Skip to content

Directed Homotopy

Defines Directed Homotopy, directed homotopy theory, d-homotopy

Let (X,dX)(X, dX) be a d-space (directed space): a topological space XX equipped with a set dXC([0,1],X)dX \subseteq C([0,1], X) of directed paths (d-paths) satisfying:

  1. every constant path cx(t)=xc_x(t) = x belongs to dXdX,
  2. if α,βdX\alpha, \beta \in dX and α(1)=β(0)\alpha(1) = \beta(0), then the concatenation αβdX\alpha * \beta \in dX,
  3. if αdX\alpha \in dX and φ:[0,1][0,1]\varphi: [0,1] \to [0,1] is continuous and non-decreasing, then αφdX\alpha \circ \varphi \in dX.

Definition. Let α,βdX\alpha, \beta \in dX be d-paths with α(0)=β(0)=x\alpha(0) = \beta(0) = x and α(1)=β(1)=y\alpha(1) = \beta(1) = y. A directed homotopy (d-homotopy) from α\alpha to β\beta is a continuous map H:[0,1]×[0,1]XH: [0,1] \times [0,1] \to X such that:

  1. H(0,t)=α(t)H(0, t) = \alpha(t) and H(1,t)=β(t)H(1, t) = \beta(t) for all t[0,1]t \in [0,1],
  2. H(s,0)=xH(s, 0) = x and H(s,1)=yH(s, 1) = y for all s[0,1]s \in [0,1] (endpoints are fixed),
  3. for each s[0,1]s \in [0,1], the map tH(s,t)t \mapsto H(s, t) is a d-path in dXdX.

If such an HH exists, α\alpha and β\beta are d-homotopic, written αdβ\alpha \simeq_d \beta.

Proposition. D-homotopy with fixed endpoints is an equivalence relation on dX(x,y)={γdX:γ(0)=x,  γ(1)=y}dX(x,y) = \{\gamma \in dX : \gamma(0) = x,\; \gamma(1) = y\}.

Proof sketch. Reflexivity: take H(s,t)=α(t)H(s,t) = \alpha(t). Symmetry: given HH from α\alpha to β\beta, the map (s,t)H(1s,t)(s,t) \mapsto H(1-s, t) is a d-homotopy from β\beta to α\alpha (the deformation parameter ss carries no directedness constraint). Transitivity: given HH from α\alpha to β\beta and KK from β\beta to γ\gamma, concatenate in ss. In each case, condition (3) is preserved because each ss-slice remains a d-path. \square

Definition. The fundamental category π1(X,dX)\vec{\pi}_1(X, dX) is the category with objects the points of XX and morphisms Hom(x,y)=dX(x,y)/d\mathrm{Hom}(x, y) = dX(x,y)/{\simeq_d}, with composition induced by concatenation of d-paths. This is well-defined: concatenation respects d-homotopy classes. The fundamental category π1(X,dX)\vec{\pi}_1(X, dX) is a category (rather than a groupoid): a d-path from xx to yy provides a morphism xyx \to y, and an inverse morphism yxy \to x exists only when a d-path from yy to xx also exists.

Proposition. Composition in π1(X,dX)\vec{\pi}_1(X, dX) is associative, and the class of the constant d-path [cx][c_x] is the identity morphism at xx.

Proof sketch. Associativity and unit laws hold up to reparametrization. A non-decreasing reparametrization of a d-path is again a d-path, so the standard reparametrization homotopies used for ordinary paths yield d-homotopies here. \square

Examples.

  • The directed interval I=([0,1],{non-decreasing continuous maps [0,1][0,1]})\vec{I} = ([0,1],\, \{\text{non-decreasing continuous maps } [0,1] \to [0,1]\}) has fundamental category isomorphic to the partial order ([0,1],)([0,1], \leq): there is a unique morphism aba \to b when aba \leq b and none when a>ba > b.
  • The directed circle S1\vec{S}^1, obtained by identifying the endpoints of I\vec{I}, satisfies π1(S1)(N,+)\vec{\pi}_1(\vec{S}^1) \cong (\mathbb{N}, +) as a monoid with one object: every morphism is an iterate of the generating loop, and reversal of the loop is absent from dXdX.
  • In concurrency theory, the state space of a system of nn processes with mutual exclusion constraints carries a natural d-space structure. Execution traces are d-paths, and two interleavings reaching the same final state are d-homotopic. Directed obstructions (d-holes) correspond to deadlocks or unreachable states.

Relations

Date created
Mathematical object
Directed space

Cite

@misc{emsenn2026-directed-homotopy,
  author    = {emsenn},
  title     = {Directed Homotopy},
  year      = {2026},
  url       = {https://emsenn.net/library/math/domains/topology/terms/directed-homotopy/},
  publisher = {emsenn.net},
  license   = {CC BY-SA 4.0}
}