nLab path




topology (point-set topology, point-free topology)

see also differential topology, algebraic topology, functional analysis and topological homotopy theory


Basic concepts

Universal constructions

Extra stuff, structure, properties


Basic statements


Analysis Theorems

topological homotopy theory



In topology, a (parametrised, oriented) path in a space XX is a map (a morphism in an appropriate category of spaces, such as a continuous function between topological space) to XX from the topological interval 𝕀=[0,1]\mathbb{I} = [0,1].

A path from aa to bb is a path ff such that f(0)=af(0) = a and f(1)=bf(1) = b.
An unparametrised path is an equivalence class of paths, such that ff and gg are equivalent if there is an increasing automorphism ϕ\phi of 𝕀\mathbb{I} such that g=fϕg = f \circ \phi. An unoriented path is an equivalence class of paths such that ff is equivalent to (xf(1x))(x \mapsto f(1 - x)).

If PP is a path, then its reverse path1, denoted P¯\overline{P}, is defined to be the composite P(t1t)P \circ ( t\mapsto 1-t ). The operation PP¯P\mapsto\overline{P} is called path reversal.

A Moore path is defined like a path, except for having another domain: replace [0,1][0,1] with the interval [0,n][0,n] for some natural number (or, more commonly, any non-negative real number) nn. All of these variations can be combined, of course. (For unoriented paths, one usually says ‘between aa and bb’ instead of ‘from aa to bb’. Also, a Moore path from aa to bb has f(n)=bf(n) = b instead of f(1)=bf(1) = b. Finally, there is not much difference between unparametrised paths and unparametrised Moore paths, since we may interpret (tnt)(t \mapsto n t) as a reparametrisation ϕ\phi.)

In graph theory, a path is a list of edges, each of which ends where the next begins. Actually, this is a special case of the above, if we use Moore paths and interpret [0,n][0,n] as the linear graph with n+1n + 1 vertices and nn edges; in this way, the other variations become meaningful. (However, as the only directed graph automorphism of [0,n][0,n] is the identity, parametrisation is trivial for directed graphs and equivalent to orientation for undirected graphs. Note that a non-Moore path is simply an edge, one of the fundamental ingredients of a graph.)


Given a Moore path ff from aa to bb and a Moore path gg from bb to cc, the concatenation of ff and gg is a Moore path f;gf ; g or gfg \circ f from aa to cc. If the domain of ff is [0,m][0,m] and the domain of gg is [0,n][0,n], then the domain of f;gf ; g is [0,m+n][0,m+n], and

(f;g)(x){f(x) xm g(xm) xm. (f ; g)(x) \coloneqq \left \{ \array { f(x) & \quad x \leq m \\ g(x-m) & \quad x \geq m .} \right .

In this way, we get a (strict) category whose objects are points in XX and whose morphisms are Moore paths in XX, with concatenation as composition. This category is called the Moore path category.

Often we are more interested in a quotient category of the Moore path category. If we use unparametrised paths (in which case we may use paths with domain 𝕀\mathbb{I} if we wish), then we get the unparametrised path category. If XX is a smooth space, then we may additionally identify paths related through a thin homotopy to get the path groupoid. Finally, if XX is a continuous space and we identify paths related through any (endpoint-preserving) homotopy, then we get the fundamental groupoid of XX.

In graph theory, the Moore path category is known as the free category on the graph.




  1. Cf. e.g. Introduction to Topology – 2, or also Section 2.1; beware that that reference, (0) like many others, uses the term “inverse path”, even though the operation of concatenation of paths does not in and of itself yield a strict groupoid, in which P¯\overline{P} would be an inverse, and (1) that it uses aa and bb for the endpoints of the interval, not the endpoints of the paths in the space XX, and (2) that it uses P P^- instead of P¯\overline{P}, which however is less suited for notational iterating (compare P¯¯=P\overline{\overline{P}}=P with (P ) =P(P^-)^-=P), and that (3) the 2008 edition has a typo: “ w(1t)w(1-t) ” in loc. cit., when inverse path gets defined, should be u(1t)u(1-t).

Last revised on August 14, 2022 at 13:11:04. See the history of this page for a list of all contributions to it.