Skip to content

Smooth Manifolds | Mathematics

An nn-dimensional topological manifold MM is a topological space that is:

  1. Hausdorff: distinct points have disjoint neighborhoods.
  2. Second countable: the topology has a countable basis.
  3. Locally Euclidean: every point pMp \in M has a neighborhood UU homeomorphic to an open subset of Rn\mathbb{R}^n.

A homeomorphism φ:UVRn\varphi : U \to V \subseteq \mathbb{R}^n is called a coordinate chart (or just chart), and (U,φ)(U, \varphi) is a coordinate neighborhood.

A smooth atlas on a topological nn-manifold MM is a collection of charts {(Uα,φα)}\{(U_\alpha, \varphi_\alpha)\} such that:

  1. The UαU_\alpha cover MM.
  2. For every pair of overlapping charts, the transition map φβφα1:φα(UαUβ)φβ(UαUβ)\varphi_\beta \circ \varphi_\alpha^{-1} : \varphi_\alpha(U_\alpha \cap U_\beta) \to \varphi_\beta(U_\alpha \cap U_\beta) is a smooth diffeomorphism.

Two atlases are compatible if their union is also a smooth atlas. A smooth structure on MM is a maximal smooth atlas.

Example 1. Rn\mathbb{R}^n with the identity chart is a smooth manifold.

Example 2. Sn={xRn+1:x=1}S^n = \{x \in \mathbb{R}^{n+1} : \|x\| = 1\} is a smooth manifold. Use stereographic projection or the 2(n+1)2(n+1) hemisphere charts.

Example 3. The general linear group GLn(R)={AMn(R):det(A)0}GL_n(\mathbb{R}) = \{A \in M_n(\mathbb{R}) : \det(A) \neq 0\} is an open subset of Rn2\mathbb{R}^{n^2}, hence a smooth manifold of dimension n2n^2.

Example 4. The real projective space RPn\mathbb{RP}^n is a smooth manifold of dimension nn.

A map f:MNf : M \to N between smooth manifolds is smooth if for every pMp \in M, there exist charts (U,φ)(U, \varphi) near pp and (V,ψ)(V, \psi) near f(p)f(p) with f(U)Vf(U) \subseteq V, such that ψfφ1\psi \circ f \circ \varphi^{-1} is smooth as a map between open subsets of Euclidean spaces.

A diffeomorphism is a smooth bijection with smooth inverse. If MM and NN are diffeomorphic, we write MNM \cong N.

Proposition 1.1. Diffeomorphism is an equivalence relation on the class of smooth manifolds.

For a smooth manifold MM of dimension nn, the tangent space TpMT_p M at pMp \in M can be defined in several equivalent ways:

Definition (Derivations). A tangent vector at pp is a linear map v:C(M)Rv : C^\infty(M) \to \mathbb{R} satisfying the Leibniz rule:

v(fg)=f(p)v(g)+v(f)g(p)v(fg) = f(p)\, v(g) + v(f)\, g(p)

The space of all such derivations is TpMT_p M, an nn-dimensional vector space.

Definition (Curves). A tangent vector is an equivalence class of smooth curves γ:(ε,ε)M\gamma : (-\varepsilon, \varepsilon) \to M with γ(0)=p\gamma(0) = p, where γ1γ2\gamma_1 \sim \gamma_2 if they have the same derivative in any chart.

In local coordinates (x1,,xn)(x^1, \ldots, x^n), a basis for TpMT_p M is given by the partial derivative operators {/xip}\{\partial/\partial x^i|_p\}.

The differential. For a smooth map f:MNf : M \to N, the pushforward or differential dfp:TpMTf(p)Ndf_p : T_p M \to T_{f(p)} N is defined by:

dfp(v)(g)=v(gf)df_p(v)(g) = v(g \circ f)

for gC(N)g \in C^\infty(N). In coordinates, dfpdf_p is represented by the Jacobian matrix.

The cotangent space TpMT_p^* M is the dual vector space to TpMT_p M. Elements are called covectors or differential 1-forms at pp.

In coordinates, the basis dual to {/xi}\{\partial/\partial x^i\} is {dxi}\{dx^i\}, defined by dxi(/xj)=δjidx^i(\partial/\partial x^j) = \delta^i_j.

The differential of a function fC(M)f \in C^\infty(M) at pp is the covector:

dfp(v)=v(f)df_p(v) = v(f)

In coordinates: df=fxidxidf = \frac{\partial f}{\partial x^i}\, dx^i.

A smooth vector field XX on MM assigns a tangent vector XpTpMX_p \in T_p M smoothly to each pMp \in M. In coordinates:

X=Xi(x)xiX = X^i(x) \frac{\partial}{\partial x^i}

where XiX^i are smooth functions.

Integral curves. A curve γ(t)\gamma(t) is an integral curve of XX if γ˙(t)=Xγ(t)\dot{\gamma}(t) = X_{\gamma(t)}. The flow ϕt\phi_t of XX is a one-parameter family of diffeomorphisms.

Lie bracket. The Lie bracket of two vector fields X,YX, Y is:

[X,Y](f)=X(Y(f))Y(X(f))[X, Y](f) = X(Y(f)) - Y(X(f))

In coordinates: [X,Y]i=XjjYiYjjXi[X, Y]^i = X^j \partial_j Y^i - Y^j \partial_j X^i.

Problem 1. Show that S1S^1 is a smooth manifold by constructing an atlas with two charts.

Solution. Use stereographic projection from the north and south poles. For U1=S1{(0,1)}U_1 = S^1 \setminus \{(0, 1)\}, φ1(x,y)=x/(1y)\varphi_1(x, y) = x/(1 - y). For U2=S1{(0,1)}U_2 = S^1 \setminus \{(0, -1)\}, φ2(x,y)=x/(1+y)\varphi_2(x, y) = x/(1 + y). The transition map φ2φ11(t)=1/t\varphi_2 \circ \varphi_1^{-1}(t) = 1/t is smooth on R{0}\mathbb{R} \setminus \{0\}. \blacksquare

Problem 2. Show that T2=S1×S1T^2 = S^1 \times S^1 is a smooth manifold.

Problem 3. Prove that the tangent bundle TM=pMTpMTM = \bigcup_{p \in M} T_p M is itself a smooth 2n2n-dimensional manifold.

A manifold is a space that looks like Euclidean space when you zoom in closely enough. The surface of the Earth is a two-dimensional manifold: locally it appears flat (which is why map projections work for small regions), but globally it has a different topology (it is a sphere). The key idea is that you can cover the manifold with overlapping coordinate charts, each mapping a small patch to Rn\mathbb{R}^n, and the transition maps between overlapping charts are smooth.

A smooth structure on a manifold is the collection of charts that make all transition maps smooth. This is what allows you to do calculus on manifolds: derivatives, integrals, and differential equations make sense because each chart locally looks like Euclidean space. The transition maps ensure that these local constructions are compatible across overlapping charts. Partitions of unity are the technical tool that lets you glue local constructions (like Riemannian metrics or differential forms) into global ones. Manifolds are the natural setting for general relativity (spacetime is a 4-dimensional Lorentzian manifold), gauge theory (connections on principal bundles), and many areas of modern geometry and physics.

Definition. A subset NMN \subseteq M is an embedded submanifold of dimension knk \leq n if for every pNp \in N, there exists a chart (U,φ)(U, \varphi) of MM such that φ(UN)=φ(U)(Rk×{0})\varphi(U \cap N) = \varphi(U) \cap (\mathbb{R}^k \times \{0\}).

Example. Sn1RnS^{n-1} \subseteq \mathbb{R}^n is an embedded submanifold of dimension n1n-1.

Example. The torus T2R3T^2 \subseteq \mathbb{R}^3 is an embedded submanifold of dimension 22.

Theorem 1.2. Every smooth manifold admits a partition of unity: a collection of smooth functions {ρα}\{\rho_\alpha\} such that supp(ρα)\mathrm{supp}(\rho_\alpha) is locally finite, each ρα0\rho_\alpha \geq 0, and αρα=1\sum_\alpha \rho_\alpha = 1.

Partitions of unity are used to construct global objects (Riemannian metrics, connections) from local data.

Problem 4. Show that S1×S1S^1 \times S^1 is diffeomorphic to the torus T2T^2 embedded in R3\mathbb{R}^3.

Problem 5. Construct an atlas for RP2\mathbb{RP}^2 and verify that the transition maps are smooth.

Mistake 1: Confusing a chart with an atlas A chart (U,φ)(U, \varphi) maps a single open set to Rn\mathbb{R}^n, while an atlas is a collection of charts that cover the entire manifold. A single chart is not sufficient for most manifolds (e.g., S1S^1 requires at least two charts). An atlas must have smooth transition maps on all overlaps.

Mistake 2: Assuming embedded submanifolds are the same as immersed submanifolds An embedded submanifold has the subspace topology, while an immersed submanifold may have a finer topology. For example, a dense line on a torus is an immersed submanifold but not embedded. The key difference is that embedded submanifolds are locally flat in the ambient manifold, while immersed submanifolds may self-intersect globally.

Mistake 3: Forgetting that partitions of unity are necessary for gluing local constructions Local objects like Riemannian metrics or differential forms can be defined on each chart, but gluing them into a global object requires a partition of unity. Without it, there is no canonical way to combine local data that agree on overlaps. Partitions of unity always exist on smooth manifolds, which is a key reason smoothness is assumed.

flowchart TD
A[1_Smooth Manifolds] --> B[Key Concepts]
A --> C[Core Principles]
A --> D[Practical Applications]
B --> E[Fundamental definitions]
C --> F[Design patterns]
D --> G[Real-world usage]