Get AI summaries of any video or article — Sign up free
Manifolds 21 | Tangent Space (Definition via tangent curves) thumbnail

Manifolds 21 | Tangent Space (Definition via tangent curves)

5 min read

Based on The Bright Side of Mathematics's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.

TL;DR

Tangent spaces for abstract manifolds are defined using equivalence classes of differentiable curves through a point p, not by derivatives in an ambient R^n.

Briefing

Tangent spaces for abstract manifolds can be defined without any ambient Euclidean space by using derivatives of curves—then identifying curves that share the same first-order behavior in local coordinates. The key move is to replace “differentiate a curve inside R^n” with “differentiate the coordinate representation of a curve,” so tangent vectors become equivalence classes of tangent curves rather than literal vectors in some surrounding space.

For an embedded submanifold M ⊂ R^n, tangent vectors come from the derivative γ′(0) of a smooth curve γ in M passing through a point p. That derivative naturally lives in R^n, so the tangent space is a subspace of R^n. The difficulty with an abstract manifold is that there may be no ambient R^n at all—only the manifold M and its smooth charts. Without an ambient space, the derivative of γ itself is not directly defined, so a substitute is needed.

The solution uses charts. Given a chart H around p (mapping into R^K, where K is the manifold’s dimension), any differentiable curve γ through p can be pushed to the coordinate level as H ∘ γ. Differentiability of γ is defined so that H ∘ γ is an ordinary differentiable map between Euclidean spaces; then the tangent vector in coordinates is simply (H ∘ γ)′(0). This produces a well-defined tangent vector in R^K for each curve.

To make the construction intrinsic, curves that produce the same coordinate derivative are treated as equivalent. If two curves γ and α through p have identical derivatives at 0 after applying a chart (and this equivalence does not depend on which chart is used, thanks to smooth transition maps), then they represent the same tangent vector. Each equivalence class—often called a “box” of curves sharing the same first-order behavior—becomes an element of the tangent space.

The tangent space T_pM is defined as the set of all such equivalence classes of differentiable curves through p, typically written using a quotient notation. For embedded submanifolds, this abstract tangent space matches the classical one: there is a natural bijection between the subspace of R^n given by γ′(0) and the equivalence class determined by γ. As a result, many calculations can treat the two tangent spaces as effectively the same.

Finally, the tangent space must carry vector space operations. Addition and scalar multiplication are made well defined by transporting equivalence classes to the coordinate level (via the induced map on tangent vectors), performing the usual operations in R^K, and then returning to the equivalence-class description. With these operations shown to be independent of chart choices, T_pM becomes a genuine vector space for abstract manifolds—setting up the machinery needed for later developments in the series.

Cornell Notes

The tangent space of an abstract smooth manifold T_pM is built from curves through p, not from an ambient Euclidean space. For a chart H near p, any differentiable curve γ through p is converted to coordinates as H∘γ, and the coordinate tangent vector is (H∘γ)′(0). Curves are declared equivalent when they yield the same coordinate derivative (and this does not depend on which chart is used). The tangent space T_pM is the set of these equivalence classes. For embedded submanifolds, this construction matches the usual tangent space in R^n via a natural bijection, and vector space operations are defined by pushing to coordinates, adding/scaling in R^K, then returning to equivalence classes.

Why can’t tangent vectors on an abstract manifold be defined by differentiating curves directly like in R^n?

On an embedded submanifold M ⊂ R^n, a curve γ(t) lies in R^n, so γ′(0) is a vector in R^n. For an abstract manifold, there may be no surrounding R^n in which γ(t) lives, so the derivative of γ itself is not defined. The workaround is to use charts: the coordinate map H∘γ lands in R^K, where ordinary derivatives make sense.

How does a chart turn a curve on M into something differentiable in the usual sense?

Take a chart H around p mapping into R^K. For a curve γ through p, form the coordinate representation H∘γ. The manifold’s differentiability requirement for γ is exactly that H∘γ is an ordinary differentiable map between Euclidean spaces. Then the tangent vector in coordinates is (H∘γ)′(0).

What does it mean for two curves to represent the same tangent vector?

Two curves γ and α through p are equivalent if their coordinate derivatives agree: (H∘γ)′(0) = (H∘α)′(0). The equivalence relation is designed so the conclusion does not depend on which chart is used, because smooth transition maps preserve first-order behavior. Equivalent curves form one equivalence class, and that class is the tangent vector.

How is T_pM defined in the abstract setting?

T_pM is the set of equivalence classes of differentiable curves through p. Each class groups all curves that share the same first-order derivative in local coordinates. The tangent space is therefore intrinsic: it depends only on the manifold’s smooth structure, not on any embedding.

How do addition and scalar multiplication become well defined on equivalence classes?

Vector operations are defined by moving to the coordinate level where tangent vectors are actual vectors in R^K. An induced map sends an equivalence class (a curve’s tangent data) to its coordinate tangent vector. Addition and scaling are performed in R^K, and then the result is mapped back to an equivalence class. One must check that the outcome is independent of the chosen chart, which follows from the compatibility of charts via transition maps.

Review Questions

  1. Given two curves through p, what specific equality must hold in local coordinates for them to be equivalent?
  2. How does the construction of T_pM avoid dependence on an ambient space like R^n?
  3. What steps are required to define vector addition on T_pM using coordinate-level operations?

Key Points

  1. 1

    Tangent spaces for abstract manifolds are defined using equivalence classes of differentiable curves through a point p, not by derivatives in an ambient R^n.

  2. 2

    A chart H converts a curve γ on M into a coordinate curve H∘γ in R^K, where the derivative at 0 is well defined.

  3. 3

    Curves are equivalent when their coordinate derivatives at 0 agree; this equivalence is independent of the chart used due to smooth transition maps.

  4. 4

    The tangent space T_pM is the set of all such equivalence classes of curves through p.

  5. 5

    For embedded submanifolds, the abstract tangent space matches the classical tangent space via a natural bijection.

  6. 6

    Vector space operations on T_pM are defined by pushing equivalence classes to coordinate tangent vectors in R^K, performing addition/scalar multiplication there, and mapping back in a chart-independent way.

Highlights

Abstract tangent vectors are created by differentiating coordinate representations (H∘γ)′(0), sidestepping the need for an ambient Euclidean space.
The tangent space T_pM is an intrinsic quotient: equivalence classes of curves with identical first-order behavior in local coordinates.
A natural bijection links the abstract tangent space to the classical tangent space for embedded submanifolds, letting both be treated interchangeably in practice.
Addition and scalar multiplication on T_pM are made well defined by transporting operations to R^K and returning to equivalence classes.

Topics