RIMS-1707
Some topological aspects of
4-fold symmetric quandle invariants of 3-manifolds
By
Eri HATAKENAKA and Takefumi NOSAKA
October 2010
R ESEARCH I NSTITUTE FOR M ATHEMATICAL S CIENCES
KYOTO UNIVERSITY, Kyoto, Japan
Some topological aspects of
4-fold symmetric quandle invariants of 3-manifolds
Eri Hatakenaka and Takefumi Nosaka
Abstract
The paper relates the 4-fold symmetric quandle homotopy (cocycle) invariants with topological objects. We show that the 4-fold symmetric quandle homotopy invariants are at least as powerful as the Dijkgraaf-Witten invariants. As an application, for an odd primep, we show that the quandle cocycle invariant of a link in S3 using the Mochizuki 3-cocycle is equivalent to the Dijkgraaf-Witten invariant with respect toZ/pZof the double covering ofS3 branched along the link. We also reconstruct the Chern-Simons invariant of closed 3-manifolds as a quandle cocycle invariant via the extended Bloch group, in analogy to [IK].
Keywords Quandle, symmetric quandle, quandle cocycle invariant, link, 3-manifold, branched covering, Dijkgraaf-Witten invariant, bordism group, Chern-Simons invariant, the extended Bloch group.
Contents
1 Introduction 2
2 Review: 4-fold symmetric quandle homotopy invariant 4 3 4-fold symmetric quandle homotopy invariants as natural transforma-
tions 6
3.1 Preliminaries: the symmetric link quandle and the associated group . . . . 7 3.2 Gec-colorings of labeled diagrams . . . 8 3.3 A fundamental symmetric quandle of a 3-manifold . . . 11 3.4 4-fold symmetric quandle homotopy invariants as natural transformations . 12 4 Formulas for the connected sum and the opposite orientation 13
5 Dijkgraaf-Witten invariant 16
5.1 Preliminary: bordism Dijkgraaf-Witten invariant . . . 16 5.2 From Π4f2,ρ(Gec) to the oriented bordism group Ω3(G, c) . . . 18 5.3 Applications . . . 21 6 Chern-Simons invariant as a quandle cocycle invariant 23 6.1 Review: 4-fold symmetric quandle cocycle from normalized group cocycle 23 6.2 Review: Cheeger-Chern-Simons class from the extended Bloch group. . . . 25 6.3 Chern-Simons invariant as a quandle cocycle invariant . . . 26 6.4 Recovery of Chern-Simons invariant from the multiplication by 6. . . 27
1 Introduction
Our motivation stems from a topic of Dijkgraaf-Witten invariant described in (1) be- low. Let M be an oriented compact 3-manifold. Dijkgraaf and Witten [DW] discussed the relation between Chern-Simons action and Wess-Zumino-Witten term on M through the cohomology of the Eilenberg-MacLane space H3(K(G,1);Z), where G is a compact Lie group. If M has no boundary, then the quantum field theory of the Chern-Simons functional in H3(K(G,1);C/Z) can be interpreted as an obstruction class in the oriented bordism group Ω3(K(G,1)). Furthermore, one of their noteworthy results is that, when G is finite, the path integral of the distribution function of a 3-cocycle ψ ∈H3(K(G,1);A) reduces to a finite sum
DWψ(M) = ∑
f∈Homgrp(π1(M),G)
⟨f∗(ψ),[M]⟩ ∈Z[A], (1)
where [M] ∈ H3(M;A) is the fundamental class of M. From a mathematical viewpoint, Wakui [W] rigorously formulated DWψ(M) of certain 3-cocycles ψ ∈ H3(G;A) in terms of a triangulation of M. Recently, the first author [H1] reformulated the Wakui formula as a “quandle cocycle invariant” of links, using the fact that any closed 3-manifold is a 4-fold branched covering of S3.
Inspired by her work, the second author [N2] introduced a 4-fold symmetric quandle homotopy invariant of 3-manifolds using a quandleGec. Here,Gis a finite group andc∈G is its central element satisfying c2 = e, and Gec is a certain quandle defined with respect to the pair of (G, c). The invariant of M is defined to be a set of “Gec-colorings” with a grading by an abelian group Π4f2,ρ(Gec). Here, ifM is a 4-fold branched covering of S3 along a linkL, aGec-coloring is roughly a homomorphismπ1(S3\L)→G4oS4 compatible with the monodromy π1(S3\L)→S4. On the other hand, the group Π4f2,ρ(Gec) is defined to be a certain link bordism group ofGec-colorings. The previous paper [N2] studied the quandle structure of Gec and estimated Π4f2,ρ(Gec) in an algebraic viewpoint. It is shown [N2, §7]
that the invariant produces the above quandle cocycle invariants considered in [H1].
In this paper, we topologically study the 4-fold symmetric quandle homotopy invariant.
For this, we give a topological interpretation of the Gec-colorings, and relate the group Π4f2,ρ(Gec) with some topological objects. We first show a natural bijection between G3 × Homgrp(π1(M), G) and the set of Gec-colorings (Theorem 3.3). Note that the set of Gec- colorings is a classical invariant and is independent of the central element c∈G.
We next give our invariants some functoriality (§3.4). We introduce a fundamental symmetric quandle SQ(M) of M (Definition 3.5). SQ(M) is roughly defined to be a universal quandle representing the allGec-colorings. Using the bijection of Theorem 3.3 and the universality, we show that the quandle SQ(M) is quandle isomorphic to G(M^)c(M), where G(M) := π1(M)× Z/2Z and c(M) := (e,1) ∈ G(M) (Corollary 3.6). Further, we define the fundamental class to be a canonical class of a link bordism using SQ(M)
(Definition 3.7). Consequently, the study of the 4-fold homotopy invariant of M is a research of the fundamental class using relativity to other 4-fold symmetric quandles Gec
(see §3.4 for detail).
Next, in order to study the group Π4f2,ρ(Gec) mentioned above, we compare the4-fold symmetric quandle homotopy invariants with the Dijkgraaf-Witten invariants given in (1). For this, we take a perspective of a bordism group rather than the state sum formula.
We then work with the bordism Dijkgraaf-Witten invariant defined by using an oriented bordism group Ω3(G, c) of the pair of (G, c) (see §5.1). From the viewpoint regarding Π4f2,ρ(Gec) as a certain link bordism, we canonically obtain an epimorphism Π4f2,ρ(Gec) → Ω3(G, c), and show that the bordism invariant is derived from the 4-fold symmetric quandle homotopy invariant (Theorem 5.4). We remark that, whenc=e, the bordism Dijkgraaf- Witten invariant produces DWψ(M) in (1) for any 3-cocycle ψ ∈ H3(K(G,1);A) (see Remark 5.1); hence, so does the 4-fold symmetric quandle homotopy invariant. For the future, it is a problem whether the epimorphism is isomorphic or not (Problem 5.7). If this is isomorphic, the two invariants in Theorem 5.4 are equivalent.
As an application of bordism groups, we succeed in giving a topological interpretation of two combinatorial invariants of links. For an odd m ∈ Z, the dihedral quandle Rm is defined to beZ/mZ with a binary operationx∗y= 2y−x. Then the quandle homotopy invariant of oriented linksL⊂S3with respect toRm is defined by a combinatorial method (see, e.g., [FRS, N1]). Let ML be the double branched cover of the link L. Then we show that the quandle homotopy invariant ofL is equivalent to the Dijkgraaf-Witten invariant of ML with respect to G = Z/mZ (Corollary 5.9). As a special case, if m is a prime p, then it is known [N1] that “the quandle cocycle invariant Φθp(L) ∈ Z[t]/(tp −1) using Mochizuki 3-cocycle” is equivalent to the quandle homotopy invariant of links, leading to an equality Φθp(L) =atnDWψ(ML) for somea, n∈Z(Corollary 5.11). As a corollary, we compute the Dijkgraaf-Witten invariants of some 3-manifolds (Example 5.12, 5.13, 5.14), using known values of the quandle cocycle invariants Φθp(L) in [AS, Iwa1, Iwa2].
In another direction, whenG=SL(2;C) orP SL(2;C), we discuss the Cheeger-Chern- Simons class Cb2 ∈ H3(G;C/4π2Z). Given a homomorphism f : π1(M)→ G, the Chern- Simons invariant of M is defined to be the pairing ⟨f∗(Cb2),[M]⟩ ∈C/4π2Z. It had been a long-standing problem to provide a computation of the Cheeger-Chern-Simons class and the Chern-Simons invariant. Dupont [Dup] gave an answer modulo π2Q. Lifting his formula, Neumann [Neu] has obtained an explicit formula forCb2 with G=P SL(2;C) via the extended Bloch groupBb(C), and a computation of the Chern-Simons invariant in term of a triangulation of M. Further, Dupont, Goette and Zickert succeeded in an extension of the formula suitable for G=SL(2;C) [DG, DZ]. From quandle theory, Inoue-Kabaya [IK] in ’09 reconstructed the Chern-Simons invariant of knot complements S3 \K as a quandle cocycle invariant, using Bb(C). In this paper, as an analogy, for G = SL(2;C), we reconstruct the Chern-Simons invariant of closed 3-manifolds as a quandle cocycle
invariant throughBb(C), using the Dupont formula [Dup] and a result in [H1] (Theorem 6.5 and §6.4). Similar to Inoue-Kabaya’s result [IK], a benefit of our reformulation is that we can combinatorially compute the Chern-Simons invariant only from the homomorphism f : π1(M) → SL(2;C) and the monodromy ϕ : π1(S3 \L) → S4 of a 4-fold branched covering M →S3, without using triangulation of M.
Lastly, we outline the reconstruction. The point follows from the Dupont formula [Dup] rather than Bb(C). He reformulated 6Cb2 as a function on a configuration space Conf4(SL(2;R)). The reformulation is adequate for a result of the first author [H1];
hence, we succeed in reconstructing 6 multiple of the Chern-Simons invariant (Theorem 6.5). Finally, the Chern-Simons invariant makes a recovery from the multiplication by 6, using the Dijkgraaf-Witten invariants with respect to cyclic groups (§6.4).
This paper is organized as follows. In §2, we review some notation of 4-fold sym- metric quandle homotopy invariants. In §3, we give a topological interpretation of the Gec-colorings. In §4, we show the connected sum formula. In §5, we compare the 4-fold symmetric quandle homotopy invariant with the bordism Dijkgraaf-Witten invariant. In
§6, we reformulate the Chern-Simons invariant as a quandle cocycle invariant.
2 Review: 4-fold symmetric quandle homotopy invariant
We briefly review some notation of 4-fold symmetric quandle homotopy invariants in [N2,
§2 and 3]. Throughout this paper, manifolds are assumed to be C∞-smooth, oriented, connected and compact. Unless§5, we assume that manifolds have no boundary.
We first review symmetric quandles introduced by Kamada [Kam]. Asymmetric quan- dle is a triple of a set X, a binary operation ∗ on X and an involution ρ : X → X satisfying that, for any x, y, z ∈ X, x∗x =x, (x∗y)∗z = (x∗z)∗(y∗z), ρ(x∗y) = ρ(x)∗y, (x∗y)∗ρ(y) =x. For example,S :={(ij)∈S4} withx∗y:=yxy andρ(x) = x is a symmetric quandle. We give another example introduced in [N2, Example 4.1] as follows. We consider a pair of a group G and its central element c∈G such that c2 = e, wheree∈Gis the identity element. Such a pair of (G, c) is called a cored group. Putting T12 :={(i, j)∈Z2|1≤i, j ≤4, i ̸=j}, we defineGecto be a quotient setG×T12/∼, where the equivalence relation ∼ on G×T12 is defined by (g, i, j) ∼ (g−1c, j, i), for (i, j) ∈ T12
and g ∈G. We equip Gec with an operation ∗:Gec×Gec→Gec defined by (g, i, j)∗(g′, i, j) = (g′g−1g′, i, j), (g, i, j)∗(g′, j, k) = (gg′, i, k), (g, i, j)∗(g′, k, l) = (g, i, j),
where i, j, k, l are distinct indices. Further, define ρ : Gec → Gec by ρ(g, i, j) = (gc, i, j).
Then (Gec, ρ) is a symmetric quandle. Moreover, putting a projection pGe
c : Gec → S which sends (g, i, j) to (ij)∈ S, the triple of (Gec, ρ, pGe
c) satisfies the axioms of the 4-fold symmetric quandle (see [N2, Definition 3.1] for detail). Remark that if G = {e}, then Gee∼=S.
For simplicity, in this paper we denote three elements (e,1,2), (e,2,3), (e,3,4) ∈ Gec bye12, e23, e34, respectively.
We review Xρ-colorings. Let Dbe an unoriented link diagram on R2. For a symmetric quandle (X, ρ), anXρ-coloringofDis a mapC :{the two normal orientations on arcs of D}
→X satisfying the following two conditions:
(X1) For the two orientations α1, α2 of the same arc as shown in Figure 1, the colors satisfy C(α1) =ρ(
C(α2))
. (Hence, we will later draw the only one color of the two).
(X2) At each crossing shown in Figure 1, the three orientations satisfyC(γ) = C(α)∗C(β).
α
1α
2C(α1) =ρ( C(α2))
α β
γ
C(γ) =C(α)∗C(β)
Figure 1: The conditions of symmetric colorings on orientations.
Note that the conditions (X1)(X2) are well-defined by the axioms of symmetric quandles.
Denote by ColX,ρ(D) the set of all Xρ-colorings of D. It is known [KO, Proposition 6.2]
that, if two diagramsD1andD2are related by Reidemeister moves, there exists a bijection between ColX,ρ(D1) and ColX,ρ(D2).
For a symmetric quandle (X, ρ), an (X, ρ)-set is a set Λ equipped with a map ∗ : Λ×X −→ Λ satisfying (λ∗x)∗x′ = (λ∗x′)∗(x∗x′) and (λ∗x)∗ρ(x) = λ for any λ∈Λ and x, x′ ∈X. For example,X is an (X, ρ)-setitself by the quandle operation. An XΛ-coloring of D is defined to be an Xρ-coloring of D with an assignment of elements of Λ to each complementary regions of D such that, for each regions separated by the arc with a color x∈X, the colors and assignments satisfy the following picture.
x [λ]
[λ′]
λ∗x=λ′. (λ, λ′ ∈Λ)
We will interpret 3-manifolds as Sid-colorings, where S = {(ij) ∈ S4} as above. It is well-known that any 3-manifold M is a 4-fold simple covering of S3 branched over a link L with its monodromy ϕ :π1(S3\L) →S4. Remark that ϕ is surjective and sends each meridian of L to a transposition inS4 (see, e.g., [N2,§2.2]). A link diagram D of Lwith such a monodromyϕ is called alabeled diagramand denoted byDϕ. Notice that a labeled diagram can be regarded as an Sid-coloring of D by Wirtinger presentation (see [PS, §24]
for detail). A labeled diagram Dϕ is said to be 3-fold, if its subdiagram of Dϕ labeled by (34) is a single unknot as shown in Figure 2. It is known that M can be regarded as a 3-fold labeled diagram Dϕ (see, e.g., [R, §10.D]). We fix three orientations α12, α23, α34 of three distinct arcs inDϕlabeled by (12),(23),(34)∈S4as shown in Figure 2, respectively.
We denote by Colee12,e23,e34
Gc,ρ (Dϕ) the subset of Gec-colorings C of D such that pGe
c(C) =Dϕ and these orientations αij are colored by eij = (e, i, j)∈Gec, for (ij)∈ {(12),(23),(34)}.
D ϕ (34)
(23) (12)
α34 α23
α12
Figure 2: A 3-fold labeled diagram with three orientationsα12, α23, α34.
For a cored group (G, c), we review an abelian group Π4f2,ρ(Gec) defined in [N2, §3.1].
Π4f2,ρ(Gec) is defined by a quotient set of all Gec-colorings of all link diagrams D modulo Reidemeister moves, symmetric concordance relations and MI, II moves. Here,symmetric concordance relationsandMI, II movesare local moves as shown in Figure 3 and 4, respec- tively. Then we impose an abelian structure on Π4f2,ρ(Gec) by letting disjoint union define our multiplication. For a labeled diagram Dϕ, we put a map Ξ4fe
Gc(Dϕ;•) : Colee12,e23,e34
Gc,ρ (Dϕ)→ Π4f2,ρ(Gec) which sends a Gec-coloring of Dϕ to the canonical class (see [N2, (3)]).
Let M be a 3-manifold presented by a 3-fold labeled diagram Dϕ. When G is finite, the second author defined a 4-fold symmetric quandle homotopy invariant of M by the following formula (see [N2, Definition 3.3 and Lemma 4.6]).
Ξ4fe
Gc(M) = |G|3 ∑
C∈Colee12,e23,e34
Gc,ρ (Dϕ)
Ξ4fe
Gc(Dϕ;C)∈Z[Π4f2,ρ(Gec)]. (2) The definition does not depend on the choice of labeled diagrams and the three arcs (see [N2, Theorem 3.4 and Lemma 4.6]).
a ρ(a)
a ρ(a)
a ρ(a) ρ(a)
a
a ρ(a)
∅
Figure 3: symmetric concordance relations. Herea∈Gec.
3 4-fold symmetric quandle homotopy invariants as natural trans- formations
In§3.2, we give a topological meaning ofGec-colorings. In§3.3, we introduce a fundamental symmetric quandle SQ(M) of a 3-manifold M, and give an interpretation of Gec-colorings
(g, i, j) (h, j, k) (g, i, j) (h, j, k)
(g, i, j) (h, j, k) (gh, i, k)
(f, k, l) (g, i, j)
(f, k, l) (g, i, j)
(g, i, j) (f, k, l)
Figure 4: MI, II moves of Gec-colorings. Here, (g, i, j), (h, j, k), (f, k, l)∈Gec, andi, j, k, lare distinct.
as a representable functor usingSQ(M). Further, we interpret a 4-fold symmetric quandle homotopy invariant as a natural transformation.
3.1 Preliminaries: the symmetric link quandle and the associated group Let L be an unoriented link in S3. We briefly review the symmetric link quandle of L introduced by Kamada [Kam]. Let SQ(L) be the set of homotopy classes of all pairs of (D, γ), where D means an oriented meridian disk of L and γ means a path in S3 \ L starting from a point of the boundary ∂D and ending at a fixed base point in S3\L. We equip SQ(L) with a binary operation given by
[(D1, γ1)]∗[(D2, γ2)] := [(D1, γ1·γ2−1·∂D2·γ2)],
and an involution ρ of SQ(L) given by ρ([(D, γ)]) = [(−D, γ)], where −D stands for the disk D with the opposite orientation. Then SQ(L) turns out to be a symmetric quandle (see also [KO, Example 2.4]).
We will explain the correspondence (3) below. Let (X, ρ) be a symmetric quandle and Dan unoriented link diagram of L. Let us denote by HomsQnd(SQ(L), X) the set of maps SQ(L) → X preserving the operations ∗ and ρ, which are called (symmetric quandle) homomorphisms. Kamada [Kam] gave a canonical bijection
Q(•) : ColX,ρ(D) −→ HomsQnd(SQ(L), X), (3)
where, for anXρ-coloringC,Q(C) is defined to be a homomorphism sending the meridian disk of an arc α to the color of C onα. This bijection is analogous to [Joy, §16].
We study labeled diagrams from a view ofSQ(L). Recall that any labeled diagramDϕ
can be regarded as anSid-coloring. By substituting the bijection (3) toX =S, we see that the map ϕ ∈ HomsQnd(SQ(L),S) associated to Dϕ through (3) is surjective. We denote three meridian disks obtained from the previous arcs α12, α23, α34, by D12, D23, D34 ∈ SQ(L), respectively. Then, as a restriction of the bijection (3), we obtain a bijection
Colee12,e23,e34
Gc,ρ (Dϕ)≃Hom(4sQndD12,D23,D34)(e12,e23,e34)(SQ(L),Gec), (4) where Hom(4sQndD12,D23,D34)(e12,e23,e34)(SQ(L),Gec) is defined to be the set of symmetric homo- morphisms f : SQ(L) → Gec satisfying f(Dij) = eij = (e, i, j) ∈ Gec and pGe
c ◦f = ϕ ∈ HomsQnd(SQ(L),S).
We review the associated group [KO] of a symmetric quandle (X, ρ) defined by the following presentation:
As(X)ρ =⟨x∈X | y·(x∗y) =x·y, ρ(x) = x−1 (x, y ∈X)⟩.
Notice that a symmetric quandle homomorphism f : (X, ρ) → (X′, ρ′) induces a group homomorphism As(f) : As(X)ρ→As(X′)ρ′.
Lastly, we discuss a canonical map iX :X →As(X)ρ defined by iX(x) =x. Note that ifX =Gec, theniGe
c is injective (cf. [N2, Lemma 3.8]). Further, we consider the case where X is the link quandle SQ(L). Put a map iL : SQ(L) → π1(S3\L) given by that, for a meridian disk D, iL(D)∈ π1(S3 \L) corresponds with a loop of the boundary ∂D. Then iL passes to a group homomorphism As(SQ(L))ρ → π1(S3 \L). It can be verified that this is an isomorphism by Wirtinger presentation (cf. [Joy, §14 and 15]). Moreover, we can check that iL is entirely the above mapiX :X →As(X)ρ with X =SQ(L), and that the image Im(iL) consists of the conjugacy classes of the meridians of L.
3.2 Gec-colorings of labeled diagrams Our objective is to show Theorem 3.3.
Given a labeled diagram Dϕ which presents a 3-manifold M, we first introduce the associated cored group denoted by (Gϕ, cϕ) as follows. Recall the associated monodromy representationϕ:π1(S3\L)−→S4in§3.1. Let (S4)1denote a subgroup{σ ∈S4 |σ(1) = 1} (∼= S3). Let S^3\L be the 4-fold (unbranched) covering associated to ϕ. Then an isomorphismπ1(S^3\L)∼=ϕ−1((S4)1) is known (see, e.g., [H1,§3.1]) and the boundary of S^3\L consists of 3♯L-tori. Let D be a link diagram of L, and l the number of the arcs of D. For 1 ≤ t ≤ l, we let met,j ∈ π1(S^3\L) be the meridian associated obtained from each lifted arcs, wherej ∈ {1,2,3}. We may assume that, for 1≤t ≤l, met,1 ∈π1(S^3\L) is obtained from the branched locus of index 2. Then, we define a normal subgroup of π1(S^3\L) by
Nϕ =⟨[met,1, π1(S^3\L)], met,1(met′,1)−1, (met,1)2, met,2, met,3 (1≤t, t′ ≤l)⟩, (5) where [met,1, π1(S^3\L)] are their commutator subgroups and the symbol ‘⟨ ⟩’ temporarily stands for the normal closure in π1(S^3\L). Define a group Gϕ =π1(S^3\L)/Nϕ. Notice π1(M) ∼= π1(S^3\L)/⟨met,j (1 ≤ t ≤ l, 1 ≤ j ≤ 3)⟩ by Van Kampen theorem, leading a canonical epimorphism Gϕ −→ π1(M). We can verify the epimorphism is a central extension, and the kernel is either Z/2Z or 0. Let us denote by cϕ a generator of the kernel. We define a cored group (Gϕ, cϕ) as required.
Proposition 3.1. Let (G, c) be a cored group, andDϕ a labeled diagram which presents a 3-manifold M. Then, there is a canonical bijection
Colee12,e23,e34
Gc,ρ (Dϕ)≃HomGrpc(Gϕ, G), (6)
where HomGrpc(Gϕ, G) is defined to be the set of group homomorphisms f : Gϕ → G satisfying f(cϕ) = c∈G.
Proof. In the case ofc=e, the first author proved this theorem (see [H1, Proposition 3.5]).
In general ofc, our proof is analogous to her proof. Then, we sketch a proof of Proposition 3.1. We often regardC ∈Colee12,e23,e34
Gc,ρ (Dϕ) as a homomorphism ψ :SQ(L)→Gec by (4).
First, given a symmetric quandle homomorphism ψ : SQ(L) → Gec, we construct Ψ ∈ HomGrpc(Gϕ, G) as follows. We consider an injection χ : Gec → G4 oS4 defined by χ(g, i, j) = (a1, a2, a3, a4,(ij)), where g = ai = ca−j1 and ak = e (k ̸= i, j). Here S4 acts on G4 by the transformations of components of G4. The map χ induces a group homomorphism ¯χ : As(Gec) → G4 oS4. Further, by the previous subsection, we have a commutative diagram:
SQ(L)
iL
ψ // Ge _c
iGce
v
χ
((R
RR RR RR RR RR RR RR R
π1(S3\L)
As(ψ) //As((Gec)ρ) χ¯ //G4 oS4
Denote by ¯Ψ a composite group homomorphism ¯χ◦ As(ψ). Hence, for each meridian m∈π1(S3\L), if the associated arc of Dϕ is colored by (g, i, j)∈Gec, then
Ψ(m) = (a¯ 1, a2, a3, a4,(ij)), where g =ai =ca−j1 and ak =e (k ̸=i, j). (7) Let us recall that π1(S^3\L) ∼= ϕ−1((S4)1) is a subgroup of π1(S3 \L). Denote by Ψe the restriction of ¯Ψ on π1(S^3\L). Then, the image of Ψ is contained in the subgroupe G4o(S4)1 ⊂G4oS4. Letπ1G:G4o(S4)1 −→Gbe the projection on the first component in G4. Then by (7) we can check that the composite homomorphism πG1 ◦Ψ sends eache meridians met,1 ∈ π1(S^3\L) to c ∈ G. Therefore πG1 ◦Ψ induces a homomorphism Ψ :e (Gϕ, cϕ)−→(G, c) as desired.
Conversely, given Ψ∈HomGrpc(Gϕ, G), we will construct aGec-coloring. Put the canon- ical sections1G:G−→(S4)1nG4 ofπG1. We defineΨ to be a compositee s1G◦Ψ◦πGϕ, where πGϕ is the canonical projectionπ1(S^3\L)−→ Gϕ. According to [H1], we have known the group presentations ofπ1(S3\L) andπ1(S^3\L), although we do not write them (see [H1,
§3.1] for more details). Then we can verify that π1(S3 \L) is generated by elements of π1(S^3\L) and three elements iL(D12), iL(D23), iL(D34)∈π1(S3\L), that is,
π1(S3\L) = ⟨π1(S^3\L), iL(D12), iL(D23), iL(D34)⟩.
Moreover, we can verify that Ψ uniquely extends to a homomorphism ¯e Ψ :π1(S3\L)−→
G4oS4 given by ¯Ψ(iL(Dij)) = χ(eij)∈G4oS4 (see [H1, Page 278]). Then ¯Ψ satisfies the condition (7). Hence, Im( ¯Ψ) is contained in Im(χ). We thus have a mapψ :SQ(L)−→Gec given by (χ)−1 ◦Ψ◦iL (see the diagram below). Further, since χ(Gec) ⊂ G4 oS4 is the
conjugacy class of (c, e, e, e,(12)), we can check that the map ψ is a symmetric quandle homomorphism SQ(L)→Gec. By the bijection (4), we obtain the required Gec-coloring of Dϕ whose arcs αij are colored by (e, i, j)∈Gec.
The two constructions give the required 1 : 1 correspondence. For an understanding of the proof, we put the following commutative diagram:
Gϕ Ψ
π1(S^3\L)
πGϕ
oo
Ψe
//π1(S3\L)
Ψ¯
SQ(L)
ψ
iL
oo
G π G4o(S4)1
1
oo G //G4oS4oo χ ? _Gec
We next discuss the projection Gϕ →π1(M) mentioned above.
Lemma 3.2. Let M be a 3-manifold. There exists a 3-fold labeled diagram Dϕ which presents M such that the projection has a splitting: Gϕ∼=π1(M)×Z/2Z.
Proof. It is shown [HMT] that there exists a 3-fold irregular branched coveringp:M →S3 such that the set of points at which pfails to be a local homeomorphism bounds a disk in M. We put the associated 3-fold labeled diagramDϕ.
Let us construct a homomorphism f : Gϕ → Z/2Z as follows. Let ϕ : π1(S3 \K) → S3 be the monodromy, where K is a knot in S3. Let ˜l (resp. m)˜ ∈ π1(S^3\K) be the longitude (resp. meridian) of a torus in S^3\K of local index 2. Put a map ιM : H1(S^3\K;Z) → H1(M;Z) induced by the inclusion S^3\K ,→ M. Since ˜l bounds a disk in M, we have ιM(˜l) = 0∈ H1(M;Z). Therefore, by a Mayer-Vietoris argument, we conclude that H1(S^3\K;Z) ∼=H1(M;Z)⊕Z, where the direct summand Z is generated by ˜m ∈H1(S^3\K;Z). Putting the projection H1(S^3\K;Z) →Z, we define a composite homomorphism by
π1(S^3\K)−−−→H1(S^3\K;Z)−−−−→proj. Z−−−−→proj. Z/2Z,
where the first map is the abelinization. Therefore, from the definition ofGϕ, the composite induces a required homomorphism f :Gϕ→Z/2Z satisfying f(cϕ) = 1.
On the other hand, recall that the projection Gϕ →π1(M) is a central extension, and that the kernel is either 0 or Z/2Z. Since f gives a crossed section of Gϕ → π1(M), we easily see Gϕ ∼=π1(M)×Z/2Zas desired.
In conclusion, we give a topological interpretation of Gec-colorings:
Theorem 3.3. For a cored group (G, c) and a labeled diagram Dϕ which presents a 3- manifold M, we thus have a bijection
ColGe
c,ρ(Dϕ)≃G3 ×Homgrp(π1(M), G), (8)
where ColGe
c,ρ(Dϕ) is the set of Gec-colorings C satisfying pGe
c(C) = Dϕ ∈ColS,id(D).
Proof. Since the set ColGe
c,ρ(Dϕ) depends on only M (see [N2, Proposition 3.2]), we may choose a 3-fold labeled diagram Dϕ in Lemma 3.2. Since Gϕ =π1(M)×Z/2Z, we notice a bijection HomGrpC(Gϕ, G) ≃ Homgrp(π1(M), G). Further, a bijection ColGe
c,ρ(Dϕ) ≃ G3×Colee12,e23,e34
Gc,ρ (Dϕ) is shown [N2, Lemma 4.6]. Hence, the required bijection is obtained from Proposition 3.1.
As a result, for a finite cored group (G, c), the cardinally of Gec-colorings is a classical invariant, and does not depend on the central element c∈G. Hence, our next step in §4 is to study the group Π4f2,ρ(Gec).
Incidentally, as a corollary, we give a topological interpretation of colorings of core quandles. Given a group G, we equip G with a symmetric quandle operation of g∗h = hg−1h and ρ= idG, called a core quandle.
Corollary 3.4. LetDbe a link diagram of a linkL, and Ga group. Denote by QGthe core quandle on G. LetML be the double branched covering of S3 branched over the link. Then the set of QG-colorings ColQG,id(D) is in 1:1 correspondence with G×Homgrp(π1(ML), G).
Proof. By Figure 5, we obtain a labeled diagram Dϕ from D, where we equip all arcs of D with labels (12) ∈ S and add two unknots labeled by (23) and (34). Then Dϕ presents ML. Note that the core quandle QG is isomorphic to the subquandle composed of {(g,1,2)∈Gee} by definitions. Hence, a QG-coloring of D is regarded as a Gee-coloring of the labeled diagram Dϕ, i.e., a homomorphism π1(ML)→G by Theorem 3.3.
D D
(23) (34)
Figure 5: A labeled diagramDϕ obtained from a link diagramD.
3.3 A fundamental symmetric quandle of a 3-manifold
For a 3-manifoldM, we will define a fundamental symmetric quandle ofM and investigate its property.
Let Dϕ be a labeled diagram which presents M. Recall the associated symmetric quandle epimorphism ϕ : SQ(L) → S in §3.1. We consider the following equivalent relations on SQ(L):
RL3,ϕ :=⟨xij ∗yjk =ρ(yjk)∗xij (xij ∈ϕ−1(ij), yjk ∈ϕ−1(jk))⟩ R4,ϕL :=⟨zij ∗wkl=zij (zij ∈ϕ−1(ij), wkl∈ϕ−1(kl))⟩
Then, we define the quotient symmetric quandle SQ(L)/⟨R3,ϕL , R4,ϕL ⟩. It goes without saying that the quotient quandle satisfies the axioms of the 4-fold symmetric quandle by definition (see [N2, Definition 4.1]). By a discussion similar to [N2, Proposition 3.2], if two labeled diagrams Dϕ and D′ϕ′ are related by some finite sequences of Reidemeis- ter moves and MI, MII moves with G = {e}, then we can obtain a symmetric quandle isomorphism SQ(L)/⟨R3,ϕL , RL4,ϕ⟩ ∼= SQ(L′)/⟨R3,ϕL′ ′, R4,ϕL′′⟩. Thus, by the result in [Apo], SQ(L)/⟨RL3,ϕ, R4,ϕL ⟩ does depend on only the 3-manifold M (see also [N2, Theorem 2.1]).
Definition 3.5. For a labeled diagram Dϕ of a 3-manifold M, we define a fundamental symmetric quandle of M by the quandleSQ(L)/⟨R3,ϕL , R4,ϕL ⟩. We denote it bySQ(M).
Assume that Dϕ is 3-fold. We use notation D12, D23, D34 ∈ SQ(L) in §3.1. Re- call the category of 4-fold symmetric quandles denoted by Qnd4s (see [N2, Corollary 4.3]). The objects of Qnd4s consist of Gec with respect to cored groups (G, c). Let us denote by HomQnd4s(SQ(M),Gec) the set of morphisms in Qnd4s from SQ(M) to Gec (see [N2, §4.1] for detail). Remark a natural bijection HomQnd4s(SQ(M),Gec) ≃ Hom(4sQndD12,D23,D34)(e12,e23,e34)(SQ(M),Gec) described in [N2, Remark 4.4]. By the correspon- dence (4) and Proposition 3.1, we thus have a bijection
HomQnd4s(SQ(M),Gec)≃Colee12,e23,e34
Gc (Dϕ). (9)
Although the definition of SQ(M) seems ad hoc, we give its concrete presentation as follows:
Corollary 3.6. For a 3-manifold M, there exists a symmetric quandle isomorphism SQ(M)∼=G(M^)c(M), where (
G(M), c(M))
is a cored group (π1(M)×Z/2Z,(e,1)).
Proof. Let Dϕ be a 3-fold labeled diagram which presents M in Lemma 3.2. Recall an equivalence of categories betweenQnd4sand the category of cored groups (see [N2, Corol- lary 4.3]). Hence, there exists a bijection
HomGrpc(G(M), G)≃HomQnd4s(G(M^)c(M),Gec),
for any cored group (G, c). By the canonical bijections (6) and (9), we have a natural equivalence of the following functors fromQnd4s to the category of sets:
HomQnd4s(SQ(M),e•c)≃HomQnd4s(G(M^)c(M),e•c).
Hence, by Yoneda embedding, we conclude SQ(M)∼=G(M^)c(M).
3.4 4-fold symmetric quandle homotopy invariants as natural transformations Furthermore, we define a fundamental class of M, and give an interpretation of the 4-fold symmetric quandle homotopy invariant as a natural transformation.
We fix a 3-fold labeled diagram Dϕ which presents a 3-manifold M. Let us regard HomQnd4s(SQ(M),e•) as a functor from the category of 4-fold symmetric quandles. Fur- ther, we interpret the group Π4f2,ρ(e•) described in §2 as such a functor. We now iden- tify HomQnd4s(SQ(M),Gec) with Colee12,e23,e34
Gc,ρ (Dϕ) by (9). Thus the map Ξ4fe
Gc(Dϕ,†) : Colee12,e23,e34
Gc,ρ (Dϕ)→Π4f2,ρ(Gec) can be regarded as a natural transformation:
Ξ4fe• (Dϕ;†) : HomQnd4s(SQ(M),e•)−→Π4f2,ρ(e•). (10) Let us consider a set of such natural transformations: by Yoneda lemma, we have a bijection
Nat(
HomQnd4s(SQ(M),e•),Π4f2,ρ(e•))
≃Π4f2,ρ(SQ(M)),
which sends Ξe4f• (Dϕ;†) to Ξ4fSQ(M)(Dϕ; idSQ(M)), where idSQ(M) is the identity map of SQ(M).
Definition 3.7. LetM be a 3-manifold, andSQ(M) the fundamental symmetric quandle of M. A fundamental class of M is defined to be Ξ4fSQ(M)(Dϕ; idSQ(M))∈Π4f2,ρ(SQ(M)).
By the naturality, we can reformulate the formula (2) of the 4-fold homotopy invariant as Ξ4fe
Gc(M) = |G|3 · ∑
F∈HomQnd4s(SQ(M),Gec)
F∗(
Ξ4fSQ(M)(Dϕ; idSQ(M)))
∈Z[Π4f2,ρ(Gec)]. (11)
In conclusion, the study of the 4-fold symmetric quandle homotopy invariant of M is roughly the research of Π4f2,ρ(SQ(M)) and of the fundamental class using relativity to other 4-fold symmetric quandles Gec.
Remark 3.8. We compare the fundamental classes of knots with those of 3-manifolds. In the theory of quandle homotopy invariants valued in π2(BX), the second author showed that, for any non-trivial knotsK, the homotopy group of the “knot quandle” is isomorphic toZ generated by the “fundamental class” (see [N1, Corollary 4.17]). On the other hand, on the 4-fold homotopy invariant of 3-manifolds M, Π4f2,ρ(SQ(M)) is always neither Znor generated by the fundamental class, but Π4f2,ρ(SQ(M)) does depend onM.
4 Formulas for the connected sum and the opposite orientation
In this section, we show the formulas of the 4-fold quandle homotopy invariant for the connected sum and the opposite orientation.
Proposition 4.1. LetM1 and M2 be 3-manifolds. LetM1#M2 denote the connected sum of M1 and M2. For a finite cored group (G, c),
Ξ4fGe
c(M1)·Ξ4fGe
c(M2) = |G|3 ·Ξ4fGe
c(M1#M2) ∈Z[Π4f2,ρ(Gec)]. (12)