fails the other way, are the other option.) This is one of the sharpest illustrations on this whole site of Axioms and Definitions’s central claim: axioms are chosen, not discovered facts about a single true geometry. Euclidean geometry isn’t “the” geometry — it’s a geometry, the one you get by choosing all five postulates, no more privileged in principle than the alternative you get by choosing differently.

17.2 The coordinate version

René Descartes (and, independently, Pierre de Fermat) introduced analytic geometry around 1637: represent points as tuples of numbers and geometric relationships as algebraic equations — a circle of radius \(r\) becomes \(x^2+y^2=r^2\), a line becomes a linear equation. This is the bridge between geometry and algebra you’ve used your entire computational life without necessarily seeing it named, and it lets Euclidean geometry be redefined entirely structurally, with no synthetic postulates at all: Euclidean space is an affine space whose vector space of displacements carries an inner product. That one sentence packages points (affine space), distances and angles (the induced norm and the Cauchy–Schwarz-guaranteed angle), and everything else Euclid’s postulates were reaching for — built entirely from structures this site has already defined.

This isn’t a computational shortcut standing in for “real” geometry; for the ordinary flat case, it’s a fully rigorous alternative axiomatization, provably equivalent to Euclid’s synthetic one. It’s another instance of the map’s historical-versus-formal split: Euclid’s postulates came roughly two thousand years before Descartes’s coordinates, and Descartes’s coordinates came roughly two hundred years before the vector-space and inner-product axioms modern geometry actually rests on today.

17.3 If you’ve done linear algebra or physics

You already know:

• The dot product \(u \cdot v\) — the standard inner product on \(\mathbb{R}^n\).
• The distance formula \(\sqrt{\sum_i (x_i-y_i)^2}\) — the metric that inner product induces, exactly as Inner Product Spaces derives it.
• A rotation matrix — a linear map that happens to preserve that inner product (an orthogonal transformation), which is precisely why rotating a shape never changes distances or angles within it.

None of these were arbitrary computational rules; each is a specific instance of a structure this site has now named.

17.4 A first pass at what comes after flat

Once geometry is redefined structurally rather than synthetically, a natural question opens up: must a “geometric” space look like \(\mathbb{R}^n\) everywhere, or only nearby each point? The surface of a sphere is curved globally — it’s not an affine space, no single flat coordinate grid covers all of it — but zoom into any small patch and it looks just like a flat plane. A manifold is, informally, exactly this: a topological space that locally resembles \(\mathbb{R}^n\) near every point, via a patchwork of coordinate charts, even when it doesn’t globally. Non-Euclidean geometries turn out to be one important source of manifolds — spaces curved in the large that are still locally Euclidean. This is genuinely a first pass: manifolds, and the differential geometry built on them, get only a preview later in this site — see Advanced Topics — and the one-paragraph version above is meant only to attach a name and a mental picture to the idea, not to define it rigorously.

17.5 Where this leads

This chapter closes the structural arc this site has built from What Is a Mathematical Structure? onward — sets, groups, rings and fields, vector spaces, inner product spaces, metric spaces, topological spaces, and affine spaces, assembled here into the geometry you already knew computationally. From here, this site’s table of contents continues into Limits and Continuity — real analysis, built directly on the continuity Topological Spaces already defined. Advanced Topics previews differential geometry (making the manifold sketch above precise), and — already foreshadowed at the end of Linear Algebra — Machine Learning picks up the high-dimensional geometry that shows up constantly there.