Book 7A

© 2026 NeckPinch · www.neckpinch.com · All rights reserved.

Course 7Book 7A: Topology and the Fundamental GroupChapter 5

Computing π₁

Seifert–van Kampen, surfaces, projective spaces and knots.

20 min read · Updated Oct 3, 2026

Read with Lee, Introduction to Topological Manifolds, chapter 9 (some group theory: free groups, presentations, free products) and chapter 10 (the Seifert–van Kampen theorem and its applications to graphs, cell complexes, surfaces and knots), or Hatcher, Algebraic Topology, section 1.2.

In this chapter · 7 sections
  1. 5.1Knotted DNA
  2. 5.2Free groups and presentations
  3. 5.3The Seifert–van Kampen theorem
  4. 5.4Surfaces, projective spaces and the torus
  5. 5.5Knot groups
  6. 5.6History
  7. 5.7Exercises

The circle is the only space whose fundamental group 7A.4 The Fundamental Group computed from scratch. Most spaces of interest are built by gluing simpler pieces, and the Seifert–van Kampen theorem computes the fundamental group of a union from the fundamental groups of the pieces and of their overlap. With it, every surface's fundamental group follows from a single picture, and so does the fundamental group of the complement of a knot. The results are given by presentations: generators, one for each independent loop, and relations, one for each way of filling a loop in.

The mechanism to remember is that gluing in a disc along a loop kills that loop, and gluing in higher-dimensional cells changes nothing. So the fundamental group of a cell complex is determined by its 1-cells and 2-cells. For three-manifolds this means it can be read off from a Heegaard diagram (10A.2 Building Three-Manifolds).

By the end of this chapter you will be able to:

  • work with free groups, presentations and free products, and abelianise a presentation;
  • state the Seifert–van Kampen theorem and apply it to wedges of circles and to attaching cells;
  • compute π1\pi_1 of every closed surface, of RPn\mathbb{R}P^n, and of the torus;
  • write the Wirtinger presentation of a knot group, and show that the trefoil is knotted;
  • explain why the fundamental group distinguishes all closed surfaces.

Knotted DNA

In the world Model Topoisomerases and DNA knots

Many DNA molecules in cells are closed loops: the circular chromosomes of bacteria, plasmids, and the DNA of some viruses. A closed loop of DNA in three-dimensional space is a knot, and it can also be linked with other loops. Its knot type cannot change by any motion of the molecule in which the strands do not pass through each other. Yet cells must untangle DNA to copy and divide it, and the enzymes that do so, type II topoisomerases, work by cutting both strands of one segment of the double helix, passing another segment through the gap, and resealing: they change the knot type by a crossing change. Other enzymes, site-specific recombinases, cut and reconnect strands in ways that produce specific knots and links, and the knot types of the products have been used to deduce how the enzymes act.

Knot types can be told apart in the laboratory: a knotted circular DNA molecule of a given length is more compact than an unknotted one, and in gel electrophoresis knots of different types travel different distances. Andrzej Stasiak and colleagues showed that the distance travelled is closely correlated with the average crossing number of the knot's tightest configuration ("Electrophoretic mobility of DNA knots", Nature, 1996). Proteins can be knotted too: William Taylor found a deep figure-eight knot in the backbone of a plant enzyme in 2000 (Nature, 2000), and many knotted proteins are now known.

Mathematically, a knot is distinguished from the unknot by its knot group, the fundamental group of its complement in space. The complement of a knot is a three-manifold (10A.1 A Zoo of Three-Manifolds), and this chapter computes its fundamental group from a picture of the knot.

Free groups and presentations

The free group F(S)F(S) on a set SS consists of reduced words in the letters s±1s^{\pm1}, s∈Ss \in S (words with no adjacent ss−1ss^{-1} or s−1ss^{-1}s), multiplied by concatenation and cancellation. The free group on one letter is Z\mathbb{Z}; on two letters aa, bb it is non-abelian (ab≠baab \neq ba) and enormous. Every group generated by SS is a quotient of F(S)F(S).

A presentation ⟨S∣R⟩\langle S \mid R\rangle, with RR a set of words, is the quotient of F(S)F(S) by the normal closure of RR, the smallest normal subgroup containing RR: the "largest" group generated by SS in which the relations r=1r = 1 hold. For example ⟨a∣an⟩≅Z/n\langle a \mid a^n\rangle \cong \mathbb{Z}/n and ⟨a,b∣aba−1b−1⟩≅Z2\langle a, b \mid aba^{-1}b^{-1}\rangle \cong \mathbb{Z}^2. Deciding whether two presentations give isomorphic groups is in general undecidable (Adian and Rabin, 1955–58), which is part of why three-manifold topology is hard.

The abelianisation Gab=G/[G,G]G^{\mathrm{ab}} = G/[G, G] forces all generators to commute; for a finite presentation it is a finitely generated abelian group computed by integer linear algebra on the relations. It is the first homology group H1H_1 (7A.8 Homology in Brief).

The free product G∗HG * H is the group of reduced words alternating between non-identity elements of GG and of HH; if G=⟨S∣R⟩G = \langle S \mid R\rangle and H=⟨S′∣R′⟩H = \langle S' \mid R'\rangle, then G∗H=⟨S∪S′∣R∪R′⟩G * H = \langle S\cup S' \mid R\cup R'\rangle. Given homomorphisms i:K→Gi : K \to G and j:K→Hj : K \to H, the amalgamated free product G∗KHG *_K H adds the relations i(k)=j(k)i(k) = j(k) for all k∈Kk \in K.

The Seifert–van Kampen theorem

Theorem 5.1 Seifert–van Kampen

Let X=U∪VX = U\cup V with UU, VV open, and UU, VV and U∩VU\cap V path-connected, all containing the basepoint x0x_0. Then the homomorphism π1(U)∗π1(V)→π1(X)\pi_1(U) * \pi_1(V) \to \pi_1(X) induced by the inclusions is surjective, and its kernel is the normal subgroup generated by the elements i∗(ω)j∗(ω)−1i_*(\omega)j_*(\omega)^{-1} for ω∈π1(U∩V)\omega \in \pi_1(U\cap V), where ii, jj are the inclusions of U∩VU\cap V into UU and VV. That is,

π1(X)≅π1(U)∗π1(U∩V)π1(V).\pi_1(X) \cong \pi_1(U) *_{\pi_1(U\cap V)}\pi_1(V).

Proof. (Sketch; Lee, chapter 10, or Hatcher, section 1.2.) Surjectivity. Given a loop γ\gamma in XX, the open sets γ−1(U)\gamma^{-1}(U) and γ−1(V)\gamma^{-1}(V) cover II, so by the Lebesgue number lemma (2B.3 Compactness) there is a subdivision 0=t0<⋯<tm=10 = t_0 < \dots < t_m = 1 with each piece γ([tk−1,tk])\gamma([t_{k-1}, t_k]) in UU or in VV. Join each γ(tk)\gamma(t_k) to x0x_0 by a path in U∩VU\cap V (possible since it is path-connected and contains both points, when γ(tk)\gamma(t_k) lies in U∩VU\cap V, which can be arranged). Inserting these paths and their reverses writes γ\gamma as a product of loops each lying in UU or in VV. The kernel. A loop in U∩VU\cap V can be read either as a loop in UU or as one in VV, so the stated elements are in the kernel. That nothing else is, is proved by subdividing a null-homotopy I×I→XI\times I \to X into small squares, each mapped into UU or VV, and moving across it one square at a time.

Wedges of circles. The wedge S1∨S1S^1\vee S^1 (two circles joined at a point, a figure eight) is covered by two open sets, each a circle with a short whisker of the other, whose intersection is a contractible cross. So π1(S1∨S1)≅Z∗Z=F(a,b)\pi_1(S^1\vee S^1) \cong \mathbb{Z} * \mathbb{Z} = F(a, b), free on the two loops. The same works for any number of circles, and for any connected graph: its fundamental group is free, on as many generators as edges outside a spanning tree.

Corollary 5.2 Attaching cells

Let YY be obtained from a path-connected space XX by attaching a 2-dimensional disc along a loop α:S1→X\alpha : S^1 \to X. Then π1(Y)≅π1(X)/N\pi_1(Y) \cong \pi_1(X)/N, where NN is the normal subgroup generated by the class of α\alpha. Attaching an nn-dimensional cell with n≥3n \geq 3 doesn't change the fundamental group.

Proof. Take UU = YY minus the centre of the disc, which deformation retracts onto XX, and VV = the open disc, which is contractible; U∩VU\cap V is a punctured disc, with π1=Z\pi_1 = \mathbb{Z} generated by a loop around the centre, which is homotopic in UU to α\alpha. Van Kampen gives π1(X)∗1\pi_1(X) * 1 modulo the normal closure of [α][\alpha]. For an nn-cell with n≥3n \geq 3, U∩V≃Sn−1U\cap V \simeq S^{n-1} is simply connected (7A.4 The Fundamental Group), so nothing is added or killed.

Surfaces, projective spaces and the torus

A closed surface described by a polygon word (7A.3 Manifolds and Surfaces) is a cell complex: after gluing, the boundary of the polygon becomes a wedge of circles, one for each letter (all corners are identified in the normal forms), and the polygon is a 2-cell attached along the word. So:

surface presentation of π1\pi_1 abelianisation
S2S^2 trivial 00
torus T2T^2 ⟨a,b∣aba−1b−1⟩≅Z2\langle a, b \mid aba^{-1}b^{-1}\rangle \cong \mathbb{Z}^2 Z2\mathbb{Z}^2
genus gg, Σg\Sigma_g ⟨a1,b1,…,ag,bg∣[a1,b1]⋯[ag,bg]⟩\langle a_1, b_1, \dots, a_g, b_g \mid [a_1, b_1]\cdots[a_g, b_g]\rangle Z2g\mathbb{Z}^{2g}
RP2\mathbb{R}P^2 ⟨a∣a2⟩≅Z/2\langle a \mid a^2\rangle \cong \mathbb{Z}/2 Z/2\mathbb{Z}/2
Nk=#kRP2N_k = \#^k\mathbb{R}P^2 ⟨a1,…,ak∣a12⋯ak2⟩\langle a_1, \dots, a_k \mid a_1^2\cdots a_k^2\rangle Zk−1⊕Z/2\mathbb{Z}^{k-1}\oplus\mathbb{Z}/2

Here [a,b]=aba−1b−1[a, b] = aba^{-1}b^{-1} is the commutator. The abelianisations are all different, so the fundamental group distinguishes every closed surface from every other, which completes the proof that the surfaces in the classification theorem are pairwise distinct (7A.3 Manifolds and Surfaces). And only S2S^2 has trivial fundamental group: a simply connected closed surface is a sphere.

Projective spaces. RPn\mathbb{R}P^n has a cell structure with one cell in each dimension 0,1,…,n0, 1, \dots, n; the 2-cell is attached to the circle RP1\mathbb{R}P^1 by a map of degree 22. So π1(RPn)=⟨a∣a2⟩=Z/2\pi_1(\mathbb{R}P^n) = \langle a \mid a^2\rangle = \mathbb{Z}/2 for every n≥2n \geq 2, by Corollary 5.2. In particular π1(RP3)=Z/2\pi_1(\mathbb{R}P^3) = \mathbb{Z}/2, a Path milestone that 7A.6 Covering Spaces proves again with covering spaces and interprets physically.

Figure 5.1. Van Kampen for the torus. UU is the torus minus a point, which deformation retracts onto the wedge of the two circles aa and bb (the glued edges of the square), so π1(U)=F(a,b)\pi_1(U) = F(a, b). VV is a disc. U∩VU\cap V is an annulus, whose generator is the loop around the puncture, homotopic in UU to aba−1b−1aba^{-1}b^{-1}. Hence π1(T2)=⟨a,b∣aba−1b−1⟩≅Z2\pi_1(T^2) = \langle a, b \mid aba^{-1}b^{-1}\rangle \cong \mathbb{Z}^2.

Knot groups

A knot is an embedded circle K⊂R3K \subset \mathbb{R}^3 (or S3S^3), and its knot group is π1(R3∖K)\pi_1(\mathbb{R}^3\setminus K). Equivalent knots, those that can be deformed into each other through embeddings, have homeomorphic complements and isomorphic groups. The unknot's complement deformation retracts onto a circle linking it, so its group is Z\mathbb{Z}.

The Wirtinger presentation. Draw the knot as a diagram in the plane with breaks at undercrossings, so that it falls into arcs. Give the knot an orientation. Each arc kk contributes a generator xkx_k: a loop from a basepoint above the plane that dips down under the arc and back, going around it in the right-handed sense. At each crossing, where an overcrossing arc xkx_k separates two undercrossing arcs xix_i and xjx_j, there is one relation,

xj=xkxixk−1orxj=xk−1xixk,x_j = x_kx_ix_k^{-1}\qquad\text{or}\qquad x_j = x_k^{-1}x_ix_k,

depending on the sign of the crossing. (Van Kampen, applied to the complement cut along the plane of the diagram, proves this.) One of the relations always follows from the others.

For the trefoil, with three arcs xx, yy, zz and three crossings (Figure 5.2), the Wirtinger presentation reduces to

π1(R3∖trefoil)≅⟨x,y∣xyx=yxy⟩≅⟨a,b∣a2=b3⟩,\pi_1(\mathbb{R}^3\setminus\text{trefoil}) \cong \langle x, y \mid xyx = yxy\rangle \cong \langle a, b \mid a^2 = b^3\rangle,

the second form by a=xyxa = xyx, b=xyb = xy (Exercise 5.5). This group is non-abelian: sending x↦(1 2)x \mapsto (1\ 2) and y↦(2 3)y \mapsto (2\ 3) in the symmetric group S3S_3 respects the relation and is onto a non-abelian group (Exercise 5.6). Since the unknot's group is Z\mathbb{Z}, abelian, the trefoil is knotted: no deformation through embeddings untangles it. Its abelianisation is Z\mathbb{Z}, the same as the unknot's; every knot group has abelianisation Z\mathbb{Z}, so the non-abelian structure is essential.

Figure 5.2. The trefoil, drawn from the parametrisation (sin⁡t+2sin⁡2t, cos⁡t−2cos⁡2t)(\sin t + 2\sin2t,\ \cos t - 2\cos2t) with heights −sin⁡3t-\sin3t deciding which strand passes over (computed). Its three arcs give the Wirtinger generators xx, yy, zz; the three crossings give relations such as z=xyx−1z = xyx^{-1}, which reduce to xyx=yxyxyx = yxy.
Where this goes Knot complements and three-manifolds

The complement of a knot in S3S^3 is a three-manifold, and removing a neighbourhood of the knot leaves a compact three-manifold whose boundary is a torus. Cutting out such a neighbourhood and gluing it back differently, Dehn surgery, produces new closed three-manifolds; every closed orientable three-manifold arises by surgery on some link in S3S^3 (10A.2 Building Three-Manifolds). Knot complements are also the source of many hyperbolic three-manifolds (10A.6 Hyperbolic Three-Manifolds): the complement of the figure-eight knot carries a complete hyperbolic metric, and its geometry is determined by its fundamental group (Mostow rigidity). For three-manifolds the fundamental group carries an enormous amount of information, which is one reason the Poincaré conjecture is stated in terms of it.

History

Walther Dyck introduced group presentations in 1882 and Heinrich Tietze developed them in 1908. Max Dehn studied knot groups and proved in 1910 that the trefoil is knotted and distinct from its mirror image (the latter in 1914). Wilhelm Wirtinger described his presentation in lectures in 1905; it was published by Emil Artin in 1925. Herbert Seifert (1931) and Egbert van Kampen (1933) proved the theorem that bears their names. Sergei Adian (1955–57) and Michael Rabin (1958) proved that most properties of finitely presented groups, including triviality, are undecidable. Stasiak and colleagues' work on DNA knot mobility appeared in 1996 and Taylor's knotted protein in 2000.

Recall Where we stand

Groups are presented by generators and relations; free groups have no relations, and abelianisation makes generators commute. Seifert–van Kampen computes π1(U∪V)\pi_1(U\cup V) as the amalgamated free product of π1(U)\pi_1(U) and π1(V)\pi_1(V) over π1(U∩V)\pi_1(U\cap V). Wedges of circles have free fundamental groups; attaching a 2-cell along a loop kills that loop, and higher cells change nothing. So π1(Σg)=⟨ai,bi∣∏[ai,bi]⟩\pi_1(\Sigma_g) = \langle a_i, b_i \mid \prod[a_i, b_i]\rangle, π1(RPn)=Z/2\pi_1(\mathbb{R}P^n) = \mathbb{Z}/2 for n≥2n \geq 2, and the fundamental group distinguishes all closed surfaces, with only S2S^2 simply connected. Knot groups have Wirtinger presentations; the trefoil group ⟨x,y∣xyx=yxy⟩\langle x, y \mid xyx = yxy\rangle maps onto S3S_3, so the trefoil is knotted. 7A.6 Covering Spaces relates fundamental groups to covering spaces.

Exercises

Exercise 5.3 The figure eight

Show that π1(S1∨S1)\pi_1(S^1\vee S^1) is non-abelian by finding a homomorphism onto S3S_3. Show that the loops aa, bb, aba−1b−1aba^{-1}b^{-1} are all non-trivial.

Solution

Send a↦(1 2)a \mapsto (1\ 2), b↦(2 3)b \mapsto (2\ 3); since the group is free, any assignment extends to a homomorphism. The image is generated by two transpositions, hence all of S3S_3, which is non-abelian, so ab≠baab \neq ba in the free group. Since transpositions are their own inverses, aba−1b−1↦((1 2)(2 3))2aba^{-1}b^{-1} \mapsto \big((1\ 2)(2\ 3)\big)^2; the product of two transpositions sharing one point is a 3-cycle, and the square of a 3-cycle is a 3-cycle, so the image is non-trivial.

Exercise 5.4 The genus-2 surface

Write down π1(Σ2)\pi_1(\Sigma_2) from the octagon word a1b1a1−1b1−1a2b2a2−1b2−1a_1b_1a_1^{-1}b_1^{-1}a_2b_2a_2^{-1}b_2^{-1}. Show that it maps onto F(a1,a2)F(a_1, a_2) (set b1=b2=1b_1 = b_2 = 1), so it is non-abelian, and compute its abelianisation.

Exercise 5.5 Two forms of the trefoil group

In ⟨x,y∣xyx=yxy⟩\langle x, y \mid xyx = yxy\rangle, let a=xyxa = xyx and b=xyb = xy. (a) Show a2=b3a^2 = b^3. (b) Show that x=b−1ax = b^{-1}a and y=a−1b2y = a^{-1}b^2, so aa, bb generate, and that the relation a2=b3a^2 = b^3 implies xyx=yxyxyx = yxy. Conclude that the two presentations give isomorphic groups.

Solution

(a) b3=xyxyxy=(xyx)(yxy)=(xyx)(xyx)=a2b^3 = xyxyxy = (xyx)(yxy) = (xyx)(xyx) = a^2, using yxy=xyxyxy = xyx. (b) b−1a=y−1x−1xyx=xb^{-1}a = y^{-1}x^{-1}xyx = x, and a−1b2=x−1y−1x−1xyxy=x−1xy=ya^{-1}b^2 = x^{-1}y^{-1}x^{-1}xyxy = x^{-1}xy = y. Conversely, with x=b−1ax = b^{-1}a, y=a−1b2y = a^{-1}b^2: xyx=b−1a a−1b2 b−1a=axyx = b^{-1}a\,a^{-1}b^2\,b^{-1}a = a and yxy=a−1b2b−1aa−1b2=a−1b3=a−1a2=ayxy = a^{-1}b^2b^{-1}aa^{-1}b^2 = a^{-1}b^3 = a^{-1}a^2 = a.

Exercise 5.6 The trefoil is knotted

Check that x↦(1 2)x \mapsto (1\ 2), y↦(2 3)y \mapsto (2\ 3) satisfies xyx=yxyxyx = yxy in S3S_3, and conclude that the trefoil group is non-abelian, so the trefoil is not the unknot. Show also that every knot group has abelianisation Z\mathbb{Z} (in the Wirtinger presentation, abelianising makes all generators equal).

Solution

(1 2)(2 3)(1 2)=(1 3)=(2 3)(1 2)(2 3)(1\ 2)(2\ 3)(1\ 2) = (1\ 3) = (2\ 3)(1\ 2)(2\ 3). In the abelianisation, each relation xj=xkxixk−1x_j = x_kx_ix_k^{-1} becomes xj=xix_j = x_i; the arcs of a connected diagram are linked by crossings, so all generators become equal, leaving one generator and no relation: Z\mathbb{Z}.

Exercise 5.7 The Klein bottle group

From the word abab−1abab^{-1}, π1(K)=⟨a,b∣abab−1⟩\pi_1(K) = \langle a, b \mid abab^{-1}\rangle. Show its abelianisation is Z⊕Z/2\mathbb{Z}\oplus\mathbb{Z}/2, and compare with the torus.

Exercise 5.8 Rehearsal: π1\pi_1 from a Heegaard splitting

A closed three-manifold MM can be built from a solid torus V=D2×S1V = D^2\times S^1 by attaching a 2-cell along a curve α\alpha on its boundary torus and then a 3-cell; the curve α\alpha is the meridian of the second solid torus in a genus-1 Heegaard splitting (7A.2 Compactness and Compactification, 10A.2 Building Three-Manifolds). (a) Using Corollary 5.2, show π1(M)=Z/⟨n⟩\pi_1(M) = \mathbb{Z}/\langle n\rangle, where nn is the number of times α\alpha winds around the core of VV (its class in π1(V)=Z\pi_1(V) = \mathbb{Z}). (b) For S3S^3 (two solid tori, meridian glued to longitude), n=1n = 1: π1(S3)=1\pi_1(S^3) = 1. (c) For S2×S1S^2\times S^1 (meridian glued to meridian), n=0n = 0: π1=Z\pi_1 = \mathbb{Z}. (d) For the lens space L(p,q)L(p, q), α\alpha winds pp times around the core: π1=Z/p\pi_1 = \mathbb{Z}/p.

Solution

(a) VV deformation retracts onto its core circle, so π1(V)=Z\pi_1(V) = \mathbb{Z}, generated by the core. Attaching the 2-cell along α\alpha kills [α]=n[\alpha] = n times the generator, giving Z/n\mathbb{Z}/n (Z\mathbb{Z} if n=0n = 0); the 3-cell changes nothing. (b)–(d) follow by substituting n=1n = 1, 00, pp.

© 2026 NeckPinch (www.neckpinch.com). All content in the guidebook (text, mathematics, figures and exercises) is protected by copyright. All rights reserved. No part may be copied, republished or redistributed without written permission.