Manifold

A manifold is a space in mathematics that, near each point, resembles ordinary Euclidean space of a fixed dimension, even when its overall shape is different. A circle is locally like a line, while a sphere’s surface is locally like a plane. This local resemblance concerns coordinates and continuity, not necessarily distances or curvature. Manifolds provide a common framework for topology, differential geometry, and mathematical models in physics. (preposterousuniverse.com)

Definition and local coordinates

Under the standard definition, a topological manifold of dimension $n$ is a topological space $M$ satisfying three conditions:

• Locally Euclidean: every point has an open neighborhood homeomorphic to an open subset of $\mathbb{R}^n$.
• Hausdorff: any two distinct points have disjoint open neighborhoods.
• Second-countable: the topology has a countable basis of open sets.

Here $\mathbb{R}^n$ consists of ordered $n$-tuples of real numbers. A homeomorphism is a bijection that is continuous in both directions. These requirements exclude certain pathological spaces while allowing considerable global complexity. (math.colostate.edu)

A coordinate chart is a pair $(U,\varphi)$, where $U$ is an open set in $M$ and

is a homeomorphism. Its component functions assign local coordinates to points. An atlas is a collection of charts whose domains cover the manifold. On overlapping domains, the transition map $\psi\circ\varphi^{-1}$ expresses one coordinate system in terms of another. A manifold generally requires several charts rather than one global coordinate system. (people.maths.ox.ac.uk)

Smooth and complex structures

A smooth manifold has an atlas whose transition maps possess continuous partial derivatives of every order. Such an atlas determines a maximal compatible atlas, called a smooth structure. Smoothness allows calculus to be performed consistently: a function or map is smooth when its coordinate expressions are smooth, independently of the compatible charts chosen. A topological manifold, by itself, does not specify this additional structure. (people.maths.ox.ac.uk)

Two smooth manifolds are diffeomorphic if there is a smooth bijection between them with a smooth inverse. This is a stronger equivalence than homeomorphism, because it preserves differentiable as well as topological structure. Intermediate $C^k$ structures require only $k$ continuous orders of derivatives. (people.maths.ox.ac.uk)

A complex manifold instead uses charts in $\mathbb{C}^m$, with holomorphic transition maps. Its underlying real manifold has dimension $2m$. The compatibility requirement is stronger than ordinary smoothness and connects manifold theory with complex analysis. (people.maths.ox.ac.uk)

Examples, dimension, and boundary

Euclidean space is itself a manifold. The sphere

has dimension $n$, although this realization places it in an ambient space of dimension $n+1$. Dimension therefore counts local coordinates, not the coordinates used to describe an embedding. Real projective space, whose points are lines through the origin in $\mathbb{R}^{n+1}$, is another $n$-dimensional example. (math.colostate.edu)

The torus $T^2=S^1\times S^1$ is a two-dimensional manifold. More generally, the Cartesian product of manifolds of dimensions $m$ and $n$ has dimension $m+n$. A Lie group is a smooth manifold carrying compatible group operations; rotation groups provide examples. A branching junction, by contrast, does not locally resemble an interval and is not a one-dimensional manifold. (preposterousuniverse.com)

For a manifold with boundary, charts may take values in relatively open subsets of the half-space

A closed disk is a two-dimensional manifold with boundary; its boundary is a circle. A sphere’s surface has no manifold boundary, despite bounding a ball in an ambient space. Corners require a further extension of the chart model. (people.maths.ox.ac.uk)

Tangent spaces and geometric structures

At each point $p$ of a smooth $n$-manifold, the tangent space $T_pM$ is an $n$-dimensional vector space representing infinitesimal directions. The differential of a smooth map $f:M\to N$ is a linear map

Its coordinate representation is the Jacobian matrix. Tangent vectors can be defined intrinsically, without placing the manifold inside a larger Euclidean space. (preposterousuniverse.com)

A Riemannian metric assigns a smoothly varying positive-definite inner product to each tangent space. Its local components form the metric tensor, which determines lengths, angles, and volume. On a connected manifold, distances are obtained by taking the infimum of curve lengths. A geodesic is a curve with zero covariant acceleration; geodesics need not minimize distance globally. (people.maths.ox.ac.uk)

The metric determines the Levi-Civita connection, enabling differentiation of vector fields along curves. Its curvature is encoded by the Riemann curvature tensor. Thus topology, smooth structure, and metric geometry are distinct layers: a manifold can support different metrics without changing its underlying smooth structure. (people.maths.ox.ac.uk)

Applications

In general relativity, spacetime is modeled as a four-dimensional smooth manifold with a Lorentzian metric. Unlike a Riemannian metric, this metric is indefinite, distinguishing temporal and spatial directions. Coordinate systems describe regions of spacetime rather than constituting the physical geometry itself. (preposterousuniverse.com)

In machine learning, the manifold hypothesis proposes that some high-dimensional data concentrate near a lower-dimensional manifold. Manifold learning methods seek to exploit this structure for dimensionality reduction and representation. This is a modeling assumption about particular data distributions, not a theorem that arbitrary datasets must satisfy. (arxiv.org)