Various topics in rack and quandle homology
April 14, 2010
Master’s thesis in Mathematics Jorik Mandemaker
supervisor: Dr. F.J.-B.J. Clauwens second reader: Prof. Dr. F. Keune student number: 0314145 Faculty of Science (FNWI) Radboud University Nijmegen Contents
1 Introduction 5 1.1 Racks and quandles ...... 6 1.2 Quandles and knots ...... 7 1.3 Augmented quandles and the adjoint group ...... 8 1.3.1 The adjoint group ...... 9 1.4 Connected components ...... 9 1.5 Rack and quandle homology ...... 10 1.6 Quandle coverings ...... 10 1.6.1 Extensions ...... 11 1.6.2 The fundamental group of a quandle ...... 12 1.7 Historical notes ...... 12
2 The second cohomology group of alexander quandles 14 2.1 The adjoint group of Alexander quandles ...... 15 2.1.1 Connected Alexander quandles ...... 15 2.1.2 Alexander quandles with connected components . . . . . 18 2.2 The second cohomology group of connected Alexander quandles . 19
3 Polynomial cocycles 22 3.1 Introduction ...... 23 3.2 Polynomial cochains ...... 23 3.3 A decomposition and filtration of the complex ...... 24 3.4 The 3-cocycles ...... 25 3.5 Variable Reduction ...... 27 3.6 Proof of surjectivity ...... 28 3.6.1 t = s ...... 29 3.6.2 t < s ...... 30 3.7 Final steps ...... 33
4 Rack and quandle modules 35 4.1 Rack and Quandle Modules ...... 36 4.1.1 Definitions ...... 36 4.1.2 Some examples ...... 38 4.2 Reduced rack modules ...... 39
1 4.3 Right modules and tensor products ...... 40 4.4 The rack algebra ...... 42 4.5 Examples ...... 44 4.5.1 A free resolution for trivial quandles ...... 44 4.5.2 R3 ...... 44
5 Trunks 46 5.1 Introduction ...... 47 5.2 Trunks ...... 47 5.3 -sets and their realizations ...... 48 5.4 The nerve of a trunk ...... 49
6 Homology operators 51 6.1 The augmented rack space ...... 52 6.1.1 H1(BInnX) ...... 54 6.2 Homology operators of negative degree ...... 55 0 6.2.1 H (BGX X)...... 58 6.3 Variations ...... 59 6.3.1 A different G ...... 59 6.3.2 Quandle homology ...... 59 6.4 Relation to the literature ...... 60 6.4.1 Some computer calculations ...... 60 6.5 Some other operators ...... 61
7 The cup product of the rack space 63 7.1 The cup product of the realization of a -set ...... 64 7.2 A triangulation of the realization of a -set ...... 65 7.3 The cup product ...... 67
2 Summary
The goal of this thesis project was originally to survey the current literature on racks and quandles in order to comment on it or extend it. Eventually this resulted in this thesis being a mix of various loosely related topics. Since rack and quandle theory is a rather obscure part of mathematics this thesis contains a general introduction to racks and quandles covering all the ba- sics. The first chapter contains this introduction and can generally be regarded as ”common” knowledge about quandles. The second and third chapter are both concerned with computing the co- homology of certain Alexander quandles. The former being an application of the work by Eisermann [7]. The calculation in 2.1.1 I owe to my supervisor F. Clauwens. But the rest of the chapter is my own work. The third chapter is a correction of the work by Mochizuki [17, 18]. In these articles Mochizuki claims to have found a basis for the third cohomology group of some Alexander quandles. However this list contains a coboundary, so it is clearly incorrect. The next chapter focuses on the rack and quandle modules from the work by Jackson [12, 13]. Jackson studied these modules as an alternate way of defining rack and quandle homology. For connected quandles we’ll show that those modules coincide with modules over a subring of the group ring of the adjoint group of the quandle. Chapter 5 is another introductionary chapter which serves mostly to set the terminology used in the final two chapters. It covers the basics of trunks and their classifying spaces as found in [9]. Chapter 6 explores the homology of the augmented rack space which proves a useful tool to study the homology of racks. It also shines new light on some of the work by Niebrzydowski and Przytycki [19]. The orginal idea that the augmented rack spaces is a topological monoid that produces homology operators is once again due to F. Clauwens. Section 6.1.1 and further is my own work with the exception of formula 6.6. The final chapter studies the yet unexplored the cup product on the coho- mology of -sets which in particular covers that of racks and quandles.
3 Acknowledgements
I would like to thank my thesis supervisor Frans Clauwens who was very patient with me over the course of this thesis project. He worked tirelessly on racks and quandles himself. He often pointed me in the right direction and explained many interesting things to me. In particular I have to thank him for the computation of chapter 2.1.1 and the insight that the cycles and cocycles of BGX produce homology operators (chapter 6). I would also like to thank Frans Keune for being the second reader of my thesis.
4 Chapter 1
Introduction
5 1.1 Racks and quandles
Definition 1.1.1. A quandle (Q, B, C) is a set Q with two binary operators B and C such that the following axioms hold. (Q1) a B a = a for all a in Q (Q2) (a C b) B b = a = (a B b) C b for all a, b in Q
(Q3) (a B b) B c = (a B c) B (b B c) for all a, b, c in Q A set R with two binary operators satisfying axioms (Q2) and (Q3) but not necessarily (Q1) is called a rack.
If (Q, B, C) is a quandle or rack than the operation C is uniquely determined by B. Because of this we will usually omit it from the notation and simply write (Q, B) or even just Q. Definition 1.1.2. A quandle homomorphism between two quandles X and Y is a function φ : X → Y such that φ(aBb) = φ(a)Bφ(b) and φ(aCb) = φ(a)Cφ(b) for all a, b ∈ X. The composition of two quandle homomorphisms is again a quandle homomorphism so we can speak of the category Qnd with objects quandles and morphisms the quandle homomorphism. In the same way we obtain the category of racks denoted by Rack.
Convention 1.1.3. We let both B and C associate to the left. So when we write a B b C c B d we mean ((a B b) C c) B d
Theorem 1.1.1. If (Q, B, C) is a quandle then (Q, C, B) is also a quandle called the dual quandle. Definition 1.1.4. A quandle or rack Q is said to be involutive if it satisfies
(QInv) a B b B b = a for all a, b in Q
A quandle or rack Q is said to be abelian if it satisfies
(Qab) (a B x) B (y B b) = (a B y) B (x B b) for all a, b, x, y in Q −1 Example 1.1.1. Every group G becomes a quandle by setting a B b = b ab −1 and a C b = bab . This quandle is called the conjugation quandle of G denoted by Conj(G). Every group homomorphism between two groups G and H is also a quandle homomorphism between Conj(G) and Conj(H). Thus we obtain a functor Conj : Grp → Qnd.
Example 1.1.2. If X is any set then X becomes a quandle by aBb = aCb = a called the trivial quandle on X.
6 Example 1.1.3. If M is an abelian group and T : M → M is an automorphism of M. Then M becomes a quandle by setting
x B y = T (x) + (1 − T )(y) −1 x C y = T (x − (1 − T )(y)) This quandle is called the Alexander quandle Alex(M,T ).
Example 1.1.4. The dihedral quandle Rk is an Alexander quandle where the group is Zk and the automorphism is multiplication by −1. An alternate descrip- tion is that Rk is the subquandle of Conj(Dk) consisting of all the reflections. The automorphism group Aut(Q) of a quandle Q consists of all bijective quandle homomorphisms φ from Q to itself. We let this group act on Q from the right by evaluation notated by aφ = φ(a). This means that for φ, ψ ∈ Aut(Q) their composition φψ is given by a(φψ) = (aφ)ψ. For every element a ∈ Q we have an automorphism ρa defined by ρa(b) = b B a.
Definition 1.1.5. The subgroup of Aut(Q) generated by the ρa is called the inner automorphism group of Q. This group is notated by Inn(Q). We also have the map inn : Q → Inn(Q) defined by inn(a) = ρa. Convention 1.1.6. As a subgroup of Aut(Q) Inn(Q) acts on Q. We will often b inn(b) h −1 write a for a = a B b. We will also write g = h gh for conjugation in a group G in light of the quandle Conj(G).
1.2 Quandles and knots
Quandles are closely related to knots. Every oriented knot has a quandle asso- ciated with it, called the fundamental quandle of the knot. It’s easy to read of a knot diagram.
Take one generator for each arc and for each crossing like the one pictured above add a relation of the form a B b = c or equivalently c C b = a. Note that the orientation of the arc crossing under is irrelevant. It is easy to see that this construction is independent on the knot diagram. In fact the three quandle axioms correspond directly with the three Reidemeister moves. For more details on this construction and a topological interpretation see [14].
7 1.3 Augmented quandles and the adjoint group
In the last section we considered a quandle Q and we had a map inn : Q → Inn(Q). Also Inn(Q) acted on Q. This map and action have the property that inn(b) φ φ a B b = a and inn(a ) = inn(a) . This situation can be generalized to give rise to the notion of an augmented quandle. Definition 1.3.1. A representation of a quandle (or rack) Q in a group G is a −1 map φ : Q → G such that φ(aBb) = φ(b) φ(a)φ(b) for all a, b in Q. Essentially φ is a quandle homomorphism between Q and Conj(G). Definition 1.3.2. Let φ : Q → G be a representation and let α be a right action of G on Q denoted by α(a, g) = ag. We call the pair (φ, α) an augmentation if φ(b) g g a B b = a and φ(a ) = φ(a) for all a, b ∈ Q and g ∈ G. This means that the following diagram commutes.
id×φ φ×id Q × Q / Q × G / G × G
B α conj id φ Q / Q / G
g g gφ(bg ) φ(b)g g Since for g ∈ G and a, b ∈ X we have a B b = a = a = (a B b) , we can view the action α as a group homomorphism α : G → Aut(Q) When we have a representation φ : Q → G where φ(Q) generates G and there is an α : G → Aut(Q) such that α ◦ φ = inn we just say that φ : Q → G is an augmentation.
8 1.3.1 The adjoint group Definition 1.3.3. For a rack or quandle Q we define Adj(Q) the adjoint group of Q as follows. Adj(X) = he : x ∈ Q | e e = e e i x x y y xBy So we have one generator for every element of Q and relations encompassing the quandle operation. We also obtain a map adj : Q → Adj(Q) by adj(x) = ex Note that we can rewrite the defining relation of the adjoint group as
e = e−1e e (1.1) xBy y x y So we see that adj : Q → Adj(Q) is a representation. It is in fact the universal representation of Q. For every representation φ : Q → G there exists a unique h : Adj(Q) → G such that φ = h ◦ adj. In particular there is a map ρ : Adj(Q) → Inn(Q) that sends ex to inn(x). We let Adj(Q) act on Q through this map. Combining this action with (1.1) we find that for all g ∈ Adj(Q) we have
g −1 (ex) = g exg = exg (1.2) Definition 1.3.4. Let Q be a quandle (or rack). Define K = ker(ρ : Adj(Q) → Inn(Q)). In light of (1.2) we see that K is the center of Adj(Q). Definition 1.3.5. We can turn Adj into a functor from Qnd → Grp. For 0 0 f : Q → Q define Adj(f) : Adj(Q) → Adj(Q ) by Adj(f)(ex) = ef(x). The name adjoint group comes from the fact that: Theorem 1.3.1. The functor Adj is the left adjoint to the functor Conj. In some papers (e.g. [13]) the adjoint group is called the associated group As(Q).
1.4 Connected components
Definition 1.4.1. Let Q be a quandle (or rack) and x ∈ Q. Denote the orbit of x under the action of Inn(Q) by [x]. There orbits are called the connected components of Q. We call Q connected if Inn(Q) acts transitively on Q. Also define π0(Q) = {[x] | x ∈ Q}.
We can view π0(Q) as a trivial quandle. If we do so, the map x 7→ [x] becomes a quandle homomorphism. In fact every quandle homomorphism from Q to some trivial quandle X factors through this map. In particular if Q is connected and 0 0 f : Q → Q is a quandle homomorphism then the map Q → π0(Q ), x 7→ [f(x)] factors through π0(Q). So when Q is connected its image is also connected. Since Adj(Q) acts on Q through inner automorphisms the Adj(Q)-orbits are the same as the connected components of Q. Using (1.2) we see that if x and y are in the same connected component then ex and ey are conjugate.
9 Definition 1.4.2. Let Q be a quandle (or rack). Define : Adj(Q) → Z by ◦ (ex) = 1. Denote ker by Adj(Q) . When Q is a quandle (not a rack!) the orbits under the action of Adj(Q)◦ are still the connected components. Since
e−(g)g g a a = a
1.5 Rack and quandle homology
Most of the research into quandles and racks has at least some link to the homology of racks and quandles. Most of this thesis is also concerned with this homology
R Definition 1.5.1. For a rack X let Cn (X) be the free abelian group generated n R R by X . Define a map δn : Cn (X) → Cn−1(X) as follows 0 δi (x1, . . . , xn) = (x1, . . . , xi−1, xi+1, . . . , xn) 1 δi (x1, . . . , xn) = (x1 B xi, . . . , xi−1 B xi, xi+1, . . . , xn) n X i 0 1 δn = (−1) [δi − δi ] i=1
R A simple computation verifies that δn−1 ◦ δn = 0 so that {C∗ (X), δ∗} forms a chain complex. This chain complex is called the rack complex of X. We usually omit the subscript on δ.
D R Definition 1.5.2. Let Q be a quandle. Define Cn (Q) ⊂ Cn (Q) as the subgroup generated by the (x1, . . . , xn) with xi = xi+1 for some 1 ≤ i < n. D D A simple computation shows that δ(Cn (Q)) ⊂ Cn−1(Q). This is not true in general when Q is just a rack and not a quandle. So we get a subcomplex D C∗ (Q) called the degeneracy complex. Q R D Definition 1.5.3. Let Q be a quandle. Define Cn (Q) = Cn (Q)/Cn (Q). δ Q descends to this quotient so we get a chain complex C∗ called the quandle complex. Also define
W W Hn (X) = Hn(C∗ (X)) n n W HW (X) = H (C∗ (X)) For W = R,D,Q. Of course for W = R this is also defined for racks.
1.6 Quandle coverings
The notion of quandle coverings was studied by Eisermann in [6],[7]. All of the definitions and theorems we mention here can be found in [7]. For details and proofs look there.
10 Definition 1.6.1. Let p : Q˜ → Q be a surjective quandle homomorphism. If ˜ for all a, x, y ∈ Q p(x) = p(y) implies a B x = a B y then we call p a covering. Definition 1.6.2. When p : Q˜ → Q is a covering then Q acts on Q˜. If x ∈ Q and a ∈ Q˜ we can define ax = ax˜ wherex ˜ ∈ Q˜ is such that p(˜x) = x. This does not depend on the choice ofx ˜. This action extends in the obvious way to Adj(Q).
Definition 1.6.3. Let p1 : Q1 → Q and p2 : Q2 → Q be two coverings. A covering morphism from p1 to p2 is a quandle homomorphism φ : Q1 → Q2 such that p1 = p2 ◦ φ. That is: the following diagram is commutative.
φ Q1 / Q2 @ @@ ~~ @@ ~~ p1 @@ ~~ p2 @ ~ ~ Q
Definition 1.6.4. All coverings over a given quandle Q together with the cov- ering morphisms between them form a category Cov(Q). Definition 1.6.5. The group of all covering automorphisms of a covering p : Q˜ → Q is denoted by Aut(p) and called the deck transformation group of p. We let Aut(p) act on Q˜ from the left. This action commutes with the action of Adj(Q˜). Definition 1.6.6. We call a covering p : Q˜ → Q galois if Q˜ is connected and Aut(p) acts freely and transitively on each fibre.
1.6.1 Extensions The case of galois coverings can be generalized to the notion of quandle exten- sions.
Definition 1.6.7. An extension E :Λ y Q˜ → Q of a quandle Q by a group Λ consists of a group action Λ y Q˜ and a surjective quandle homomorphism p : Q˜ → Q such that the following two axioms hold. ˜ 1. (λx) B y = λ(x B y) and x B (λy) = x B y for all λ ∈ Λ and x, y ∈ Q. 2. Λ acts freely and transitively on each fibre p−1(x). The homomorphism p in every extension is a quandle covering and when Q˜ is connected it is galois. Also every galois covering becomes an extension by the action of Aut(p) on Q˜. Definition 1.6.8. Let Λ be a group and Q be a quandle. An equivalence be- p1 p2 tween two extensions E1 :Λ y Q1 /Q and E2 :Λ y Q2 /Q is a cover- ing isomorphism φ : p1 → p2 such that λφ = φλ for all λ ∈ Λ. The set of all equivalence classes of extensions is denoted by Ext(Q, Λ).
11 1.6.2 The fundamental group of a quandle Definition 1.6.9. A pointed quandle (Q, q) is a quandle Q along with an ele- ment q ∈ Q chosen as a base point. A pointed quandle homomorphism is just a quandle homomorphism that preserves the base point. We also have pointed quandle coverings defined in the same way as regular coverings. Definition 1.6.10. A pointed quandle covering p :(Q,˜ q˜) → (Q, q) is called universal if for every pointed quandle covering p0 :(Q0, q0) → (Q, q) there exists a unique homomorphism φ :(Q,˜ q˜) → (Q0, q0) with p = p0 ◦ φ. Of course every two universal coverings are isomorphic. Definition 1.6.11. The universal covering (Q,˜ q˜) of a connected quandle (Q, q) is given by Q˜ = {(a, g) ∈ Q × Adj(Q)◦ | qg = a} −1 Withq ˜ = (q, 1) and (a, g) B (b, h) = (a B b, g · adj(a) · adj(b)). The covering p :(Q,˜ q˜) → (Q, q) is given by p(a, g) = a. For more details on this definition as well as proof that it is indeed universal see [7].
Definition 1.6.12. The fundamental group π1(Q, q) of a connected quandle (Q, q) is defined to be
◦ g π1(Q, q) = {g ∈ Adj(Q) | q = q}
Theorem 1.6.1. For every connected quandle (Q, q) and every group Λ we have
2 ∼ ∼ HQ(Q, Λ) = Ext(Q, Λ) = Hom(π1(Q, q), Λ) and also Q ∼ H2 (Q) = π1(Q, q)ab
1.7 Historical notes
The first time racks were studied in general was in 1959 in unpublished corre- spondence between J.C. Conway and G.C. Wraith [5]. They used the term wrack for the concept. The name comes from the English phrase ”wrack and ruin” in relation to the ’wrecking’ of a group by throwing away the multiplication and only retaining conjugation. They studied racks and quandles as algebraic objects but also knew of the fundamental quandle of a knot. In a less general way racks and quandles have been studied earlier. In 1942 M. Takasaki studied involutive quandles which he called keis [20]. In particular he studied dihedral quandles arising from symmetries of regular polygons. Even earlier in 1929 Burnstin and Mayer studied ”distributive groups” which are semigroups with a self left and right distributive group operation [2]. Quandles of course generalize these objects.
12 The first published work on racks is from 1982 by Joyce [14]. In this paper the term quandle was coined. This paper introduced many of the basic notions such as the connected components of quandles as well as augmented quandles and the adjoint group. Joyce also studied racks in a topological way. His main result is that the fundamental quandle of a knot is a classifying invariant of knots. In 1995 Fenn, Rourke and Sanderson introduced trunks in [9] leading to the homology theory for racks. The refinement to quandle homology comes from Carter et al. [3, 4]. The majority of the current articles related to racks and quandles deal with this homology theory. Either computing parts of it [7, 8, 15, 17, 19] or using it to construct knot and link invariants [1, 4, 10].
13 Chapter 2
The second cohomology group of alexander quandles
14 2 ∼ Since HQ(Q, Λ) = Hom(π1(Q, q), Λ) computing the adjoint group of a quan- dle will help computing the second cohomology group. In this chapter we’ll give a description of Adj(Q) for alexander quandles Q and then apply the theory from [7] to compute the second cohomology group.
2.1 The adjoint group of Alexander quandles 2.1.1 Connected Alexander quandles Let M be an abelian group and T : M → M an automorphism. We give M the structure of a quandle by setting x B y = T x + (1 − T )y for x, y ∈ M. Define C0 as the connected component of M containing 0. Since 0 B x = (1 − T )x it follows that (1 − T )M ⊂ C0. Since T ((1 − T )x) + (1 − T )y = (1 − T )(T x + y) it follows that (1 − T )M is closed under Inn(M), so C0 = (1 − T )M. Thus M is connected iff (1 − T ) is surjective. We will assume (1 − T ) to be invertible, which is automatic when M is finite. Recall that Adj(M) is the group with generators ex for x ∈ M and relations e = e−1e e . Note that for all p ∈ M xBy y x y
e−1e e−1e e−1e p 0 x+y = p + x − T x + y − T y = p 0 x 0 y (2.1)
So −1 −1 −1 e0 ex+y = γ(x, y)e0 exe0 ey (2.2) for some γ(x, y) ∈ K = ker(ρ : Adj(M) → Inn(M)). Recall that K is the center of Adj(M). We find that γ(x, 0) = γ(0, x) = 1 directly from the definition. We −1 can expand e0 ex+y+z in two ways using formula (2.2) to find γ(x, y + z)γ(y, z) = γ(x + y, z)γ(x, y) (2.3)
−1 −1 We can rewrite (2.2) as ex+y = γ(x, y)exe0 ey = γ(x, y)e0 eT −1xey, which can be written as −1 eT −1xey = γ(x, y) e0ex+y (2.4) It follows that
−1 γ(x, y) e0ex+y = eT −1xey = eyex+(1−T )y −1 = γ(T y, x + (1 − T )y) e0eT y+x+(1−T )y −1 = γ(T y, x + (1 − T )y) e0ex+y (2.5)
Here the first and third equality follow from (2.4) and the second is the defining relation of Adj(M). So we get
γ(x, y) = γ(T y, x + (1 − T y)) (2.6)
Filling in x = 0 we find γ(T y, (1 − T )y) = 1 (2.7)
15 We can apply (2.6) to compute that
γ(x, y)γ((1 − T )y, x) = γ(T y, x + (1 − T )y)γ((1 − T )y, x) = γ(T y + (1 − T )y, x)γ(T y, (1 − T )y) = γ(y, x) (2.8)
The second equation follows from (2.3). One would hope that γ is symmetric. Unfortunately this isn’t the case. But by looking at how much γ falls short of being symmetric we can find out more. Therefore we define
λ(u, v) = γ(u, v)−1γ(v, u) = γ((1 − T )v, u) (2.9)
Using (2.3) twice we find
γ(u, v + w)γ(v, w) = γ(u + v, w)γ(u, v) γ(w + v, u)γ(w, v) = γ(w, v + u)γ(v, u)
Dividing the second relation by the first we get
λ(u, v + w)λ(v, w) = λ(u + v, w)λ(u, v) (2.10)
Using (2.9) we get λ((1 − T )−1x, y)−1 = γ(x, y). Combining this with (2.3) we find
λ((1 − T )−1(x + y), w)λ((1 − T )−1x, y) = λ((1 − T )−1x, y + w)λ((1 − T )−1y, w)
For x = (1 − T )u and y = (1 − T )v this reads
λ(u + v, w)λ(u, (1 − T )v) = λ(u, (1 − T )v + w)λ(v, w) (2.11)
Dividing this by (2.10) we get
λ(u, (1 − T )v)λ(u, v)−1 = λ(u, (1 − T )v + w)λ(u, v + w)−1 (2.12)
Note that the left hand side of this formula does not contain w. Specifying to w = −v we find λ(u, (1 − T )v)λ(u, v)−1 = λ(u, −T v) (2.13) Plugging this back into (2.12) we get
λ(u, −T v) = λ(u, (1 − T )v + w)λ(u, v + w)−1 (2.14)
For a = v + w and b = −T v it reads
λ(u, b)λ(u, a) = λ(u, a + b) (2.15)
So λ is linear in the second coordinate. Using that λ(x, y) = λ(y, x)−1 it follows that λ is in fact bilinear. Since λ(x, y) = γ((1 − T )x, y) and (1 − T ) is invertible it follows that γ is also bilinear.
16 Now we know that γ is bilinear we can restate (2.6) in the following way
γ(x, y) = γ(T y, x + (1 − T )y) = γ(T y, x)γ(T y, (1 − T )y) = γ(T y, x) (2.16)
Using this we can view γ as a group homomorphism from ST (M) to Adj(M) where ST (M) is defined as
ST (M) = M ⊗ M/ < x ⊗ y − T y ⊗ x > (2.17)
So in ST (M) we have the relation [x ⊗ y] = [T y ⊗ x]. Now we know enough to give a nice description of Adj(M). Define GT (M) by GT (M) = Z × M × ST (M) (2.18)
We make GT (M) into a group by defining
(k, x, µ) (l, y, ν) = (k + l, T l(x) + y, µ + ν + [T l(x) ⊗ y]) (2.19)
It’s easy to verify that this indeed defines a group structure on GT (M). In order to show associativity one has to use the defining relation of ST (M). Note that (k, x, µ)−1 = (−k, −T −kx, −µ + [x ⊗ x]). ∼ Theorem 2.1.1. Adj(M) = GT (M).
Proof. We start by defining a homomorphism φ : Adj(M) → GT (M) by setting −1 −1 φ(ex) = (1, x, 0) and φ(ex ) = (−1, −T x, [x ⊗ x]). In order to see that this is well defined we have to verify the following
−1 −1 φ(ex ) φ(ey) φ(ex) = (−1, −T x, [x ⊗ x]) (1, y, 0) (1, x, 0) = (−1, −T −1x, [x ⊗ x]) (2, T y + x, [T y ⊗ x]) = (1, −T x + T y + x, [x ⊗ x] + [T y ⊗ x] −[T x ⊗ T y + x]) = (1, y x, 0) = φ(e ) B yBx
We’ll now define another homomorphism ψ : GT (M) → Adj(M) by setting k−1 −1 ψ(k, x, µ) = e0 exγ(µ) . We first need to check that this indeed defines a group homomorphism.
k−1 −1 l−1 −1 ψ(k, x, µ)ψ(l, y, ν) = e0 exγ(µ) e0 eyγ(ν) k+l −1 −1 −1 = e0 e0 eT lxe0 eyγ(µ + ν) k+l −1 l −1 −1 = e0 e0 eT lx+yγ(T x, y) γ(µ + ν) = ψ(k + l, T lx + y, µ + ν + [T lx ⊗ y])
Now we’ll see that φ and ψ are each others inverses
ψ(φ(ex)) = ψ(1, x, 0) = ex (2.20)
17 The other way around takes more work. First we have
k−1 −1 −1 φ(ψ(k, x, µ)) = φ(e0 ex)φ(γ(µ) ) = (k, x, 0) φ(γ(µ) ) (2.21) It’s now enough to show that φ(γ([x ⊗ y])−1) = (0, 0, [x ⊗ y]) to complete the proof.
−1 −1 −1 φ(γ([x ⊗ y]) ) = φ(ex+yexe0 ey) = (−1, −T −1(x + y), [(x + y) ⊗ (x + y)]) (1, x, 0) (−1, 0, 0) (1, y, 0) = (−1, −T −1(x + y), [(x + y) ⊗ (x + y)]) (1, x + y, [x ⊗ y]) = (0, 0, [x ⊗ y])
2.1.2 Alexander quandles with connected components We’ll now tackle another class of Alexander quandles. This time assume that M 6= C0 but that the subquandle C0 = (1 − T )M is connected and therefore (1 − T ): C0 → C0 is invertible. An example of such a quandle would be R6, the Alexander quandle where M = Z6 and T is multiplication by −1. C0 is a connected component of M, but what do the other components of M look like? First note that the map f : x 7→ x + a for some fixed a ∈ M is a quandle homomorphism. We see that f maps C0 to the connected component of a. Since f is a bijection it follows that the connected component of a is equal to a + C0. Since (1 − T ): C0 → C0 is a bijection it also follows that every connected component has to contain exactly one element of the kernel of (1 − T ). So every connected component is of the form a + C0 with a ∈ ker(1 − T ). In the rest of this section let a, b, c etc. be variables over ker(1 − T ) and let x, y, z etc. be variables over C0. We hope that we can find Adj(C0) as a subgroup of Adj(M) since we already have described Adj(C0) in the previous section. In fact we hope to find one copy of Adj(C0) for each connected component of M since they are all isomorphic. −1 −1 Note that e0 ea ∈ K since (1 − T )a = 0. Therefore e0 ea is central in Adj(M). So we get −1 −1 −1 −1 e0 eaex = e0 eag e0g = g eag = ea+x (2.22) −1 where g = e0 eu with u such that (1 − T )u = x. The first and last equalities follow from 1.2. This is already enough to tie everything together. Let A = {a|a ∈ ker(1 − T ) \{0}} and set
GT (M) := Z × C0 × ST C0 × Z[A] = GT (C0) × Z[A] (2.23)
The group structure is simply the direct product of GT (C0) and Z[A]. We’ll write 0 = 0 in Z[A] to simplify notation.
18 ∼ Theorem 2.1.2. Adj(M) = GT (M) ∼ Proof. Since GT (M) = Adj(C0) × Z[A] we can define the following map φ : Adj(M) → GT (M) by setting φ(ea+x) = (ex, a). One easily checks that this is a well-defined. The inverse ψ : GT (M) → Adj(M) is given by
X Y −1 ka ψ(g, kaa) = g · (e0 ea) (2.24) a∈A a∈A
−1 Because e0 ea is central it’s easy to see that this is also a well defined homo- morphism. We have
−1 ψφ(ea+x) = ψ(ex, a) = exe0 ea = ea+x
X Y −1 ka Y X φψ(g, kaa) = φ(g) φ(e0 ea) = (g, 0) (0, kaa) = (g, kaa) a∈A a∈A a∈A a∈A
We now find a subgroup of Adj(M) isomorphic to Adj(C0) for every con- nected component of M. These subgroups are given by
Ga = {(k, x, µ, ka) ∈ GT (C0) × Z[A]}
2.2 The second cohomology group of connected Alexander quandles
In this section let M be a connected alexander quandle. Before we get to the explicit constructions of cocycles we’ll first alter the isomorphism between Adj(M) and GT (M) a bit. 0 Definition 2.2.1. Define φ : Adj(M) → GT (M) by φ(ex) = (1, (1 − T )x, 0). To see that φ0 is an isomorphism note that φ0 = φ ◦ Adj(1 − T ) where φ is the isomorphism from the previous section. Since (1 − T ) is an isomorphism so is Adj(1 − T ). Lemma 2.2.1. For g ∈ Adj(M) we have
xg = T (g)x + 0g (2.25)
Proof. We obviously have x1 = x = T 0(x) + 0. Now suppose for some g ∈ Adj(M) relation (2.25) holds. It suffices to prove that it also holds for gey and −1 gey . A simple computation shows:
ge g (g) g x y = x B y = (T (x) + 0 ) B y = T (g)+1(x) + T (0g) + (1 − T )(y) = T (gey )(x) + 0gey
−1 A similar computation shows the same for gey .
19 Thanks to this lemma we see that for g ∈ Adj(M) we have φ0(g) = ((g), 0g, ν) for some ν ∈ ST (M). In light of this we’ll take 0 as a base point for our quandle and find g ∼ π1(M, 0) = {g ∈ Adj(M)|(g) = 0, 0 = 0} = ST (M) The universal covering of M, M˜ , is given by
M˜ = {(a, g) ∈ M × Adj(M)◦|0g = a}
As noted in [7] the first coordinate is redundant since all information is contained in Adj(M)◦. The quandle operation from 1.6.11 is then transformed into
−1 −1 g B h = e0 gh e0h
◦ We know that Adj(M) = {(k, x, µ) ∈ GT (M)|k = 0}. So this formula becomes
(0, x, µ) B (0, y, ν) = (−1, 0, 0) (0, x, µ) (0, −y, −ν + [y ⊗ y]) (1, 0, 0) (0, y, ν) = (−1, 0, 0) (0, x − y, µ − ν + [y ⊗ y] − [x ⊗ y]) (1, y, ν) = (−1, 0, 0) (1, T x − T y + y, µ − ν + [y ⊗ y] − [x ⊗ y] +ν + [T x ⊗ y] − [T y ⊗ y]) = (0, x B y, µ + [y ⊗ x] − [x ⊗ y]) = (0, x B y, µ + [(1 − T )y ⊗ x]) (2.26) The covering p : M˜ = Adj(M)◦ → M is simply given by projection on the second coordinate. The action of Adj(M) on M˜ from 1.6.2 is given by
(0, x, µ)g = (−ε(g), 0, 0) (0, x, µ) (ε(g), 0g, ν) = (0,T ε(g)(x) + 0g, µ + ν + [T ε(g)(x) ⊗ 0g]) where g ∈ Adj(M) and φ0(g) = ((g), 0g, ν). We’ll drop the first coordinate of the elements of M˜ = Adj(M)◦ from now on since it’s 0 anyway. Note that when 0 g ∈ π1(M, 0) we can identify g with ν ∈ ST (M) where φ (g) = (0, 0, ν). The action for such g then becomes the simple formula.
(x, µ)g = (x, µ + ν)
Now we have the necessary preparations out of the way. Let Λ be a group ∼ and ξ : π1(M, 0) = ST (M) → Λ. This ξ defines an right action of ST (M) on Λ given by λg = λξ(g). Like in lemma 6.5 of [7] we now construct
g M˜ α := Λ × M/˜ h(λ , x, µ) ∼ (λ, x, µ + ν)i (2.27)
0 where again g ∈ π1(M, 0) and φ (g) = (0, 0, ν). The quandle operator on M˜ α is given by 0 (λ, x, µ) B (λ , y, ν) = (λ, x B y, µ + [(1 − T )y ⊗ x])
20 M˜ α covers M by projection on the second coordinate. Λ acts on M˜ α by 0 0 ˜ λ(λ , x, µ) = (λλ , x, µ). It’s easy to verify that this turns Λ y Mα → M into an extension of M by Λ. This completes the first half of the isomorphism from Hom(π1(M, 0), Λ) to 2 HQ(M). Now we create a 2-cocycle f like in lemma 9.6 of [7]. Define a section ˜ s : M → Mα by s(x) := (1Λ, x, 0). f is defined by s(a) B s(b) = f(a, b)s(a B b). We find
s(a) B s(b) = (1, a, 0) (1, b, 0) = (1, a B b, [(1 − T )b ⊗ a]) = (ξ([(1 − T )b ⊗ a], a B b, 0) = ξ([(1 − T )b ⊗ a])s(a B b)
The third equality follows from (2.27), the defining relation of M˜ α. So f(a, b) = ξ([(1 − T )b ⊗ a]). Since 1 − T is a bijection and ξ : M ⊗ M/ ∼→ Λ it follows that f is bilinear. Also note
f(a, b) = ξ([(1 − T )b ⊗ a]) = ξ([T a ⊗ (1 − T )b]) = ξ([T a ⊗ b − T a ⊗ T b] = ξ([(T − 1)a ⊗ b]) = −f(b, a)
Finally since [a ⊗ b] = [T a ⊗ T b] we also have f(a, b) = f(T a, T b). One can verify that any map M 2 → Λ with these properties is in fact a 2-cocycle.
21 Chapter 3
Polynomial cocycles
22 3.1 Introduction
In [17, 18] Mochizuki computes the third cohomology group of a collection of Alexander quandles. His proof contains some errors that result in his list of 3-cocycles being incorrect. However his basic methodology is useful. In this chapter we correct the proof given in [18].
3.2 Polynomial cochains
In this chapter we will consider Alexander quandles where M is a finite field and the automorphism is given by multiplication. Let q be a power of a prime p and let 0, 1 6= ω ∈ Fq, consider the Alexander quandle X = Fq where the quandle operation is given by a B b = ωa + (1 − ω)b. We are going to use the field structure of X to rewrite the cochain complex with polynomials.
Definition 3.2.1. Let k be any field extension of Fq. Define n C (k, q) = {f ∈ k[U1,...,Un]|degUi (f) < q} The obvious map from Cn(k, q) to the collection of all maps Xn → k is a bijection. Definition 3.2.2. We’ll turn C∗(k, q) into a chain complex by defining δ(f) ∈ Cn+1(q) for f ∈ Cn(q) by
δ(f)(U1,...,Un+1) = n X i−1 (−1) f(ωU1, . . . , ωUi−1, ωUi + Ui+1,Ui+2,...Un+1) i=1 n X i−1 − (−1) f(U1,...,Ui−1,Ui + Ui+1,Ui+2,...,Un+1) i=1 A simple computation shows that δ2 = 0 so this does in fact define a chain ∗ complex. This chain complex is equivalent to CR(X, k). Define the chain- ∗ ∗ homomorphism φ : C (k, q) → CR(X, k) by
φ(f)(x1, . . . , xn) = f(x1 − x2, . . . , xn−1 − xn, xn) It’s easy to verify that φδ = δφ, so φ is indeed a chain-homomorphism. It’s also simple to see that φ is a bijection.
Lemma 3.2.1. Let k1, k2 be two fields such that Fq ⊂ k1 ⊂ k2. We have n ∗ n ∗ dim(H (C (k1, q)) = dim(H (C (k2, q))).
Proof. Find a set I and ai ∈ k2 for i ∈ I such that the ai form a basis of k2 n n as a k1-vectorspace. Let f ∈ C (k2, q). Then we can find gi ∈ C (k1, q) such P P n that f = aigi. We find that δf = aiδgi. So if f is a cocycle in C (k2, q) n then it is a linear combination of cocycles in C (k1, q), and analogously for coboundaries.
23 Convention 3.2.3. Because of the last lemma we can restrict ourselves to n n k = Fq and we’ll simply write C for C (Fq, q).
3.3 A decomposition and filtration of the com- plex
If f is a polynomial of homogenous total degree d then so is δ(f). So we can decompose the complex by total degree.
Definition 3.3.1. For d ∈ N define
n n Cd = {f ∈ C |f is homogenously of total degree d}
n L n It’s clear that C = Cd . d
Definition 3.3.2. We’ll define an order on the monomials of k[U1,U2,...] by
X ei X ti Ui ≤ Ui ⇔ (e1, e2,...) ≤lex (t1, t2,...) (3.1) i i
Where ≤lex is the lexicographical order. Lemma 3.3.1. If m is a monomial then all the monomials occurring in δ(m) are smaller than or equal to m. The coefficient of m in δ(m) is equal to ωd − 1.
Proof. This follows from a trivial computation.
d ∗ Theorem 3.3.2. If ω 6= 1 then the complex Cd is acyclic. n Proof. Let f ∈ Cd with δf = 0. We’ll prove that f is a coboundary with e1 en induction to the largest monomial occurring in f. Let m = U1 ...Un be the largest monomial occurring in f and let A be its coefficient. Finally let k = max{i|ei 6= 0}. Suppose n − k is even. The coefficient of m in δf is equal to A(ωd − 1) 6= 0. n−1 d −1 So n − k is odd. Therefore en = 0 and m ∈ Cd . Now f − A · (ω − 1) δm only contains monomials strictly smaller than m. Proceeding with induction concludes the proof.
Convention 3.3.3. We’ll write Tn for Un when dealing with polynomials in n Cd to emphasize what the dimension of a polynomial is.
Definition 3.3.4. Let s ∈ N and define ( ) n(s) X aps n Cd = fa(U1,...,Un−1) · Tn ∈ Cd a n(∞) n \ n(s) Cd = {f0(U1,...,Un−1) ∈ Cd } = Cd s
24 n(s+1) n(s) s n(s) n(∞) It’s clear that Cd ⊂ Cd and when p > q then Cd = Cd because the degree of Tn has to be smaller than or equal to q. It’s also easy to check n(s) n+1(s) ∗ that δ(Cd ) ⊂ Cd . So we have a filtration of the complex Cd . Definition 3.3.5. Set
n n Zd = ker(δ) ∩ Cd n(s) n(s) Zd = ker(δ) ∩ Cd n n−1 Bd = δ(Cd ) n(s) n−1(s) Bd = δ(Cd )
(s) n(s) n(s) Definition 3.3.6. There is a chain homomorphism D∗ : Cd → Cd−ps given by ! (s) X aps X (a−1)ps Dn fa(U1,...,Un−1)Tn = a · fa(U1,...,Un−1)Tn a a
(s) (s) It can be easily checked that δ ◦ Dn = Dn+1 ◦ δ.
3.4 The 3-cocycles
In [18] Mochizuki gives a list of cocycles which he claims form a basis of H3(C∗). However it turns out that some items on this list are if fact not cocycles at all and others are coboundaries. This list is corrected below. First define the following polynomial in Fp[x, y]
p−1 X 1 χ(x, y) = (−1)i−1i−1xp−iyi ≡ ((x + y)p − xp − yp) (mod p) p i=1
Definition 3.4.1. Now for a, b ∈ N define