Fundamental Group

Loops, homotopy, concatenation

Let X be a topological space and pick a basepoint x_0 ∈ X. A loop at x_0 is a continuous map γ : [0, 1] → X with γ(0) = γ(1) = x_0. Two loops γ and δ are homotopic if one can be continuously deformed into the other while keeping endpoints fixed; formally, there is a continuous H : [0, 1] × [0, 1] → X with H(s, 0) = γ(s), H(s, 1) = δ(s), and H(0, t) = H(1, t) = x_0 for all t.

Homotopy is an equivalence relation, and we write [γ] for the class containing γ. The set of all such classes is the fundamental group π_1(X, x_0). The group operation is concatenation:

(γ · δ)(s) = γ(2s)        for 0 ≤ s ≤ 1/2
            δ(2s − 1)    for 1/2 ≤ s ≤ 1

That is, traverse γ at double speed and then traverse δ at double speed. The constant loop at x_0 is the identity, the reverse loop γ̄(s) = γ(1 − s) is the inverse of γ, and concatenation is associative up to homotopy. The associativity-up-to-homotopy is enough to make π_1(X, x_0) into a group when we pass to homotopy classes. This was Henri Poincaré's invention in his 1895 paper Analysis Situs; it founded algebraic topology.

A small zoo of fundamental groups

Space	π_1	Why
Point	{e}	Only the constant loop exists.
ℝ^n (any n)	{e}	Contractible — every loop deforms to the basepoint.
S^n for n ≥ 2	{e}	Loops have measure zero; pull them off a missed point and contract radially.
S¹ (circle)	ℤ	Loops classified by winding number; universal cover is ℝ.
T² = S¹ × S¹	ℤ²	Two independent winding directions; abelian because product.
Figure-8	F₂ (free group on 2 generators)	Two loops a, b; no relation imposed.
RP² (real projective plane)	ℤ/2ℤ	One loop; going around twice gives the trivial loop (antipodal cover).
Klein bottle	⟨a, b | abab⁻¹⟩	Non-abelian; mixes winding directions.

Two patterns are immediate. First, simply connected spaces (those with trivial π_1) are precisely the ones in which every loop contracts. The phrase "simply connected" comes from this property — the space is "connected in one piece including its loops". Second, spaces built as products X × Y have π_1(X × Y) = π_1(X) × π_1(Y), which is why the torus's group is the direct product of two copies of ℤ.

Worked example: π_1(S¹) = ℤ via the universal cover

The cleanest computation in algebraic topology is the proof that π_1(S¹) = ℤ. The argument runs through the universal cover p : ℝ → S¹ given by p(t) = (cos 2πt, sin 2πt), which wraps the real line infinitely many times around the circle.

Given a loop γ : [0, 1] → S¹ with γ(0) = γ(1) = (1, 0), there is a unique lift γ̃ : [0, 1] → ℝ with γ̃(0) = 0 and p ∘ γ̃ = γ. The endpoint γ̃(1) must satisfy p(γ̃(1)) = γ(1) = (1, 0), so γ̃(1) is an integer. This integer is the winding number of γ.

The map γ → γ̃(1) is the key. We claim it is a group isomorphism π_1(S¹) → ℤ:

1. Well-defined. If γ ~ δ are homotopic, the homotopy lifts to a homotopy of γ̃ to δ̃, so the endpoints differ by a continuous function valued in ℤ — which must be constant. So homotopic loops have the same winding number.
2. Homomorphism. The lift of γ · δ is γ̃ followed by a translate of δ̃, ending at γ̃(1) + δ̃(1).
3. Surjective. The loop γ_n(s) = (cos 2πns, sin 2πns) winds n times and has γ̃_n(1) = n.
4. Injective. A loop with winding number 0 lifts to a loop in ℝ (which is contractible), so it is null-homotopic upstairs and downstairs.

The same idea — analyse loops by lifting them to a simply-connected cover — gives the fundamental group of any space with a sufficiently nice covering theory.

Computing π_1 with van Kampen's theorem

For spaces built up by gluing, the Seifert-van Kampen theorem gives a recipe. Suppose X = U ∪ V where U, V are open and U ∩ V is path-connected. Pick a basepoint in U ∩ V. Then

π_1(X) = π_1(U) ∗_{π_1(U∩V)} π_1(V)

— the amalgamated free product. Loops in U and V are independent generators, except that any loop sitting in the intersection contributes the same element regardless of which side you read it from, which gives one relation per generator of π_1(U ∩ V).

For example, the figure-8 wedge X = S¹ ∨ S¹ is the union of two open neighbourhoods of the loops, intersecting in a contractible neighbourhood of the wedge point. Then π_1(U) = ⟨a⟩ ≅ ℤ, π_1(V) = ⟨b⟩ ≅ ℤ, π_1(U ∩ V) = {e}, so π_1(X) = ⟨a⟩ ∗ ⟨b⟩ = F_2 — the free group on two generators with no relations. Loops in a figure-8 are arbitrary words in a, b, a⁻¹, b⁻¹.

For the torus, two open neighbourhoods plus an annular intersection give one relation aba⁻¹b⁻¹ = 1 — the intersection's loop reads as the commutator. So π_1(T²) = ⟨a, b | aba⁻¹b⁻¹⟩ ≅ ℤ ⊕ ℤ. The Klein bottle has the related relation abab⁻¹ = 1, which makes the group non-abelian.

Functoriality and topological invariance

π_1 is not just a single group attached to a space — it is a functor from pointed topological spaces to groups. A continuous map f : X → Y with f(x_0) = y_0 induces a homomorphism f_∗ : π_1(X, x_0) → π_1(Y, y_0) by sending [γ] to [f ∘ γ]. Composition of maps becomes composition of homomorphisms; the identity becomes the identity.

This functoriality is the source of π_1's invariant power. If X and Y are homotopy equivalent (each is a continuous deformation retract of the other), then π_1(X) and π_1(Y) are isomorphic. So if you can compute π_1(X) ≠ π_1(Y), you've proved X and Y are not homotopy equivalent — let alone homeomorphic.

The classical application: the 2-sphere and the torus are not homeomorphic, because π_1(S²) = {e} ≠ ℤ² = π_1(T²). Trying to prove this directly without an algebraic invariant — by exhibiting some bijection that fails to be continuous — is a much harder argument.

Where the fundamental group shows up

• Knot theory. Two knots K, K' are equivalent if there's an ambient homeomorphism of ℝ³ taking one to the other. Their knot groups are π_1(ℝ³ ∖ K) and π_1(ℝ³ ∖ K'). For the unknot the group is ℤ; for the trefoil knot the group is ⟨a, b | aba = bab⟩, which has elements of finite order in its abelianisation only at 6 — distinguishing the trefoil from the unknot.
• Covering space theory. Connected covering spaces of X are classified by subgroups of π_1(X) up to conjugacy. The universal cover corresponds to the trivial subgroup. This gives a complete dictionary between topology and group theory; the lifting criterion for maps becomes a question about subgroup containment.
• Galois theory in topology. The covering-space classification mirrors the Galois correspondence between intermediate fields and subgroups of a Galois group. Étale fundamental groups in algebraic geometry generalise this to schemes; Grothendieck's program turned topology and number theory into a single subject.
• Quantum field theory and anyons. In 2D, exchanges of identical particles trace loops in the configuration space whose fundamental group is the braid group B_n. The "anyon" particle statistics observed in quantum Hall systems implement non-trivial representations of B_n. In 3D, the configuration-space fundamental group is the symmetric group S_n — only bosons and fermions.
• Robot motion planning. The configuration space of a robot is a topological space; two configurations are reachable along a path iff they lie in the same path component, and obstacle-free paths between them are classified up to homotopy by π_1 of the free configuration space. RRT-style planners implicitly enumerate elements of π_1 when topology forces detours.

Higher homotopy and the Hopf fibration

Beyond loops one can ask about maps from higher spheres. The n-th homotopy group π_n(X, x_0) is the set of based-homotopy classes of maps S^n → X. For n ≥ 2 these groups are abelian (the Eckmann-Hilton argument: two binary operations on the same set that distribute over each other are equal and commutative).

The most famous higher-homotopy result is that π_3(S²) = ℤ — the 3-sphere maps non-trivially onto the 2-sphere, generated by the Hopf fibration. The Hopf map h : S³ → S² has each fibre h⁻¹(p) a circle, and any two distinct fibres are linked once in S³. The integer in π_3(S²) is the linking number of two preimage fibres.

Higher homotopy groups of spheres π_n(S^k) for n > k are notoriously difficult — the table is incomplete past about n = k + 30, and many entries are sporadic finite groups. By contrast, the fundamental group is computable for any reasonable space, which is why π_1 dominates undergraduate topology.

Variants and extensions

• Fundamental groupoid. Drop the basepoint and consider all paths up to homotopy. Get a groupoid (a category in which every morphism is invertible) whose objects are points of X. More flexible for Seifert-van Kampen-type theorems and for non-path-connected spaces.
• Homology groups H_n(X). Abelianised homotopy: H_1(X) is the abelianisation of π_1(X). For most computational purposes — Euler characteristics, intersection numbers, characteristic classes — homology is easier than homotopy and still distinguishes most pairs of spaces.
• Étale fundamental group. Algebraic-geometric analogue defined for schemes, classifying étale covers. Grothendieck showed it equals the profinite completion of the topological π_1 for complex algebraic varieties.
• Mapping class group. Loops in the space of self-homeomorphisms of a surface. For a genus-g surface, this is a group with rich combinatorial structure (the Dehn twist generators) and underwrites moduli spaces of Riemann surfaces and Teichmüller theory.
• Whitehead's theorem. A map between CW complexes that induces isomorphisms on all homotopy groups π_n is a homotopy equivalence. So homotopy theory is "captured" by all the π_n collectively — but not by any one of them.

Common pitfalls

• Confusing path-connected with simply connected. Path-connected means every two points are joined by a path. Simply connected means additionally that every loop contracts. The plane minus a point is path-connected but not simply connected; π_1 = ℤ.
• Forgetting the basepoint. π_1 is defined relative to a basepoint. Without path-connectedness the answer depends on which component you're in. Even within a path-connected component the isomorphism between π_1 at different basepoints is non-canonical.
• Treating π_1 as automatically abelian. π_1(X) is non-abelian in general — the figure-8 has free group on two letters, the Klein bottle has a non-abelian group, knot complements are usually non-abelian. Higher homotopy groups π_n for n ≥ 2 are abelian; π_1 isn't.
• Reading "trivial fundamental group" as "trivial topology". S² has trivial π_1 but non-trivial higher homotopy and rich algebraic topology. Simply connected ≠ contractible. ℝ^n is contractible (all π_n trivial); S² is simply connected but not contractible.
• Misusing van Kampen on non-open or non-path-connected pieces. Both U and V must be open; the intersection must be path-connected. If U ∩ V has multiple components you need the version with groupoids. Many "easy" applications go wrong by ignoring these hypotheses.