The Euler Characteristic Counts Holes

An old formula about polyhedra, seen with new eyes

Pick up a cube. It has 8 corners, 12 edges, and 6 flat faces. Add those with the alternating pattern vertices minus edges plus faces: 8 - 12 + 6 = 2. Now try a tetrahedron, the simplest pyramid: 4 vertices, 6 edges, 4 faces, giving 4 - 6 + 4 = 2 again. A soccer-ball-style shape, an octahedron, a dodecahedron, a squashed box, a many-sided gem — every convex polyhedron you can name returns the same 2. This stubborn answer is Euler's polyhedron formula, V - E + F = 2, noticed by Euler in 1750. What is going on that the count refuses to depend on which solid you chose?

Here is the shift that the rubber-sheet view from the last guide unlocks. The surface of every one of those solids is, topologically, a sphere — inflate any convex polyhedron and it rounds out into a ball's skin, no tearing required, so all of them are homeomorphic to one another. The number 2 is not really a fact about cubes or pyramids; it is a fact about the sphere, dressed up in whatever vertices, edges, and faces you happened to draw on it. Once you see V - E + F as a property of the underlying surface rather than the particular solid, you can ask the daring question: what does this count give for a doughnut?

Defining chi, and why a doughnut gives zero

Promote the count to a name. For any way of dividing a surface into vertices, edges, and faces, define the Euler characteristic chi = V - E + F. For a sphere chi = 2, as we saw. To find it for a torus — the surface of a doughnut — we just need one honest mesh on it. Picture the torus as a square sheet whose left edge is glued to its right and whose top is glued to its bottom; that gluing is exactly what bends a flat sheet into a tube and then a ring. Drawing the simplest grid on that glued square and counting carefully gives V = 1, E = 2, F = 1, so chi = 1 - 2 + 1 = 0.

        a                 surface       chi = V - E + F
   +---------+
   |         |            sphere            2
  b|  torus  |b           torus             0
   |  (glued)|            2-hole surface    -2
   +---------+            g-hole surface   2 - 2g
        a

  square with left~right (b) and top~bottom (a) glued:
  V = 1   E = 2 (the two glued sides a and b)   F = 1
  chi = 1 - 2 + 1 = 0

So the sphere scores 2 and the torus scores 0 — they are genuinely different surfaces, and chi can tell. Add a second hole, making a pretzel-shaped surface, and the same kind of count gives chi = -2; a three-holed surface gives -4. The pattern is exact: a closed orientable surface with g holes has chi = 2 - 2g. The number of holes here is what topologists call the genus g, the star of the very next guide. For now, read the relationship backwards: knowing chi, you recover g = (2 - chi) / 2. The Euler characteristic is, quite literally, a hole-counter wearing the disguise of an arithmetic of dots, lines, and patches.

Why the count never changes

The whole power of chi rests on one promise: that V - E + F does not depend on how you carve up the surface. If two people mesh the same torus completely differently, they must still both land on 0, or the number means nothing. To make this trustworthy we fix a clean way to carve — a triangulation, which just means covering the surface with triangular faces meeting edge-to-edge, each triangle an honest disk. Now watch what any single change to a triangulation does to the count.

1. Add a new vertex in the middle of an existing edge. That splits the one edge into two (E goes up by 1) and adds one vertex (V goes up by 1). The change to V - E + F is +1 - 1 = 0. No change.
2. Add a new vertex inside a triangle and connect it to all three corners. That adds 1 vertex, 3 edges, and turns 1 face into 3 (so F goes up by 2). The change is +1 - 3 + 2 = 0. No change.
3. Draw a new edge across a face to split it in two. That adds 1 edge and 1 face: change +0 - 1 + 1 = 0. No change either.
4. Since any two triangulations of the same surface can be turned into each other by a finite sequence of these refinement moves (and their reverses), and not one move budges V - E + F, both triangulations must give the identical number.

Each elementary move is engineered to leave the alternating sum untouched — a vertex always arrives paired with just enough edges and faces to cancel itself out. That is the real reason a cube and a tetrahedron, meshes of the very same sphere, both reported 2: you can refine one into the other without the count ever flinching. So chi is a property of the surface itself, not of any drawing on it — it is a genuine topological invariant, the same for any two surfaces that are homeomorphic.

Putting the invariant to work

An invariant earns its keep by telling shapes apart. Suppose someone hands you two closed surfaces and swears they are secretly the same — that one can be stretched into the other without cutting. Triangulate each, compute chi for both. If the numbers differ, the claim is dead on arrival: a sphere (chi = 2) can never be homeomorphic to a torus (chi = 0), and now you have a one-line proof, not just a feeling that a ball and a doughnut are different. This is the everyday job of chi across a topological space: a cheap, computable certificate of difference.

There is also a lovely arithmetic of building surfaces. To glue two surfaces into one, cut a small disk out of each and stitch them along the two new boundary circles — the connected sum, written S_1 # S_2. The Euler characteristics combine by a simple rule: chi(S_1 # S_2) = chi(S_1) + chi(S_2) - 2, because each removed disk costs a face and the seam merges shared vertices and edges in a way that subtracts a fixed 2. Glue two tori: 0 + 0 - 2 = -2, the two-holed surface, exactly as our hole-count predicted. The bookkeeping of chi and the bookkeeping of holes are the same ledger.

What chi is, honestly, and where it points next

Let me be straight about what we have and have not done. We have defined chi and shown by the refinement moves that it is unchanged under the local edits relating two triangulations of one surface. The deeper facts — that every reasonable surface can be triangulated at all, and that any two triangulations are connected by a finite chain of those moves — are real theorems whose careful proofs belong to a topology course proper, not to this rung. We have used them honestly as the load-bearing facts they are, while being clear that their full justification lives above us. What you should trust completely is the meaning and the computation: chi = V - E + F, the same for any mesh, counting holes via 2 - 2g.

And resist over-reading the number. A surface is not nothing but its Euler characteristic — chi throws away an enormous amount, keeping only one integer. Two surfaces with the same chi can still differ in orientability, as the torus and Klein bottle just showed, and surfaces with boundary or with more than two dimensions need richer invariants. chi is the first and friendliest of a whole family of topological numbers; it is the doorway, not the whole house.

Even so, this one integer is the hinge of the whole rung. The very next guide takes chi together with orientability and uses them to classify every closed surface — a complete catalogue of shapes from a single number and a yes/no. Climb higher still and chi reappears in the most unexpected place: in the surfaces rung you may meet the Gauss-Bonnet theorem, where the total curvature of a closed surface, a purely geometric integral, turns out to equal 2 pi times this same chi. The little arithmetic V - E + F that started with a cube on a table becomes the bridge on which geometry and topology finally meet.