Tutorials › Topology (Undergraduate) › Homotopic Maps as an Equivalence Relation

Homotopy of Maps and Paths · Tutorial 654 of 799

Homotopic Maps as an Equivalence Relation

Learn how to build the constant, reversed, and concatenated homotopies that make homotopy an equivalence relation on maps.

Advanced 9 min read

What You'll Learn

  • State homotopy precisely for continuous maps with a fixed domain and codomain.
  • Construct a constant homotopy from a map to itself.
  • Reverse a homotopy by changing its time parameter.
  • Concatenate two homotopies using a piecewise formula.
  • Verify continuity of the concatenated homotopy by a closed-set pasting argument.
  • Identify and correct a common error in the time rescaling for concatenation.

Homotopy Relates Maps With the Same Domain and Codomain

Homotopy is a way to say that two continuous maps can be deformed into one another through continuous maps. The definition involves a map on a product with the unit interval, so continuity is a condition on both the domain point and the time parameter. Once a homotopy is given, three simple operations produce homotopies that establish reflexivity, symmetry, and transitivity.

Let \(X\) and \(Y\) be topological spaces, with topologies understood, and let \(I=[0,1]\) have the subspace topology inherited from the standard topology on \(\mathbb{R}\). Give \(X\times I\) the product topology. We consider continuous maps from \(X\) to \(Y\).

Definition: A homotopy from \(f:X\to Y\) to \(g:X\to Y\) is a continuous map \(H:X\times I\to Y\) such that \(H(x,0)=f(x)\) and \(H(x,1)=g(x)\) for every \(x\in X\). If such an \(H\) exists, the maps \(f\) and \(g\) are homotopic, written \(f\simeq g\).

The maps \(f\) and \(g\) must have the same domain and codomain for this relation to apply. We will use the product topology in checking continuity. In particular, if \(r:I\to I\) is continuous, then \((x,t)\mapsto(x,r(t))\) is continuous: the preimage of a basic open rectangle \(U\times V\) is \(U\times r^{-1}(V)\), which is open in \(X\times I\). The same reasoning applies when the time interval is restricted to a subinterval of \(I\).

The Three Homotopy Operations

Theorem: For any topological spaces \(X,Y\), homotopy is an equivalence relation on the set of continuous maps \(X\to Y\). In other words, it is reflexive, symmetric, and transitive.

Proof. Fix continuous maps \(f,g,h:X\to Y\) as needed for each part.

Reflexivity. Define \(C:X\times I\to Y\) by \(C(x,t)=f(x)\). This is \(f\circ\pi_1\), where \(\pi_1:X\times I\to X\) is the first projection; the projection is continuous in the product topology, so \(C\) is continuous. Also \(C(x,0)=f(x)\) and \(C(x,1)=f(x)\) for every \(x\). Thus \(f\simeq f\).

Symmetry. Suppose \(H:X\times I\to Y\) is a homotopy from \(f\) to \(g\). Define \(H^{\mathrm{rev}}(x,t)=H(x,1-t)\). The function \(r(t)=1-t\) is continuous from \(I\) to \(I\), and the product map \((x,t)\mapsto(x,1-t)\) is continuous by the product-topology observation above. Hence \(H^{\mathrm{rev}}\) is continuous. Its endpoints are \(H^{\mathrm{rev}}(x,0)=H(x,1)=g(x)\) and \(H^{\mathrm{rev}}(x,1)=H(x,0)=f(x)\). Therefore \(g\simeq f\).

Transitivity. Suppose \(H\) is a homotopy from \(f\) to \(g\), and \(K\) is a homotopy from \(g\) to \(h\). Define \[ L(x,t)= \begin{cases} H(x,2t),&0\le t\le \tfrac12,\\ K(x,2t-1),&\tfrac12\le t\le 1. \end{cases} \] Both formulas are defined on their stated intervals: \(2t\in I\) when \(0\le t\le\tfrac12\), and \(2t-1\in I\) when \(\tfrac12\le t\le1\). At the joining time, the first formula gives \(H(x,1)=g(x)\), and the second gives \(K(x,0)=g(x)\). Thus the formulas agree where both apply.

For completeness, the continuity argument uses the following closed-set pasting fact. If a space is the union of finitely many closed subsets and continuous functions on those subsets agree on their overlaps, the function they define on the whole space is continuous. Indeed, for any closed \(F\subseteq Y\), its preimage on each closed subset is closed relative to that subset, hence closed in the whole space; the preimage on the union is the finite union of these closed sets and is closed. Apply this fact to the closed subsets \(X\times[0,\tfrac12]\) and \(X\times[\tfrac12,1]\) of \(X\times I\). The restrictions defining \(L\) are continuous because they are compositions of \(H\) or \(K\) with the continuous maps \((x,t)\mapsto(x,2t)\) or \((x,t)\mapsto(x,2t-1)\), respectively. The two restrictions agree at the overlap, as checked above, so \(L\) is continuous. Finally, \(L(x,0)=H(x,0)=f(x)\) and \(L(x,1)=K(x,1)=h(x)\). Thus \(f\simeq h\), proving transitivity. These constructions also work if \(X\) is empty, since the required functions and endpoint conditions are then defined on an empty domain. \(\square\)

The time rescaling is essential: each given homotopy occupies half of the new time interval. The first half runs through all of \(H\), and the second half runs through all of \(K\). This is what makes their common endpoint \(g\) the value at the join.

Worked Examples of the Formulas

Worked Example: A Constant Homotopy

Let \(X=\mathbb{R}\) and \(Y=\mathbb{R}\), both with their standard topologies, and let \(f(x)=x\). To exhibit \(f\simeq f\), define \(C:\mathbb{R}\times I\to\mathbb{R}\) by \(C(x,t)=x\).

This is the first projection, so it is continuous in the product topology. At the two endpoints, \(C(x,0)=x=f(x)\) and \(C(x,1)=x=f(x)\), for every \(x\in\mathbb{R}\). Thus \(C\) is a homotopy from \(f\) to itself.

Check: at \(x=3\), the values are \(C(3,0)=3\) and \(C(3,1)=3\), as required. The formula does not change \(x\) at any intermediate time either.

Worked Example: Reversing a Homotopy

Let \(X=\{*\}\) have its unique topology \(\{\varnothing,X\}\), and let \(Y=I\) have its subspace topology from \(\mathbb{R}\). Define \(f(* )=0\), \(g(* )=1\), and \(H(*,t)=t\). The map \(H\) is continuous: under the identification of \(\{*\}\times I\) with \(I\), it is the identity map on \(I\). Also \(H(*,0)=0=f(*)\) and \(H(*,1)=1=g(*)\).

Reverse time to obtain \(H^{\mathrm{rev}}(*,t)=H(*,1-t)=1-t\). This is continuous because \(t\mapsto1-t\) is continuous, and its endpoints are \(H^{\mathrm{rev}}(*,0)=1=g(*)\) and \(H^{\mathrm{rev}}(*,1)=0=f(*)\). It is therefore a homotopy from \(g\) back to \(f\).

Check: at \(t=\tfrac14\), the reversed formula gives \(1-\tfrac14=\tfrac34\). At \(t=\tfrac34\), it gives \(1-\tfrac34=\tfrac14\). The first value is near the endpoint \(g(*)=1\), and the second is near \(f(*)=0\), consistent with reversing the original time direction.

Worked Example: Concatenating Two Homotopies

Let \(X=\{*\}\) have its unique topology and let \(Y=\mathbb{R}\) have its standard topology. Consider the maps \(f(* )=0\), \(g(* )=1\), and \(h(* )=2\). Define \(H(*,t)=t^2\) and \(K(*,t)=1+t\). These are continuous functions on \(I\), and their endpoints are \(H(*,0)=0\), \(H(*,1)=1\), \(K(*,0)=1\), and \(K(*,1)=2\). Thus \(H\) goes from \(f\) to \(g\), and \(K\) goes from \(g\) to \(h\).

The concatenated formula is \(L(*,t)=4t^2\) for \(0\le t\le\tfrac12\), and \(L(*,t)=2t\) for \(\tfrac12\le t\le1\). The first expression comes from \(H(*,2t)=(2t)^2=4t^2\). The second comes from \(K(*,2t-1)=1+(2t-1)=2t\).

At the join, the first expression gives \(4(\tfrac12)^2=4(\tfrac14)=1\), while the second gives \(2(\tfrac12)=1\). The pieces agree, so the closed-set pasting argument proves continuity. The endpoints are \(L(*,0)=4(0)^2=0=f(*)\) and \(L(*,1)=2(1)=2=h(*)\). Hence \(L\) is a homotopy from \(f\) to \(h\).

Check: at \(t=\tfrac14\), \(L(*,\tfrac14)=4(\tfrac14)^2=4(\tfrac1{16})=\tfrac14\). At \(t=\tfrac34\), \(L(*,\tfrac34)=2(\tfrac34)=\tfrac32\). Both values use the appropriate half-interval formula, and the computed join value is \(1\) from either side.

A Common Mistake and Why the Relation Matters

Common pitfall. In the second half of the concatenation, using \(K(x,2t)\) instead of \(K(x,2t-1)\) fails to start the second homotopy at its initial value. In the preceding example, this incorrect formula would give \(K(*,2t)=1+2t\) for \(\tfrac12\le t\le1\). At the join \(t=\tfrac12\), the first piece gives \(H(*,1)=1\), but the incorrect second piece gives \(K(*,1)=2\). The formulas disagree, so they do not define the intended continuous concatenation. The corrected rescaling gives \(K(*,2(\tfrac12)-1)=K(*,0)=1\), matching the first piece. The subtraction of \(1\) maps the second half of the time interval onto the full interval \(I\), starting at time \(0\).

These three constructions show why homotopy is an equivalence relation: a map is homotopic to itself, a homotopy can be run backward, and two homotopies with a shared endpoint can be run in succession. Consequently, continuous maps with fixed domain \(X\) and codomain \(Y\) can be grouped into equivalence classes under homotopy. This grouping records which maps can be connected through a continuous family, while keeping the domain and codomain fixed.

Check Your Understanding

For each question, keep track of both the endpoint values and the time interval on which each formula is used.

  1. Write a constant homotopy from a continuous map \(f:X\to Y\) to itself, and verify its endpoints.
  2. If \(H\) runs from \(f\) to \(g\), write the reversed homotopy and compute its values at \(t=0\) and \(t=1\).
  3. If \(H\) runs from \(f\) to \(g\) and \(K\) runs from \(g\) to \(h\), write the two pieces of the concatenated homotopy. At which time do the pieces meet?
  4. For \(H(*,t)=t\) and \(K(*,t)=1+t\), where \(X=\{*\}\) and \(Y=\mathbb{R}\), compute the concatenated value at \(t=\tfrac14\) and at \(t=\tfrac34\).
  5. Explain why the formulas \(H(x,2t)\) and \(K(x,2t-1)\) agree at \(t=\tfrac12\) under the stated endpoint assumptions.