Hairy ball theorem
A Socratic walk-through of the hairy ball theorem — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why must there always be at least one place on Earth where the wind is not blowing?
Here is a claim that sounds like weather trivia and is really a theorem: at any instant, somewhere on Earth the horizontal wind is exactly zero. Not "usually", not "on average" — always, and no measurement is needed to know it.
That should be suspicious. How could anyone know something about tomorrow's atmosphere without knowing anything about the atmosphere? Ask what is doing the work. It cannot be meteorology. So what is left?
Reasoning it through
REASONING #Strip the weather away and keep only the shape of the claim. At every point of a sphere we are given an arrow lying flat against the surface, and the arrows vary continuously — nearby points carry nearly parallel arrows of nearly equal length. The claim is that at least one arrow must have length zero.
Before believing it, try to break it. On a circle, the same setup is easy: let every arrow point clockwise along the circle. Continuous, tangent, never zero. So "closed curved surface" is not the obstruction. On a flat disc, comb every hair in the same direction — also easy. Whatever goes wrong on the sphere is not going wrong on either of those.
Now try the sphere honestly. Send every arrow due east: continuous everywhere except at the two poles, where "east" has no meaning and the field has no continuous value. Push the trouble around and it does not vanish — it moves. Comb everything toward the north pole and the arrows collide there; make a whorl and the centre of the whorl is calm. Each attempt leaves a defect somewhere. That is the shape of a conservation law: something is being conserved that our combing cannot destroy.
What is conserved? Around any small loop enclosing an isolated zero, the arrow rotates through some whole number of turns as you walk the loop once. That integer is the zero's index, and it cannot change under any continuous deformation — you cannot turn a whole number into a different whole number gradually. Poincaré and Hopf showed the accounting closes: on a compact surface, the indices of all the zeros of a continuous tangent field sum to the surface's Euler characteristic, a number belonging to the shape alone.
Can we check that number ourselves? Yes, and this is the step worth doing by hand. Euler's characteristic is vertices minus edges plus faces of any polyhedral division of the surface. Divide a sphere as a tetrahedron: 4 vertices, 6 edges, 4 faces, so 4 - 6 + 4 = 2. Divide a torus by gluing the opposite sides of a single square: the four corners all become one point, the two pairs of sides become two edges, and there is one face, so 1 - 2 + 1 = 0.
Now the conclusion is forced. If a field on the sphere had no zeros, the sum of its indices would be an empty sum, namely 0 — but that sum must equal 2. Contradiction, so a zero exists. On the torus the target is 0, and no contradiction arises; indeed the explicit field "run around the tube" never vanishes. The circle likewise has characteristic 0, which is why the clockwise field worked.
I should be plain about the limit of this walk-through: the Poincaré-Hopf theorem is the real content, and I have used it rather than proved it. What the popular intuition gets wrong is worth naming, though. People explain the theorem by the sphere's curvature or its closedness. Neither is it. A doughnut is closed and curved and combs perfectly; the surface of a cube is piecewise flat and combs no better than a sphere. The obstruction is topological, and the number 2 is the whole of it.
The analogy
THE ANALOGY #Think of a national accounting rule that every region's books must balance to a fixed total set by the country itself, not by any region. You may move a deficit from one region to another, split it, or shuffle it about — but you cannot make the total anything other than what the country's constitution says it is. The calm points are the deficits; the Euler characteristic is the constitutional total.
the accounting analogy suggests the total is a convention someone chose, whereas the Euler characteristic is forced by the shape and can be computed independently by anyone who counts corners, edges and faces — and unlike money, an index can be negative or an entire zero can carry index 2 at once.
Clarifying the model
THE MODEL #Every hypothesis in the statement is load-bearing, and dropping any one destroys it.
Tangent. The arrows must lie in the surface. Real wind has a vertical component, and the theorem says nothing about the full three-dimensional velocity — only that the horizontal part vanishes somewhere. A point of pure updraught satisfies the theorem.
Continuous. Fields with a discontinuity comb freely — the "due east" field is nonvanishing at every point where it is defined, and it fails only by being undefined at two points.
Even-dimensional sphere. The folklore version, "you can't comb a sphere", is false as stated. The circle and the 3-sphere both have characteristic 0 and both admit nowhere-zero tangent fields. It is the even-dimensional spheres that resist.
And it is not the same theorem as the one about antipodal points — that two opposite points on Earth share a temperature and pressure is Borsuk-Ulam, a different result with a different proof, routinely quoted alongside this one as though they were relatives.
What observation would refute the claim? A continuous, everywhere-tangent, nowhere-zero wind field covering the entire globe at one instant. If someone produced one, the theorem would be false. Notice how little the theorem promises in exchange: it names no location, no duration, no size. The calm point may be a metre wide over open ocean and gone a second later, which is why this is a statement about shapes and not a forecast.
A picture of it
THE PICTURE #How to readEach bar is one surface's Euler characteristic, counted from corners minus edges plus faces. The bar height is the sum every tangent field's zeros must add up to, so a bar sitting flat at zero means the surface can be combed with no calm point at all — the circle, the torus and the 3-sphere. The two bars away from the line, the sphere at +2 and the two-holed surface at -2, are the surfaces where zeros are compulsory. Read the vertical position only; the horizontal order carries no meaning.
What became clearer
WHAT CLEARED #The wind result is not about air. It is about the fact that an arrow turning around a loop must turn a whole number of times, that whole numbers cannot drift continuously, and that a surface fixes the total once and for all. A sphere's total is 2 and cannot be spent down to nothing, so the calm point can be moved anywhere but never removed. A torus's total is 0, and its calm points can be cancelled away entirely.
Where to go next
ONWARD #- Poincare-Hopf in full: how a zero's index is defined, and why the indices must sum to the Euler characteristic.
- Why the theorem forbids a nowhere-zero continuous field on any even-dimensional sphere, and what happens in dimension 3.
- The Brouwer fixed point theorem, which the hairy ball theorem can be used to prove.
Key terms
TERMS #| Term | What it means |
|---|---|
| Tangent vector field | an assignment of an arrow lying in the surface to each of its points. |
| Index of a zero | the whole number of turns the field makes as you walk once around a small loop enclosing it. |
| Euler characteristic | vertices minus edges plus faces of any polyhedral division of a surface; 2 for a sphere, 0 for a torus. |
| Poincare-Hopf theorem | the indices of a continuous tangent field's zeros sum to the surface's Euler characteristic. |
Every term the collection defines is gathered in the glossary.