HOMOTOPY OF PATHS AND FUNDAMENTAL GROUP

ADITYA KARNATAKI

1. Introduction
We have seen in pointset topology the various properties of ‘shape’ - these include connectedness, com-
pactness, Hausdorffness, regularness, normalness, metrizability etc. These tell us a lot about spaces and their
nature and these provide us with useful topological invariants. It is however still quite difficult to distinguish
spaces based on these topological invariants. An example of this is Brouwer’s ‘invariance of domain’ theorem,
which approximately says, among other things, that Rm and Rn are not homeomorphic to each other when
m ̸= n. It is still quite difficult to prove this very intuitive-looking result with the tools we have at our dis-
posal till now. Algebraic topology evolved to handle such questions - it puts more invariants at our disposal
and these are invariants that are more ‘computable’ in some sense than just being abstract properties. The
trade-off is that we have to weaken the notion of two spaces being homeomorphic to a different notion that
gives us more flexibility and working space, so to speak. This notion is known as ‘homotopy equivalence’.
Today we will begin to study this idea and define a useful invariant, namely, the fundamental group(oid), of
a topological space under this weaker concept of equivalence.

2. Homotopy of paths
Homotopy is the concept of continuous deformation, parametrized by the unit interval I = [0, 1].
Definition 2.1. Let X, Y be two topological spaces. Let f, g : X → Y be two continuous maps between
them. A homotopy between f and g is a continuous map F : X × I → Y such that F (x, 0) = f (x) and
F (x, 1) = g(x) for all x ∈ X. If such an F exists, we say that f and g are homotopic and write f ≃ g. If f
is homotopic to a constant map, we say it is nullhomotopic.
We want to study paths in topological spaces, i.e. continuous f : [0, 1] → X, f (0) = x0 , f (1) = x1 . The
above notion of homotopy is not very useful if we don’t fix the endpoints x0 and x1 of these paths, because
then every path is homotopic to a constant path. (Exercise!) A better notion of ‘homotopy of paths’ only
considers homotopies which keep the end points in place.
Definition 2.2. Two paths f, g : [0, 1] → X from x0 to x1 are path homotopic if there exists F : [0, 1] ×
[0, 1] → X such that F (s, 0) = f (s), F (s, 1) = g(s), (so that it is a homotopy) and F (0, t) = x0 , F (1, t) = x1
(so that endpoints are fixed). In particular, for all t ∈ [0, 1], ft = F |[0,1]×{t} is a path from x0 to x1 . Such
an F is called a path homotopy and we write f ≃p g for this relation between f and g.
We can think of a homotopy as giving a path between the paths f and g, that is, we can think of it as
giving a function Γ : [0, 1] → {Paths in X}, where Γ(t) = ft . It is possible, with some mild assumptions on
X, to make the set of paths in into a topological space in such a way that this means of producing a Γ from
an F gives a bijection between homotopies as defined above and “paths between paths”, but we will avoid
that kind of point-set topological technicality. Instead, we will define homotopies as map from a rectangle
[0, 1] × [0, 1] which gives a nice intuition as paths of paths.
Lemma 2.3. The relations ≃ and ≃p are equivalence relations on paths.

Proof. (1) Clearly f ≃ f since the constant homotopy F (s, t) = f (s) for all t ∈ [0, 1] gives a homotopy
from f to itself.
(2) If f ≃ g with homotopy F (s, t), then the reverse homotopy G(s, t) = F (s, 1 − t) gives a homotopy
from g to f , so g ≃ f .
1
 (3) If f ≃ g and g ≃ h with homotopies F (s, t) and G(s, t), then we can concatenate the homotopies :
define H(s, t) : [0, 1]×[0, 1] → X such that H(s, t) = F (s, 2t) if 0 ≤ t ≤ 1/2, and H(s, t) = G(s, 2t−1)
if 1/2 ≤ t ≤ 1. These two formulae agree at t = 1/2, since F (s, 1) = g(s) = g(s, 0), and so H(s, t) is
well-defined and continuous by the pasting lemma, and gives a homotopy f ≃ h.
In the case of path homotopies, we can check that the above constructions preserve the requirements
F (0, t) = x0 and F (1, t) = x1 for all t, so yield path homotopies, and thus ≃p is also an equivalence
relation. □
We denote the (path) homotopy equivalence class of f by [f ].
Example 2.4. Below are some natural examples and non-examples of path homotopies.
(1) If f and g are paths in R2 (or any convex subset of Rn for any n ≥ 1), we can define the straight
line homotopy between f and g as-
F (s, t) = (1 − t)f (s) + tg(s).
For each fixed value of s, this connects f (s) with g(s) by a straight line, and this defines a path
homotopy by convexity hypothesis. So f and g are homotopic always!
(2) In the punctured plane X = R2 \ {(0, 0)}, let f, g be paths from (−1, 0) to (1, 0) such that f stays in
the upper half plane {(x, y) : y ≥ 0} and g stays in the lower half plane {(x, y) : y ≤ 0}. Then there
is no homotopy between f and g. We will prove this rigorously later.

3. Fundamental groupoid
Now we prove that the homotopy classes of paths in any space X form a category, in fact, a groupoid.
The key operation is what we have seen earlier in a similar context- composition (concatenation) of paths.
Definition 3.1. If f is a path from x to y in a topological space X, and g is a path from y to z, define a
path f ∗ g from x to z by running through f first and then through g, twice as fast-
(f ∗ g)(s) = f (2s), 0 ≤ s ≤ 1/2
and then
(f ∗ g)(s) = g(2s − 1), 1/2 ≤ s ≤ 1.
As before, this gives a well-defined path by the pasting lemma.
The point is that this gives a well-defined composition operation on path-homotopy classes as well, as
long as f (1) = g(0) : If f ≃p f ′ and g ≃p g ′ with homotopies F (s, t) and G(s, t), then f ∗ g ≃p f ′ ∗ g ′ using
the homotopy
(F ∗ G)(s, t) = F (2s, t), 0 ≤ s ≤ 1/2
and then
(F ∗ G)(s, t) = G(2s − 1, t), 1/2 ≤ s ≤ 1.
So we can define
[f ] ∗ [g] := [f ∗ g].
Note that this is still assuming f (1) = g(0), and since all paths in a path-homotopy class have the same
endpoints, it doesn’t matter which representatives f and g we choose.
Proposition 3.2. The operation ∗ is associative on path-homotopy classes, and has identity and inverses.
Proof. As it turns out, associativity is the trickiest, but not very difficult.
(1) Associativity: Given paths f, g, h with f (1) = g(0) and g(1) = h(0), the claim is that (f ∗ g) ∗ h ≃p
f ∗ (g ∗ h). Both these paths run along f , then g, then h, but the parametrizations are different,
i.e. the speeds are different. In the LHS, f and g are traversed at constant speed and h is traversed
at double that speed, and similarly for RHS, f is traversed at double the speed of g and h. The
homotopy comes from adjusting for this: define
 
4s 1+t
F (s, t) := f ,0 ≤ s ≤
1+t 4
2
 and
1+t 2+t
F (s, t) := g(4s − (1 + t)), ≤s≤
4 4
and finally,  
4s − (2 + t) 2+t
F (s, t) := h , ≤ s ≤ 1.
2−t 4
This may feel like a lot, but this is really not much. If you think of what happens at t = 0 and t = 1
for s = 1/4, 1/2, 3/4, you will be able to combine this information with straight-line homotopies in
the variable t that we have kind of seen earlier.
(2) Identity: Given x, y ∈ X, and f any path from x to y, we claim that idx := [ex ] is the identity for
this operation, where ex : I → X defined as ex (s) = x for all s is the constant path for the point x.
That is, we want to show [f ] ∗ idy = idx ∗ [f ] = f . Indeed, there is an explicit homotopy f ≃p f ∗ idy
by defining  
s
F (s, t) := f , s ∈ [0, 1 − t/2]
1 − t/2
and
F (s, t) = y, s ∈ [1 − t/2, 1].
Similarly, there is an explicit homotopy f ≃p (f ∗ idx ) by defining
G(s, t) := x, s ∈ [0, t/2]
and  
s − t/2
G(s, t) := f , s ∈ [t/2, 1].
1 − t/2
(3) Inverses: Given f a path from x to y, define the reverse path f¯(s) := f (1 − s) from y to x. Then
we claim that [f¯] is an inverse for [f ], i.e. ex ≃p f ∗ f¯ and ey ≃p f¯ ∗ f . We will only do the first
case here, leaving the second as an exercise. The homotopy in this case is quite easy to write:
F (s, t) := f (2ts) for s ∈ [0, 1/2] and F (s, t) := f (2t(1 − s)) for s ∈ [1/2, 1]. For any given t, this
runs forward in the direction of f from f (0) = x to f (t) at s = 1/2, then runs backward from f (t)
to f (0) = x at s = 1. For t = 0, we have the path ex , and for t = 1, we have the path f ∗ f¯.
□
We can package all this information in a category C: the objects of C are points of X, and Hom(x, y)
in C are simply {homotopy classes of paths from x to y}. This set is empty if x and y are not in the same
path-component of X. With the above discussion, it is clear that this is a category, because the equivalence
class of the concatenation of two paths only depends on the equivalence classes of the individual paths.
This is in fact a groupoid, since all morphisms are invertible and hence isomorphisms. This is called the
‘fundamental groupoid’ of X.

4. Fundamental Group
Since the fundamental groupoid of any non-empty non-singleton space X has more than one objects, the
inability to multiply every pair of paths prevents us from having an actual group structure. To address this,
one usually restricts to a single object of the category, i.e. we fix a base point x0 ∈ X, and only consider
paths that go from x0 to itself - i.e. loops, based at x0 .
Definition 4.1. The set of path-homotopy classes of loops based at x0 , with the concatenation operation ∗
is called the fundamental group of X for the base point x0 , denoted π1 (X, x0 ).
Example 4.2. In any convex subset X of Rn and a point x0 ∈ X, every loop at x0 is path-homotopic to
the constant path (i.e. the identity) at x0 by the straight-line homotopy : F (S, t) = (1 − t)f (s) + tx0 . So
π1 (X, x0 ) = {idx0 }.
Definition 4.3. A nonempty topological space X is simply connected if X is path connected, and for any
x0 ∈ X, π1 (X, x0 ) = {1}.
Example 4.4. Rn , convex subspaces of Rn , the n-sphere S n for n ≥ 2 (not completely obvious).
We will study the fundamental group and examples of it in detail in the next few lectures.
3