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

Manifolds 21 | Tangent Space (Definition via tangent curves) [dark version]

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 built from differentiable curves through p using equivalence classes, not from derivatives in an ambient 2n.

Briefing

Tangent spaces for abstract manifolds can be defined without embedding the manifold into any surrounding Euclidean space by using tangent curves and an equivalence relation. The key move is to replace “take the derivative of a curve inside the ambient space” with “take the derivative after pushing the curve into a chart,” where ordinary derivatives make sense. This matters because it lets tangent vectors—and the tangent space built from them—be defined purely from the manifold’s own smooth structure, not from an external coordinate container.

For a submanifold sitting inside 2n, tangent vectors are obtained by differentiating smooth curves through a point p and then interpreting the result as a vector in 2n. The transcript highlights the obstacle for an abstract manifold M: there is no ambient 2n, so the derivative of a curve cannot be interpreted the same way. The workaround is to use charts. A chart map b7 pushes local information from M into 2K, where standard calculus applies. If a curve b3 on M passes through p, then composing with the chart gives a differentiable curve in 2K, and the derivative at the parameter value 0 produces a vector in 2K.

However, different curves through p can yield the same derivative in 2K. To capture “the tangent vector” rather than “the particular curve,” the definition groups curves into equivalence classes: two curves b3 and b1 are equivalent if their derivatives agree after pushing them through a chart (equivalently, checking one chart is enough because transition maps preserve the derivative outcome). Each equivalence class is treated as a single tangent vector. The tangent space T_pM is then the set of all such equivalence classes of differentiable curves through p.

The transcript also connects this abstract construction back to the familiar embedded case. For a submanifold, the equivalence-class tangent vector corresponds naturally to the usual derivative vector in 2n. This correspondence is described as a well-defined projection: because all curves in an equivalence class share the same derivative in the chart coordinates, the class determines the same tangent vector in the embedded picture. As a result, calculations can often be done using either viewpoint without changing the outcome.

Finally, the tangent space must be more than a set: it needs vector space operations. Addition and scalar multiplication are defined by transporting equivalence classes to the chart level (where vectors in 2K can be added and scaled), performing the operations there, and then mapping back to equivalence classes. The transcript emphasizes that these operations must be well defined—independent of which chart is used—so the smooth structure of the manifold guarantees consistency. With that, T_pM becomes a genuine vector space for abstract manifolds, setting up the machinery needed for later results in the series.

Cornell Notes

For an abstract smooth manifold M, tangent vectors are defined using differentiable curves through a point p and an equivalence relation. A curve b3 is pushed into local coordinates via a chart, producing a differentiable curve in 2K; the derivative at 0 gives a vector in 2K. Curves are considered equivalent when these derivatives match (checking one chart suffices due to compatibility under transition maps), and each equivalence class represents a tangent vector. The tangent space T_pM is the set of all equivalence classes of such curves. For embedded submanifolds, this construction matches the usual tangent space in 2n via a natural projection, and vector space operations (addition and scalar multiplication) are defined by performing them in chart coordinates and then returning to equivalence classes.

Why can’t tangent vectors on an abstract manifold be defined by differentiating curves directly in an ambient space?

An abstract manifold has no surrounding Euclidean space like 2n. Without an ambient target, the derivative of a curve b3: (-b5,b5) a3 M cannot be interpreted as a vector in some fixed 2n. The transcript resolves this by using charts: composing b3 with a chart map sends the curve into 2K, where ordinary derivatives exist.

How does the chart-based derivative turn a curve into a tangent vector?

Given a differentiable curve b3 through p, the chart map b7 (denoted as H after the transcript’s notation) produces a differentiable map into 2K. The tangent vector is then defined as the derivative of this coordinate representation at the parameter value 0, yielding an element of 2K.

What is the equivalence relation on curves, and why is it needed?

Different curves through p can produce the same derivative in chart coordinates. To avoid treating those curves as different tangent vectors, the definition declares two curves equivalent if their derivatives agree after pushing them to 2K via the chart. The equivalence class is what represents the tangent vector, not the specific curve.

Why does checking equivalence in one chart suffice?

The transcript notes that the equivalence condition does not depend on which chart is used: transition maps between charts preserve the derivative outcome. Therefore, verifying the derivative agreement in one chart guarantees the same result in any other compatible chart.

How are addition and scalar multiplication defined on T_pM?

Vector space operations are defined by moving to chart coordinates: an equivalence class (tangent vector) is mapped to its representative vector in 2K, where addition and scalar multiplication are standard. After performing the operation in 2K, the result is mapped back to an equivalence class in T_pM. Well-definedness requires that the outcome does not depend on the chosen chart.

How does the abstract tangent space relate to the tangent space of a submanifold embedded in 2n?

For a submanifold, there are two tangent notions: the embedded one (derivatives in 2n) and the abstract one (equivalence classes of curves). The transcript describes a natural bijection/projection between them: the equivalence class determines the same derivative vector in the embedded picture, so the two tangent spaces can be treated as the same for calculations.

Review Questions

  1. Given a curve b3 through p, what exact object is differentiated to produce the tangent vector in the chart-based definition?
  2. What does it mean for two curves to be equivalent in the tangent-space construction, and what role do transition maps play?
  3. How would you define addition of two tangent vectors in T_pM using the chart-level representation?

Key Points

  1. 1

    Tangent spaces for abstract manifolds are built from differentiable curves through p using equivalence classes, not from derivatives in an ambient 2n.

  2. 2

    A chart pushes a curve b3 on M into 2K, where the derivative at parameter 0 produces a coordinate vector.

  3. 3

    Two curves represent the same tangent vector when their chart-coordinate derivatives agree; the equivalence class is the tangent vector.

  4. 4

    Equivalence can be checked in one chart because transition maps preserve the derivative-based condition.

  5. 5

    For embedded submanifolds, the equivalence-class tangent space matches the usual tangent space in 2n via a natural projection/bijection.

  6. 6

    T_pM becomes a vector space by defining addition and scalar multiplication through chart-level operations in 2K and mapping back to equivalence classes, with well-definedness across charts.

Highlights

The absence of an ambient space is handled by pushing curves into chart coordinates, where derivatives are defined.
A tangent vector is not a curve itself but an equivalence class of curves whose coordinate derivatives match at the point.
Transition maps ensure the curve-equivalence relation—and thus T_pM—doesn’t depend on which chart is used.
Vector space operations on T_pM are defined by doing arithmetic in 2K after chart-pushing, then returning to equivalence classes.

Topics