Manifolds: Curved Spaces Without Embedding
A sphere is curved, but how do you describe its geometry without referring to the 3D space it sits in? A manifold is a space that locally looks like ordinary flat space — you can draw coordinate grids on small patches — but globally may be curved, have holes, or wrap around.
This is the mathematical machinery that lets us talk about curved spacetime without asking “curved into what?”
The key idea: coordinate charts
You can’t cover a sphere with a single flat map without distortion (cartographers have known this for centuries). But you can cover it with two overlapping maps — for instance, one centred on the north pole and one on the south, each covering slightly more than a hemisphere.
Each map is a chart: a smooth, invertible function from a patch of the manifold to $\mathbb{R}^n$. A collection of charts covering the whole manifold is an atlas. Where charts overlap, there are smooth transition functions relating one set of coordinates to another.
A manifold is a space equipped with such an atlas. Dimension $n$ means each chart maps to $\mathbb{R}^n$.
| Example | Dimension | Charts needed |
|---|---|---|
| Circle $S^1$ | 1 | 2 |
| Sphere $S^2$ | 2 | 2 |
| Torus $T^2$ | 2 | 4+ (depends on choice) |
| Spacetime | 4 | Varies by topology |
Two stereographic projections cover the sphere. Drag the point on the sphere to see its coordinates in both charts. Near the poles, one chart distorts wildly while the other is well-behaved.
Tangent vectors
At each point $p$ on a manifold, there’s a tangent space $T_pM$ — the set of all possible velocity vectors of curves passing through $p$. On a 2D surface, the tangent space at each point is a 2D plane; on a 4D manifold, it’s a 4D vector space.
In coordinates $(x^1, \ldots, x^n)$, the tangent space is spanned by the partial derivative operators $\partial/\partial x^i$. A tangent vector is:
$$v = v^i \frac{\partial}{\partial x^i}$$
(summing over $i$ — this is the Einstein summation convention). The components $v^i$ depend on the coordinate choice, but the vector $v$ itself doesn’t. This distinction between the geometric object and its coordinate representation is the central theme of differential geometry.
Intrinsic vs extrinsic
Gauss’s Theorema Egregium (1827, “remarkable theorem”) proved that the curvature of a surface can be determined entirely from measurements within the surface — without reference to the space it’s embedded in. Ants on a sphere can detect the curvature by, say, measuring that the angles of a triangle sum to more than 180°.
This is why manifolds matter for physics: spacetime is a 4D manifold, and there’s no 5D space it sits in. All the geometry — curvature, distances, geodesics — must be describable intrinsically, using only coordinates on the manifold itself.
Smooth structure
“Smooth” means the transition functions between charts are infinitely differentiable ($C^\infty$). This ensures that calculus works globally: you can differentiate functions, define vector fields, and eventually build the tensor machinery needed for general relativity.
Not all manifolds are smooth (there are topological manifolds that resist smooth structure), but all the manifolds in physics are.