This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Tensor products of quandles and 1-handles attached to surface-links

Seiichi Kamada Department of Mathematics, Osaka University, Toyonaka, Osaka 560-0043, Japan [email protected]
Abstract.

A quandle is an algebra with two binary operations satisfying three conditions which are related to Reidemeister moves in knot theory. In this paper we introduce the notion of the (canonical) tensor product of a quandle. The tensor product of the knot quandle or the knot symmetric quandle of a surface-link in 44-space can be used to classify or construct invariants of 11-handles attaching to the surface-link. We also compute the tensor products for dihedral quandles and their symmetric doubles.

Key words and phrases:
quandles, tensor products, surface-links, 11-handles
2010 Mathematics Subject Classification:
57Q45

1. Introduction

A quandle X=(X,,¯)X=(X,\ast,\overline{\ast}) is a set XX with two binary operations \ast and ¯\overline{\ast} satisfying conditions related to Reidemeister moves in knot theory (cf. [6, 14] and Section 2). A symmetric quandle is a quandle equipped with an involution called a good involution ([8, 9, 12]). For a given quandle XX, there is a symmetric quandle (D(X),ρ)(D(X),\rho), called the symmetric double of XX (Example 2.3). Quandles and symmetric quandles are used in knot theory for construction of invariants of links and surface-links.

In this paper, we define the tensor product XXX\otimes X of a quandle XX. We give a method of computing the tensor product D(X)D(X)D(X)\otimes D(X) of the symmetric double D(X)D(X) from the tensor product XXX\otimes X of XX (Theorem 3.4). Let τ:XXXX\tau:X\otimes X\to X\otimes X be the involution switching the components, i.e., τ(x,y)=(y,x)\tau(x,y)=(y,x). When XX is equipped with a good involution ρ:XX\rho:X\to X, we have another involution ρ\rho (=ρρ=\rho\otimes\rho) :XXXX:X\otimes X\to X\otimes X with ρ(x,y)=(ρ(x),ρ(y))\rho(x,y)=(\rho(x),\rho(y)). We consider the quotient set XX/τX\otimes X/\langle\tau\rangle (XX/ρX\otimes X/\langle\rho\rangle or XX/τ,ρX\otimes X/\langle\tau,\rho\rangle) of XXX\otimes X by the equivalence relation generated by τ\tau (by ρ\rho or by τ\tau and ρ\rho). It will be seen that the tensor product XXX\otimes X itself and the quotient set XX/τX\otimes X/\langle\tau\rangle (or XX/τ,ρX\otimes X/\langle\tau,\rho\rangle) of XXX\otimes X are useful for studying 11-handles attached to surface-links.

We compute the tensor product RnRnR_{n}\otimes R_{n} for the dihedral quandle RnR_{n} of every order nn (Theorems 5.1 and 5.3) in Section 5. Once we know the elements of RnRnR_{n}\otimes R_{n}, we can obtain the quotient set RnRn/τR_{n}\otimes R_{n}/\langle\tau\rangle (Corollaries 5.2 and 5.4). For the symmetric double (D(Rn),ρ)(D(R_{n}),\rho) of RnR_{n}, using Theorem 3.4, we can compute D(Rn)D(Rn)D(R_{n})\otimes D(R_{n}) from RnRnR_{n}\otimes R_{n} (Theorems 6.1 and 6.5), and then we obtain the quotient sets D(Rn)D(Rn)/τD(R_{n})\otimes D(R_{n})/\langle\tau\rangle, D(Rn)D(Rn)/ρD(R_{n})\otimes D(R_{n})/\langle\rho\rangle and D(Rn)D(Rn)/τ,ρD(R_{n})\otimes D(R_{n})/\langle\tau,\rho\rangle in Section 6.

By a surface-link, we mean a closed surface embedded in 4\mathbb{R}^{4}. The surface may be disconnected or non-orientable. When it is orientable and oriented, it is called an oriented surface-link. A surface-link is called a surface-knot when it is connected. When FF is oriented, the knot quandle Q(F)Q(F) is defined (cf. [3, 6, 8, 9, 12, 14] and Example 2.5). For any surface-link FF, the knot symmetric quandle (Q~(F),ρ)(\widetilde{Q}(F),\rho) is defined (cf. [8, 9, 12] and Example 2.6).

Using tensor products of quandles, we discuss classification and invariants of 11-handles attached to surface-links.

F. Hosakawa and A. Kawauchi [5] studied unknotted surface-links and surgery along 11-handles. They proved that any oriented surface-link can be transformed to an unknotted oriented surface-link by surgery along oriented 11-handles. For non-orientable surface-links, an analoguos result was shown by the author in [7], that is, any non-orientable surface-link can be transformed to an unknotted non-orientable surface-link by surgery along 11-handles. In this sense, surgery along a 11-handle is often called an unknotting operation for surface-links. Surgery along a 11-handle is also used to construct examples of surface-knots and surface-links.

J. Boyle [1] classified oriented 11-handles attached to an oriented surface-knot FF up to strong equivalence in terms of the knot group G(F):=π1(4F)G(F):=\pi_{1}(\mathbb{R}^{4}\setminus F) and its peripheral subgroup. They are in one-to-one correspondence to double cosets of the knot group G(F)G(F) by the peripheral subgroup. The author [10] extended Boyle’s argument to the case where FF is a surface-knot which may be unoriented or non-orientable. However the arguments in [1] and [10] using knot groups are not applied directly to surface-links which are not connected. Even for surface-knots, classifications of 11-handles stated in terms of the knot groups given in [10] are not easy to discuss. The argument using tensor products discussed in this paper is simple and applicable to any surface-link which may be disconnected or non-orientable.

For a surface-link FF, let s(F)\mathcal{H}^{\rm s}(F) (or w(F)\mathcal{H}^{\rm w}(F)) denote the set of strong (or weak) equivalence classes of 11-handles attached to FF. (See Section 7 for the definition of strong/weak equivalence.)

Theorem 1.1.

Let FF be a surface-link and (Q~(F),ρ)(\widetilde{Q}(F),\rho) the knot symmetric quandle of FF. There are bijections

(1) s(F)Q~(F)Q~(F)andw(F)Q~(F)Q~(F)/τ,ρ.\mathcal{H}^{\rm s}(F)\longleftrightarrow\widetilde{Q}(F)\otimes\widetilde{Q}(F)\quad\mbox{and}\quad\mathcal{H}^{\rm w}(F)\longleftrightarrow\widetilde{Q}(F)\otimes\widetilde{Q}(F)/\langle\tau,\rho\rangle.

This theorem is proved as Theorems 8.2 and 9.2.

When FF is an oriented surface-link, a 11-handle attached to FF is called an oriented 11-handle if the orientation is compatible with the orientation of FF (Definition 7.5). Let s(F)ori\mathcal{H}^{\rm s}(F)^{\rm ori} (or w(F)ori\mathcal{H}^{\rm w}(F)^{\rm ori}) denote the set of strong (or weak) equivalence classes of oriented 11-handles attached to FF.

Theorem 1.2.

Let FF be an oriented surface-link and Q(F)Q(F) the knot quandle of FF. There are bijections

(2) s(F)oriQ(F)Q(F)andw(F)oriQ(F)Q(F)/τ.\mathcal{H}^{\rm s}(F)^{\rm ori}\longleftrightarrow Q(F)\otimes Q(F)\quad\mbox{and}\quad\mathcal{H}^{\rm w}(F)^{\rm ori}\longleftrightarrow Q(F)\otimes Q(F)/\langle\tau\rangle.

This theorem is proved as Theorems 8.4 and 9.4.

Theorems 1.1 and 1.2 give complete classifications of all 11-handles and oriented 11-handles in terms of the tensor products of quandles. When Q~(F)\widetilde{Q}(F) is finite, the tensor product Q~(F)Q~(F)\widetilde{Q}(F)\otimes\widetilde{Q}(F) and its quotient Q~(F)Q~(F)/τ,ρ\widetilde{Q}(F)\otimes\widetilde{Q}(F)/\langle\tau,\rho\rangle are finite and we can list the elements theoretically. However, when Q~(F)\widetilde{Q}(F) is infinite or has a large cardinality, it is difficult to compute or list the elements of the tensor product and its quotient. In such a case, it is useful to construct an invariant. Theorems 1.1 and 1.2 can be used to construct invariants of 11-handles as follows:

Theorem 1.3.

Let FF be a surface-link and (Q~(F),ρ)(\widetilde{Q}(F),\rho) the knot symmetric quandle of FF. Let f:(Q~(F),ρ)(X,ρ)f:(\widetilde{Q}(F),\rho)\to(X,\rho) be a symmetric quandle homomorphism to a finite symmetric quandle (X,ρ)(X,\rho). We have invariants of 11-handles,

(3) s(F)XXandw(F)XX/τ,ρ.\mathcal{H}^{\rm s}(F)\longrightarrow X\otimes X\quad\mbox{and}\quad\mathcal{H}^{\rm w}(F)\longrightarrow X\otimes X/\langle\tau,\rho\rangle.
Theorem 1.4.

Let FF be an oriented surface-link and Q(F)Q(F) the knot quandle of FF. Let f:Q(F)Xf:Q(F)\to X be a quandle homomorphism to a finite quandle XX. We have invariants of 11-handles,

(4) s(F)oriXXandw(F)oriXX/τ.\mathcal{H}^{\rm s}(F)^{\rm ori}\longrightarrow X\otimes X\quad\mbox{and}\quad\mathcal{H}^{\rm w}(F)^{\rm ori}\longrightarrow X\otimes X/\langle\tau\rangle.

This paper is organized as follow: In Section 2 we recall quandles and symmetric quandles. In Section 3 the tensor product of a quandle is defined. For a quandle XX, we give a method of computing the tensor product D(X)D(X)D(X)\otimes D(X) of the symmetric double D(X)D(X) from the tensor product XXX\otimes X (Theorem 3.4). In Section 4 for a quandle XX or a symmetric quandle (X,ρ)(X,\rho), involutions τ\tau and ρ\rho on the tensor product XXX\otimes X and the quotients XX/τX\otimes X/\langle\tau\rangle, XX/ρX\otimes X/\langle\rho\rangle and XX/τ,ρX\otimes X/\langle\tau,\rho\rangle are defined. In Section 5 the tensor product RnRnR_{n}\otimes R_{n} of the dihedral quandle RnR_{n} is computed. In Section 6 the tensor product D(Rn)D(Rn)D(R_{n})\otimes D(R_{n}) and its quotients for the symmetric double D(Rn)D(R_{n}) of RnR_{n} are computed. In Section 7 the definition of strong/weak equivalence of 11-handles is given. We recall the notion of chords attached to a surface-link and a relationship between 11-handles and chords. In Section 8 we discuss strong equivalence classes of 11-handles and show a bijection from the strong equivalence classes of 11-handles (or oriented 11-handles) attached to FF to the tensor product Q~(F)Q~(F)\widetilde{Q}(F)\otimes\widetilde{Q}(F) (or Q(F)Q(F)Q(F)\otimes Q(F)) (Theorems 8.2 and 8.4). In Section 9 we discuss weak equivalence classes of 11-handles and show a bijection from the weak equivalence classes of 11-handles (or oriented 11-handles) attached to FF to the quotient of the tensor product Q~(F)Q~(F)/τ,ρ\widetilde{Q}(F)\otimes\widetilde{Q}(F)/\langle\tau,\rho\rangle (or Q(F)Q(F)/τQ(F)\otimes Q(F)/\langle\tau\rangle) (Theorems 9.2 and 9.4). In Section 10 we give examples and discuss invariants of 11-handles.

This work was supported by JSPS KAKENHI Grant Numbers JP19H01788 and JP17H06128.

2. Quandles and symmetric quandles

A quandle is a set XX with two binary operations \ast and ¯\overline{\ast} satisfying the following three conditions:

  • (Q1)

    For any xXx\in X, xx=xx\ast x=x.

  • (Q2)

    For any x,yXx,y\in X, (xy)¯y=x(x\ast y)\,\overline{\ast}\,y=x and (x¯y)y=x(x\,\overline{\ast}\,y)\ast y=x.

  • (Q3)

    For any x,y,zXx,y,z\in X, (xy)z=(xz)(yz)(x\ast y)\ast z=(x\ast z)\ast(y\ast z).

This notion was introduced independently by Joyce [6] and Matveev [14]. The three conditions correspond to three basic moves on knot diagrams, called Reidemeister moves. A symmetric quandle is a pair (X,ρ)(X,\rho) of a quandle XX and an involution ρ:XX\rho:X\to X, called a good involution on XX, satisfying the following conditions:

  • (S1)

    For any x,yXx,y\in X, ρ(xy)=ρ(x)y\rho(x\ast y)=\rho(x)\ast y,

  • (S2)

    For any x,yXx,y\in X, xρ(y)=x¯yx\ast{\rho(y)}=x\,\overline{\ast}\,y.

A symmetric quandle homomorphism f:(X,ρ)(Y,ρ)f:(X,\rho)\to(Y,\rho) means a quandle homomorphism f:XYf:X\to Y with fρ=ρff\circ\rho=\rho\circ f. Refer to [8, 9, 12] for details on symmetric quandles and related topics.

We here give some examples of quandles and symmetric quandles.

Example 2.1.

(Dihedral quandles) Let X=/n={0,1,,n1}X=\mathbb{Z}/n\mathbb{Z}=\{0,1,\cdots,n-1\}. Define ab=a¯b=2baa\ast b=a\,\overline{\ast}\,b=2b-a. Then X=(X,,¯)X=(X,\ast,\,\overline{\ast}\,) is a quandle, which is called the dihedral quandle of order nn. It is denoted by RnR_{n} in this paper.

Let ρ:RnRn\rho:R_{n}\to R_{n} be the identity map. It is a good involution, and we have a symmetric quandle (Rn,ρ)(R_{n},\rho).

Suppose that n=2mn=2m and let ρ:RnRn\rho^{\prime}:R_{n}\to R_{n} be the map sending xx to x+mx+m for all xRnx\in R_{n}. Then ρ\rho^{\prime} is a good involution of RnR_{n}, and we have another symmetric quandle (Rn,ρ)(R_{n},\rho^{\prime}) whose underlying quandle is the dihedral quandle. Good involutions on dihedral quandles are classified in [12].

Example 2.2.

(Conjugation quandles) Let GG be a group. Let XX be GG as a set. Define ab=b1aba\ast b=b^{-1}ab and a¯b=bab1a\,\overline{\ast}\,b=bab^{-1}. Then X=(X,,¯)X=(X,\ast,\,\overline{\ast}\,) is a quandle, which is called the conjugation quandle of GG (cf. [6, 14]).

Let ρ:XX\rho:X\to X be the map sending xx to x1x^{-1} for all xXx\in X. It is a good involution on the conjugation quandle, and we have a symmetric quandle (X,ρ)(X,\rho) (cf. [8, 9, 12]).

Example 2.3.

(Symmetric doubles, cf. [8, 9, 12]) Let X=(X,,¯)X=(X,\ast,\,\overline{\ast}\,) be a quandle. Let X+X^{+} and XX^{-} be copies of XX. For each xXx\in X, we denote by x+x^{+} and xx^{-} the elements of X+X^{+} and XX^{-} corresponding to xx. Let D(X)=X+XD(X)=X^{+}\amalg X^{-} and define binary operations \ast and ¯\overline{\ast} on D(X)D(X) by

(5) x±y+=(xy)±,x±y=(x¯y)±,x±¯y+=(x¯y)±,x±¯y=(xy)±,\begin{array}[]{ll}x^{\pm}\ast y^{+}=(x\ast y)^{\pm},&x^{\pm}\ast y^{-}=(x\,\overline{\ast}\,y)^{\pm},\\ x^{\pm}\,\overline{\ast}\,y^{+}=(x\,\overline{\ast}\,y)^{\pm},&x^{\pm}\,\overline{\ast}\,y^{-}=(x\ast y)^{\pm},\end{array}

where x±x^{\pm} means x+x^{+} or xx^{-}, respectively. Then D(X)=(D(X),,¯)D(X)=(D(X),\ast,\,\overline{\ast}\,) is a quandle, which we call the symmetric double of XX. (The subquandle X+X^{+} of D(X)D(X) can be identified with the original quandle XX by identifying x+x^{+} with xx. Elements x+x^{+} and xx^{-} are often denoted by xx and x¯\overline{x}, respectively.) The involution ρ:D(X)D(X)\rho:D(X)\to D(X) interchanging x+x^{+} and xx^{-} for all xXx\in X is a good involution. The symmetric quandle (D(X),ρ)(D(X),\rho), with this particular ρ\rho, is also referred to as the symmetric double of XX.

Remark 2.4.

Let XX and YY be quandles and let f:XYf:X\to Y be a quandle homomorphism. Let ff also denote the map f:D(X)D(Y)f:D(X)\to D(Y) sending xϵx^{\epsilon} to f(x)ϵf(x)^{\epsilon} for xXx\in X and ϵ{+,}\epsilon\in\{+,-\}. Then f:D(X)D(Y)f:D(X)\to D(Y) is a quandle homomorphism. Moreover, when (D(X),ρ)(D(X),\rho) and (D(Y),ρ)(D(Y),\rho) are the symmetric doubles of XX and YY, the map f:D(X)D(Y)f:D(X)\to D(Y) is a symmetric quandle homomorphism, i.e., it is a quandle homomorphism with f(ρ(a))=ρ(f(a))f(\rho(a))=\rho(f(a)) for aD(X)a\in D(X).

Example 2.5.

(Knot quandles, cf. [3, 6, 9, 14]) Let KK be an oriented closed nn-manifold embedded in n+2\mathbb{R}^{n+2}. Let Q(K)={[(D,a)]}{Q}(K)=\{[(D,a)]\} be the set of homotopy classes of pairs (D,a)(D,a) such that DD is a positively oriented meridian disk of KK and aa is an arc in the knot exterior connecting DD and the base point.

Define

(6) [(D1,a1)][(D2,a2)]=[(D1,a1a21D2a2)],[(D1,a1)]¯[(D2,a2)]=[(D1,a1a21(D2)a2)].\begin{array}[]{lll}~{}[(D_{1},a_{1})]\ast[(D_{2},a_{2})]&=&[(D_{1},a_{1}\cdot a_{2}^{-1}\cdot\partial D_{2}\cdot a_{2})],\\ ~{}[(D_{1},a_{1})]\,\overline{\ast}\,~{}[(D_{2},a_{2})]&=&[(D_{1},a_{1}\cdot a_{2}^{-1}\cdot\partial(-D_{2})\cdot a_{2})].\end{array}

See Figure 1. The knot quandle Q(K){Q}(K) of KK is (Q(K),,¯)({Q}(K),\ast,\overline{\ast}). The knot quandle of KK can be computed from a diagram of KK (cf. [2, 6, 9, 14]).

Figure 1. Knot quandle operations
Example 2.6.

(Full knot quandles and knot symmetric quandles, cf. [8, 9, 12]) Let KK be a closed nn-manifold embedded in n+2\mathbb{R}^{n+2}, which may be non-orientable. Let Q~(K)={[(D,a)]}\widetilde{Q}(K)=\{[(D,a)]\} be the set of homotopy classes of pairs (D,a)(D,a) such that DD is an oriented meridian disk of KK and aa is an arc in the knot exterior connecting DD and the base point. Define binary operations \ast and ¯\,\overline{\ast}\, on Q~(K)\widetilde{Q}(K) by the same way as the operations of the knot quandle above. Then (Q~(K),,¯)(\widetilde{Q}(K),\ast,\overline{\ast}) is a quandle, which we call the full knot quandle Q~(K)\widetilde{Q}(K) of KK.

The standard involution is ρ:Q~(K)Q~(K)\rho:\widetilde{Q}(K)\to\widetilde{Q}(K), ρ[(D,a)]=[(D,a)]\rho[(D,a)]=[(-D,a)], which is a good involution. See Figure 2. The knot symmetric quandle of KK is the pair (Q~(K),ρ)(\widetilde{Q}(K),\rho) of the full knot quandle and its standard involution.

Remark 2.7.

When KK is oriented, the knot quandle Q(K)Q(K) is a subquandle of Q~(K)\widetilde{Q}(K), and the knot symmetric quandle (Q~(K),ρ)(\widetilde{Q}(K),\rho) is identified with the symmetric double (D(Q(K)),ρ)(D(Q(K)),\rho) of Q(K)Q(K) (cf. [8, 9, 12]).

Figure 2. The standard involution ρ\rho
Remark 2.8.

For a quandle XX, the associated group of XX is defined by

(7) As(X)=xXxy=y1xyfor x,yX.As(X)=\langle x\in X\mid x\ast y=y^{-1}xy\quad\mbox{for $x,y\in X$}\rangle.

It is known that for an oriented closed nn-manifold KK embedded in n+2\mathbb{R}^{n+2}, the knot group G(K)=π1(n+2K)G(K)=\pi_{1}(\mathbb{R}^{n+2}\setminus K) is isomorphic to As(Q(K))As(Q(K)) (cf. [3, 6, 14]).

For a symmetric quandle (X,ρ)(X,\rho), the associated group of (X,ρ)(X,\rho) is defined by

(8) As(X,ρ)=xXxy=y1xy,ρ(x)=x1for x,yX.As(X,\rho)=\langle x\in X\mid x\ast y=y^{-1}xy,~{}\rho(x)=x^{-1}\quad\mbox{for $x,y\in X$}\rangle.

For a closed nn-manifold KK embedded in n+2\mathbb{R}^{n+2}, the knot group G(K)G(K) is isomorphic to As(Q~(K),ρ)As(\widetilde{Q}(K),\rho) (Theorem 4.5 of [9]).

3. Tensor products of quandles

Suppose that a group GG acts on a quandle X1X_{1} from the right and on a quandle X2X_{2} from the left; x1(gg)=(x1g)gx_{1}\cdot(gg^{\prime})=(x_{1}\cdot g)\cdot g^{\prime} and (gg)x2=g(gx2)(gg^{\prime})\cdot x_{2}=g\cdot(g^{\prime}\cdot x_{2}). Consider the equivalence relation on X1×X2X_{1}\times X_{2} generated by

(9) (x1g,x2)(x1,gx2)for x1X1,x2X2, and gG.(x_{1}\cdot g,\,x_{2})\sim(x_{1},\,g\cdot x_{2})\quad\mbox{for $x_{1}\in X_{1},\,x_{2}\in X_{2},$ and $g\in G.$}
Definition 3.1.

The tensor product of X1X_{1} and X2X_{2} over GG, denoted by X1GX2X_{1}\otimes_{G}X_{2}, is the set of all equivalence classes [(x1,x2)][(x_{1},x_{2})] for (x1,x2)X1×X2(x_{1},x_{2})\in X_{1}\times X_{2}. An element [(x1,x2)][(x_{1},x_{2})] of X1GX2X_{1}\otimes_{G}X_{2} is also denoted by [x1,x2][x_{1},x_{2}], x1x2x_{1}\otimes x_{2} or (x1,x2)(x_{1},x_{2}).

Let XX be a quandle. Let F(X)F(X) be the free group generated by XX. The free group F(X)F(X) acts on XX from the right by

(10) xg:=(((xϵ1x1)ϵ2x2))ϵmxmx\cdot g:=(((x\ast^{\epsilon_{1}}x_{1})\ast^{\epsilon_{2}}x_{2})\cdots)\ast^{\epsilon_{m}}x_{m}

for xXx\in X and g=x1ϵ1x2ϵ2xmϵmGg=x_{1}^{\epsilon_{1}}x_{2}^{\epsilon_{2}}\cdots x_{m}^{\epsilon_{m}}\in G, where 1=\ast^{1}=\ast and 1=¯\ast^{-1}=\overline{\ast}. For example, xabc1d=(((xa)b)¯c)dx\cdot abc^{-1}d=(((x\ast a)\ast b)\,\overline{\ast}\,c)\ast d. This action is called the canonical action of F(X)F(X) on XX. The free group F(X)F(X) acts on XX from the left by

(11) gx:=xg1=(((xϵmxm)ϵm1xm1))ϵ1x1g\cdot x:=x\cdot g^{-1}=(((x\ast^{-\epsilon_{m}}x_{m})\ast^{-\epsilon_{m-1}}x_{m-1})\cdots)\ast^{-\epsilon_{1}}x_{1}

for xXx\in X and g=x1ϵ1x2ϵ2xmϵmGg=x_{1}^{\epsilon_{1}}x_{2}^{\epsilon_{2}}\cdots x_{m}^{\epsilon_{m}}\in G.

Definition 3.2.

The canonical tensor product of XX is the tensor product XF(X)XX\otimes_{F(X)}X of XX over the canonical action of F(X)F(X). In this paper, we refer to the canonical tensor product as the tensor product of XX and denote it simply by XXX\otimes X.

Note that

(12) XX=X×X/(x1g,x2)(x1,gx2)=X×X/(x1g,g1x2)(x1,x2)=X×X/(x1g,x2g)(x1,x2).\begin{array}[]{rl}X\otimes X&=X\times X/(x_{1}\cdot g,\,x_{2})\sim(x_{1},\,g\cdot x_{2})\\ &=X\times X/(x_{1}\cdot g,\,g^{-1}\cdot x_{2})\sim(x_{1},\,x_{2})\\ &=X\times X/(x_{1}\cdot g,\,x_{2}\cdot g)\sim(x_{1},\,x_{2}).\\ \end{array}

Thus, the tensor product XXX\otimes X is the orbit space of X×XX\times X by the right action of F(X)F(X) with (x1,x2)g:=(x1g,x2g)(x_{1},\,x_{2})\cdot g:=(x_{1}\cdot g,\,x_{2}\cdot g). We also refer to this action of F(X)F(X) on X×XX\times X, (x1,x2)g:=(x1g,x2g)(x_{1},\,x_{2})\cdot g:=(x_{1}\cdot g,\,x_{2}\cdot g), as the canonical right action of F(X)F(X) on X×XX\times X.

Example 3.3.

Let R3R_{3} be the dihedral quandle of order 33. As a set, R3=/3={0,1,2}R_{3}=\mathbb{Z}/3\mathbb{Z}=\{0,1,2\} and R3×R3={(i,j)i,j{0,1,2}}R_{3}\times R_{3}=\{(i,j)\mid i,j\in\{0,1,2\}\}. Note that (i,j)(ik,jk)=(ik,jk)=(2ki,2kj)(i,j)\sim(i\cdot k,j\cdot k)=(i\ast k,j\ast k)=(2k-i,2k-j) for any i,j,ki,j,k.

For example, (0,0)(01,01)=(2,2)(0,0)\sim(0\ast 1,0\ast 1)=(2,2) and (0,1)(01,11)=(2,1)(0,1)\sim(0\ast 1,1\ast 1)=(2,1).

Considering the canonical right actions of 0,10,1 and 22 on R3×R3R_{3}\times R_{3}, we see that

(13) [0,0]={(0,0),(1,1),(2,2)}[0,1]={(0,1),(0,2),(1,0),(1,2),(2,0),(2,1)},\begin{array}[]{ll}~{}[0,0]&=\{(0,0),(1,1),(2,2)\}\\ ~{}[0,1]&=\{(0,1),(0,2),(1,0),(1,2),(2,0),(2,1)\},\end{array}

and

(14) R3R3={[0,0],[0,1]}.R_{3}\otimes R_{3}=\{[0,0],[0,1]\}.

In Section 5 we compute the tensor product RnRnR_{n}\otimes R_{n} for every dihedral qaundle RnR_{n}.

We introduce a method of computing D(X)D(X)D(X)\otimes D(X) from XXX\otimes X.

Let XX be a quandle and let D(X)D(X) be the symmetric double. Note that D(X)=X+X={x+,xxX}D(X)=X^{+}\amalg X^{-}=\{x^{+},x^{-}\mid x\in X\}. For each element (x,y)X×X(x,y)\in X\times X and ϵ,δ{+,}\epsilon,\delta\in\{+,-\}, let (x,y)ϵ,δ(x,y)^{\epsilon,\delta} denote the element (xϵ,yδ)D(X)×D(X)(x^{\epsilon},y^{\delta})\in D(X)\times D(X). Note that every element of D(X)×D(X)D(X)\times D(X) can be written in this form. For a subset EE of X×XX\times X and ϵ,δ{+,}\epsilon,\delta\in\{+,-\}, let Eϵ,δE^{\epsilon,\delta} denote the subset of D(X)×D(X)D(X)\times D(X) with

(15) Eϵ,δ={(x,y)ϵ,δ(x,y)E}.E^{\epsilon,\delta}=\{(x,y)^{\epsilon,\delta}\mid(x,y)\in E\}.
Theorem 3.4.

For any quandle XX,

(16) D(X)D(X)={Eϵ,δEXXandϵ,δ{+,}}.D(X)\otimes D(X)=\{E^{\epsilon,\delta}\mid E\in X\otimes X~{}\mbox{{and}}~{}\epsilon,\delta\in\{+,-\}\}.

We provide two lemmas.

Let α+:F(X)F(D(X))\alpha^{+}:F(X)\to F(D(X)) be the homomorphism determined by α+(x)=x+\alpha^{+}(x)=x^{+} for xXx\in X. (Then α+(x1)=(x+)1\alpha^{+}(x^{-1})=(x^{+})^{-1} for xXx\in X.)

Let β:F(D(X))F(X)\beta:F(D(X))\to F(X) be the homomorphism determined by β(x+)=x\beta(x^{+})=x and β(x)=x1\beta(x^{-})=x^{-1} for xXx\in X. (Then β((x+)1)=x1\beta((x^{+})^{-1})=x^{-1} and β((x)1)=x\beta((x^{-})^{-1})=x for xXx\in X.)

Lemma 3.5.

Let xx, x0Xx_{0}\in X, gF(X)g\in F(X), hF(D(X))h\in F(D(X)), and ϵ{+,}\epsilon\in\{+,-\}.

  • (1)

    If x=x0gx=x_{0}\cdot g, then xϵ=(x0)ϵα+(g)x^{\epsilon}=(x_{0})^{\epsilon}\cdot\alpha^{+}(g).

  • (2)

    If xϵ=(x0)ϵhx^{\epsilon}=(x_{0})^{\epsilon}\cdot h, then x=x0β(h)x=x_{0}\cdot\beta(h).

Proof.  It is directly seen from the definition of the canonical action. ∎

Lemma 3.6.

Let EXXE\in X\otimes X and E=[(x0,y0)]E=[(x_{0},y_{0})]. Then Eϵ,δ=[(x0,y0)ϵ,δ]E^{\epsilon,\delta}=[(x_{0},y_{0})^{\epsilon,\delta}].

Proof.  Let (x,y)ϵ,δ(x,y)^{\epsilon,\delta} be any element of Eϵ,δE^{\epsilon,\delta}. By definition, (x,y)E(x,y)\in E. There is an element gF(X)g\in F(X) with (x,y)=(x0,y0)g(x,y)=(x_{0},y_{0})\cdot g. By Lemma 3.5, (x,y)ϵ,δ=(x0,y0)ϵ,δα+(g)(x,y)^{\epsilon,\delta}=(x_{0},y_{0})^{\epsilon,\delta}\cdot\alpha_{+}(g). Thus, (x,y)ϵ,δ[(x0,y0)ϵ,δ](x,y)^{\epsilon,\delta}\in[(x_{0},y_{0})^{\epsilon,\delta}]. Hence, Eϵ,δ[(x0,y0)ϵ,δ]E^{\epsilon,\delta}\subset[(x_{0},y_{0})^{\epsilon,\delta}].

Conversely, let (x,y)ϵ,δD(X)×D(X)(x,y)^{\epsilon,\delta}\in D(X)\times D(X) be any element of [(x0,y0)ϵ,δ][(x_{0},y_{0})^{\epsilon,\delta}]. There is an element hF(D(X))h\in F(D(X)) with (x,y)ϵ,δ=(x0,y0)ϵ,δh(x,y)^{\epsilon,\delta}=(x_{0},y_{0})^{\epsilon,\delta}\cdot h. By Lemma 3.5, (x,y)=(x0,y0)β(h)E(x,y)=(x_{0},y_{0})\cdot\beta(h)\in E. Thus, (x,y)ϵ,δEϵ,δ(x,y)^{\epsilon,\delta}\in E^{\epsilon,\delta}. Hence, [(x0,y0)ϵ,δ]Eϵ,δ[(x_{0},y_{0})^{\epsilon,\delta}]\subset E^{\epsilon,\delta}. ∎

Proof of Theorem 3.4.  It follows from Lemma 3.6. ∎

The following is seen from Example 3.3 and Theorem 3.4.

Example 3.7.

Let D(R3)D(R_{3}) be the symmetric double of R3R_{3}. Then

(17) D(R3)D(R3)={[0,0]ϵ,δ,[0,1]ϵ,δϵ,δ{+,}},D(R_{3})\otimes D(R_{3})=\{[0,0]^{\epsilon,\delta},[0,1]^{\epsilon,\delta}\mid\epsilon,\delta\in\{+,-\}\},

where

(18) [0,0]ϵ,δ={(0,0)ϵ,δ,(1,1)ϵ,δ,(2,2)ϵ,δ},[0,1]ϵ,δ={(0,1)ϵ,δ,(0,2)ϵ,δ,(1,0)ϵ,δ,(1,2)ϵ,δ,(2,0)ϵ,δ,(2,1)ϵ,δ},\begin{array}[]{ll}~{}[0,0]^{\epsilon,\delta}&=\{(0,0)^{\epsilon,\delta},(1,1)^{\epsilon,\delta},(2,2)^{\epsilon,\delta}\},\\ ~{}[0,1]^{\epsilon,\delta}&=\{(0,1)^{\epsilon,\delta},(0,2)^{\epsilon,\delta},(1,0)^{\epsilon,\delta},(1,2)^{\epsilon,\delta},(2,0)^{\epsilon,\delta},(2,1)^{\epsilon,\delta}\},\\ \end{array}

for ϵ,δ{+,}\epsilon,\delta\in\{+,-\}. Thus D(R3)D(R3)D(R_{3})\otimes D(R_{3}) consists of 88 elements.

4. Involutions on tensor products

In this section we introduce involutions on tensor products.

Let XXX\otimes X be the tensor product of a quandle XX. Let τ:X×XX×X\tau:X\times X\to X\times X be the involution of X×XX\times X switching the components, τ(x1,x2)=(x2,x1)\tau(x_{1},x_{2})=(x_{2},x_{1}). Since τ\tau commutes with the canonical right action of F(X)F(X) on X×XX\times X, it induces an involution on XXX\otimes X, which we denote also by

(19) τ:XXXX,[x1,x2][x2,x1].\tau:X\otimes X\to X\otimes X,\quad[x_{1},x_{2}]\mapsto[x_{2},x_{1}].

Let ρ:XX\rho:X\to X be an involution. It induces an invitation ρ(=ρρ):X×XX×X\rho~{}(=\rho\otimes\rho):X\times X\to X\times X with ρ(x1,x2)=(ρ(x1),ρ(x2))\rho(x_{1},x_{2})=(\rho(x_{1}),\rho(x_{2})). Note that ρ:X×XX×X\rho:X\times X\to X\times X commutes with the canonical right action of F(X)F(X) on X×XX\times X if and only if ρ:XX\rho:X\to X satisfies the first condition (S1) of a good involution.

Let ρ:XX\rho:X\to X be a good involution. Then the involution ρ:X×XX×X\rho:X\times X\to X\times X induces an involution on XXX\otimes X, which we denote also by

(20) ρ:XXXX,[x1,x2][ρ(x1),ρ(x2)].\rho:X\otimes X\to X\otimes X,\quad[x_{1},x_{2}]\mapsto[\rho(x_{1}),\rho(x_{2})].

We denote by XX/τX\otimes X/\langle\tau\rangle (XX/ρX\otimes X/\langle\rho\rangle or XX/τ,ρX\otimes X/\langle\tau,\rho\rangle, resp.) the set of equivalence classes of XXX\otimes X by the equivalence relation generated by τ\tau (by ρ\rho or by τ\tau and ρ\rho, resp.).

Example 4.1.

Let R3R_{3} be the dihedral quandle of order 33. Recall that R3R3={[0,0],[0,1]}R_{3}\otimes R_{3}=\{[0,0],[0,1]\} as in Example 3.3. Since τ([0,0])=[0,0]\tau([0,0])=[0,0] and τ([0,1])=[0,1]\tau([0,1])=[0,1], we have

(21) R3R3/τ={{[0,0]},{[0,1]}}.R_{3}\otimes R_{3}/\langle\tau\rangle=\{\{[0,0]\},\{[0,1]\}\}.

In Section 5 we compute RnRn/τR_{n}\otimes R_{n}/\langle\tau\rangle for every dihedral qaundle RnR_{n}.

Example 4.2.

Let (D(R3),ρ)(D(R_{3}),\rho) be the symmetric double of R3R_{3}. Recall that D(R3)D(R3)={[0,0]ϵ,δ,[0,1]ϵ,δ,ϵ,δ{+,}}D(R_{3})\otimes D(R_{3})=\{[0,0]^{\epsilon,\delta},[0,1]^{\epsilon,\delta},\mid\epsilon,\delta\in\{+,-\}\} as in Example 3.7. Note that for k=0,1k=0,1,

(22) τ([0,k]+,+)=[0,k]+,+,τ([0,k],)=[0,k],,τ([0,k]+,)=[0,k],+,ρ([0,k]+,+)=[0,k],,ρ([0,k]+,)=[0,k],+.\begin{array}[]{ll}\tau([0,k]^{+,+})=[0,k]^{+,+},&\tau([0,k]^{-,-})=[0,k]^{-,-},\\ \tau([0,k]^{+,-})=[0,k]^{-,+},&\\ \rho([0,k]^{+,+})=[0,k]^{-,-},&\rho([0,k]^{+,-})=[0,k]^{-,+}.\\ \end{array}

Thus we see the following.

  • D(R3)D(R3)/τD(R_{3})\otimes D(R_{3})/\langle\tau\rangle consists of

    {[0,k]+,+},{[0,k],},{[0,k]+,,[0,k],+} for k=0,1\{[0,k]^{+,+}\},\{[0,k]^{-,-}\},\{[0,k]^{+,-},[0,k]^{-,+}\}\quad\mbox{ for $k=0,1$. }
  • D(R3)D(R3)/ρD(R_{3})\otimes D(R_{3})/\langle\rho\rangle consists of

    {[0,k]+,+,[0,k],},{[0,k]+,,[0,k],+} for k=0,1.\{[0,k]^{+,+},[0,k]^{-,-}\},\{[0,k]^{+,-},[0,k]^{-,+}\}\quad\mbox{ for $k=0,1$.}
  • D(R3)D(R3)/τ,ρD(R_{3})\otimes D(R_{3})/\langle\tau,\rho\rangle consists of

    {[0,k]+,+,[0,k],},{[0,k]+,,[0,k],+} for k=0,1.\{[0,k]^{+,+},[0,k]^{-,-}\},\{[0,k]^{+,-},[0,k]^{-,+}\}\quad\mbox{ for $k=0,1$.}

The following lemma is straightforward and we omit the proof.

Lemma 4.3.

For a quandle homomorphism f:XYf:X\to Y, we have a map

(23) ff:XXYY,[x1,x2][f(x1),f(x2)].f\otimes f:X\otimes X\to Y\otimes Y,\quad[x_{1},x_{2}]\mapsto[f(x_{1}),f(x_{2})].

It induces a map XX/τYY/τX\otimes X/\langle\tau\rangle\to Y\otimes Y/\langle\tau\rangle. Moreover, for a symmetric quandle homomorphism f:(X,ρ)(Y,ρ)f:(X,\rho)\to(Y,\rho), the map fff\otimes f induces maps XX/ρYY/ρX\otimes X/\langle\rho\rangle\to Y\otimes Y/\langle\rho\rangle and XX/τ,ρYY/τ,ρX\otimes X/\langle\tau,\rho\rangle\to Y\otimes Y/\langle\tau,\rho\rangle.

5. Tensor products of dihedral quandles

In this section we compute tensor products RnRnR_{n}\otimes R_{n} and the quotient set RnRn/τR_{n}\otimes R_{n}/\langle\tau\rangle for every dihedral quandle RnR_{n}.

Theorem 5.1.

Let Rn=R2m+1R_{n}=R_{2m+1} be the dihedral quandle of order n=2m+1n=2m+1. The tensor product RnRnR_{n}\otimes R_{n} consists of m+1m+1 elements, E(0),,E(m)E(0),\dots,E(m), where

(24) E(k)={(i,i+k),(i,ik)iRn} for k=0,1,,m.
E(k)=\{(i,i+k),(i,i-k)\mid i\in R_{n}\}\quad\mbox{ for $k=0,1,\dots,m$.}\\

We prove this theorem later.

Since τ(E(k))=E(k)\tau(E(k))=E(k) for k=0,1,,mk=0,1,\dots,m, we have the following.

Corollary 5.2.

Let Rn=R2m+1R_{n}=R_{2m+1} be the dihedral quandle of order n=2m+1n=2m+1. The quotient set RnRn/τR_{n}\otimes R_{n}/\langle\tau\rangle consists of m+1m+1 elements, {E(k)}\{E(k)\} for k=0,1,,mk=0,1,\dots,m.

When n=3n=3 and m=1m=1, E(0)=[0,0]E(0)=[0,0] and E(1)=[0,1]E(1)=[0,1] in Examples 3.3 and  4.1.

Let Rn=R2mR_{n}=R_{2m} be the dihedral quandle of order n=2mn=2m. We denote by Rn,evenR_{n,{\rm even}} (or Rn,oddR_{n,{\rm odd}}) the subset of RnR_{n} consisting of the elements represented by even numbers (or odd numbers). For example, when n=8n=8, then R8,even={0,2,4,6}R_{8,{\rm even}}=\{0,2,4,6\} and R8,odd={1,3,5,7}R_{8,{\rm odd}}=\{1,3,5,7\}.

Theorem 5.3.

Let Rn=R2mR_{n}=R_{2m} be the dihedral quandle of order n=2mn=2m. The tensor product RnRnR_{n}\otimes R_{n} consists of n+2n+2 (=2m+2=2m+2) elements,

(25) E(k)0andE(k)1for k=0,1,,m,E(k)_{0}\quad\mbox{and}\quad E(k)_{1}\quad\mbox{for $k=0,1,\dots,m$,}

where

(26) E(k)0={(i,i+k),(i,ik)iRn,even}for k=0,1,,m,E(k)1={(i,i+k),(i,ik)iRn,odd}for k=0,1,,m.\begin{array}[]{ccll}E(k)_{0}&=&\{(i,i+k),(i,i-k)\mid i\in R_{n,{\rm even}}\}&\mbox{for $k=0,1,\dots,m$,}\\ E(k)_{1}&=&\{(i,i+k),(i,i-k)\mid i\in R_{n,{\rm odd}}\}&\mbox{for $k=0,1,\dots,m$.}\\ \end{array}

We prove this theorem later.

It is obvious that

(27) τ(E(k)0)=E(k)0,τ(E(k)1)=E(k)1for k=0,2,4,m,τ(E(k)0)=E(k)1for k=1,3,5,m.\begin{array}[]{ll}\tau(E(k)_{0})=E(k)_{0},\quad\tau(E(k)_{1})=E(k)_{1}&\mbox{for $k=0,2,4,\dots\leq m$,}\\ \tau(E(k)_{0})=E(k)_{1}&\mbox{for $k=1,3,5,\dots\leq m$.}\\ \end{array}
Corollary 5.4.

Let Rn=R2mR_{n}=R_{2m} be the dihedral quandle of order n=2mn=2m. The quotient set RnRn/τR_{n}\otimes R_{n}/\langle\tau\rangle consists of 3(m+1)/23(m+1)/2 elements (if mm is odd) or 3m/2+23m/2+2 elements (if mm is even), which are listed below:

(28) {E(k)0},{E(k)1}for k=0,2,4,m,{E(k)0,E(k)1}for k=1,3,5,m.\begin{array}[]{ll}\{E(k)_{0}\},\quad\{E(k)_{1}\}&\mbox{for $k=0,2,4,\dots\leq m$},\\ \{E(k)_{0},E(k)_{1}\}&\mbox{for $k=1,3,5,\dots\leq m$}.\\ \end{array}

Let dn:Rn×Rn0={0}d_{n}:R_{n}\times R_{n}\to\mathbb{N}_{0}=\mathbb{N}\cup\{0\} be the distance on Rn=/nR_{n}=\mathbb{Z}/n\mathbb{Z} induced from the usual distance d0:×0d_{0}:\mathbb{Z}\times\mathbb{Z}\to\mathbb{N}_{0} on \mathbb{Z} with d0(x1,x2)=|x1x2|d_{0}(x_{1},x_{2})=|x_{1}-x_{2}|. Since dn(x1x,x2x)=dn(x1¯x,x2¯x)=d(x1,x2)d_{n}(x_{1}\ast x,x_{2}\ast x)=d_{n}(x_{1}\,\overline{\ast}\,x,x_{2}\,\overline{\ast}\,x)=d(x_{1},x_{2}) for any x1,x2,xRnx_{1},x_{2},x\in R_{n}, we have

(29) dn(x1g,x2g)=dn(x1,x2) for any x1,x2Rn and gF(Rn).d_{n}(x_{1}\cdot g,x_{2}\cdot g)=d_{n}(x_{1},x_{2})\quad\mbox{ for any $x_{1},x_{2}\in R_{n}$ and $g\in F(R_{n})$.}
Lemma 5.5.

Let Rn=R2m+1R_{n}=R_{2m+1} be the dihedral quandle of order n=2m+1n=2m+1. Let k{0,1,,m}k\in\{0,1,\dots,m\}. If dn(x,y)=kd_{n}(x,y)=k, then there exists an element gF(Rn)g\in F(R_{n}) with xg=0x\cdot g=0 and yg=ky\cdot g=k.

Proof.  It is known that RnR_{n} with n=2m+1n=2m+1 is connected as a quandle, i.e., for any xRnx\in R_{n} there exists an element gF(Rn)g\in F(R_{n}) with xg=0x\cdot g=0. In fact, let g=x/2Rng=x/2\in R_{n} if xx is even, or g=(n+x)/2Rng=(n+x)/2\in R_{n} if xx is odd. Then xg=0x\ast g=0.

Thus we have the assertion with k=0k=0.

Suppose 0<km0<k\leq m and dn(x,y)=kd_{n}(x,y)=k. Let g1F(Rn)g_{1}\in F(R_{n}) be an element with xg1=0x\cdot g_{1}=0. By (29), we have yg1=ky\cdot g_{1}=k or k-k. If yg1=ky\cdot g_{1}=k, then g=g1g=g_{1} is a desired one. If yg1=ky\cdot g_{1}=-k, then g=g10g=g_{1}0 is a desired one. ∎

Proof of Theorem 5.1.  For any (x,y)Rn×Rn(x,y)\in R_{n}\times R_{n}, dn(x,y){0,1,,m}d_{n}(x,y)\in\{0,1,\dots,m\}. Since

(30) E(k)={(x,y)Rn×Rndn(x,y)=k},E(k)=\{(x,y)\in R_{n}\times R_{n}\mid d_{n}(x,y)=k\},

we see that E(0),E(1),,E(m)E(0),E(1),\dots,E(m) cover Rn×RnR_{n}\times R_{n} and E(k)E(k)=E(k)\cap E(k^{\prime})=\emptyset for kkk\neq k^{\prime}. By Lemma 5.5, E(k)[0,k]E(k)\subset[0,k]. Conversely, by (29), we see that [0,k]E(k)[0,k]\subset E(k). ∎

Lemma 5.6.

Let Rn=R2mR_{n}=R_{2m} be the dihedral quandle of order n=2mn=2m. Let k{0,1,,m}k\in\{0,1,\dots,m\}. If dn(x,y)=kd_{n}(x,y)=k and if xx is even (or odd), then there exists an element gF(Rn)g\in F(R_{n}) with xg=0x\cdot g=0 and yg=ky\cdot g=k (or xg=1x\cdot g=1 and yg=1+ky\cdot g=1+k).

Proof.  It is known that RnR_{n} with n=2mn=2m has two connected components Rn,evenR_{n,{\rm even}} and Rn,oddR_{n,{\rm odd}}, i.e., xRn,evenx\in R_{n,{\rm even}} (or xRn,oddx\in R_{n,{\rm odd}}) if and only if there exists an element gF(Rn)g\in F(R_{n}) with xg=0x\cdot g=0 (or xg=1x\cdot g=1). In fact, if xRn,evenx\in R_{n,{\rm even}} (or xRn,oddx\in R_{n,{\rm odd}}), then let g=x/2g=x/2 (or g=(1+x)/2g=(1+x)/2) and we have xg=0x\cdot g=0 (or xg=1x\cdot g=1). Conversely, if xg=0x\cdot g=0 (or xg=1x\cdot g=1) for some gF(Rn)g\in F(R_{n}), then xRn,evenx\in R_{n,{\rm even}} (or xRn,oddx\in R_{n,{\rm odd}}).

Thus we have the assertion with k=0k=0.

Suppose 0<km0<k\leq m, dn(x,y)=kd_{n}(x,y)=k and xRn,evenx\in R_{n,{\rm even}} (or xRn,oddx\in R_{n,{\rm odd}}). Let g1F(Rn)g_{1}\in F(R_{n}) be an element with xg1=0x\cdot g_{1}=0 (or xg1=1x\cdot g_{1}=1). By (29), we have yg1=ky\cdot g_{1}=k or k-k (or yg1=1+ky\cdot g_{1}=1+k or 1k1-k). If yg1=ky\cdot g_{1}=k (or yg1=1+ky\cdot g_{1}=1+k), then g=g1g=g_{1} is a desired one. If yg1=ky\cdot g_{1}=-k (or yg1=1ky\cdot g_{1}=1-k), then g=g10g=g_{1}0 (or g=g11g=g_{1}1) is a desired one. ∎

Proof of Theorem 5.3.  Since

(31) E(k)0={(x,y)Rn×Rndn(x,y)=kandxRn,even}andE(k)1={(x,y)Rn×Rndn(x,y)=kandxRn,odd},\begin{array}[]{ll}E(k)_{0}&=\{(x,y)\in R_{n}\times R_{n}\mid d_{n}(x,y)=k~{}\mbox{and}~{}x\in R_{n,{\rm even}}\}~{}\mbox{and}\\ E(k)_{1}&=\{(x,y)\in R_{n}\times R_{n}\mid d_{n}(x,y)=k~{}\mbox{and}~{}x\in R_{n,{\rm odd}}\},\end{array}

we see that E(k)0,E(k)1E(k)_{0},E(k)_{1} for k=0,1,,mk=0,1,\cdots,m cover Rn×RnR_{n}\times R_{n} and E(k)iE(k)i=E(k)_{i}\cap E(k^{\prime})_{i^{\prime}}=\emptyset unless k=kk=k^{\prime} and i=ii=i^{\prime}. By Lemma 5.5, E(k)0[0,k]E(k)_{0}\subset[0,k] and E(k)1[1,1+k]E(k)_{1}\subset[1,1+k]. Conversely, by (29) and by the fact that Rn,evenR_{n,{\rm even}} and Rn,oddR_{n,{\rm odd}} are the connected components of RnR_{n}, we see that [0,k]E(k)0[0,k]\subset E(k)_{0} and [1,1+k]E(k)1[1,1+k]\subset E(k)_{1}. ∎

6. Tensor products of symmetric doubles of dihedral quandles

In this section we discuss tensor products of symmetric doubles of dihedral quandles, and their quotient sets by involutions.

Let RnR_{n} be the dihedral quandle of order nn and D(Rn)D(R_{n}) its symmetric double. Note that D(Rn)={i+,iiRn}D(R_{n})=\{i^{+},i^{-}\mid i\in R_{n}\}. We use the notation in Theorem 3.4.

Theorem 6.1.

Let Rn=R2m+1R_{n}=R_{2m+1} be the dihedral quandle of order n=2m+1n=2m+1. The tensor product D(Rn)D(Rn)D(R_{n})\otimes D(R_{n}) consists of 4(m+1)4(m+1) (=2n+2=2n+2) elements, E(k)ϵ,δE(k)^{\epsilon,\delta} for k=0,1,,mk=0,1,\dots,m and ϵ,δ{+,}\epsilon,\delta\in\{+,-\}, where

(32) E(k)ϵ,δ={(i,i+k)ϵ,δ,(i,ik)ϵ,δiRn} for k=0,1,,m.
E(k)^{\epsilon,\delta}=\{(i,i+k)^{\epsilon,\delta},(i,i-k)^{\epsilon,\delta}\mid i\in R_{n}\}\quad\mbox{ for $k=0,1,\dots,m$.}\\

Proof.  It is a consequence of Theorems 3.4 and 5.1. ∎

It is obvious from the definition of E(k)ϵ,δE(k)^{\epsilon,\delta} that

(33) τ(E(k)+,+)=E(k)+,+,τ(E(k),)=E(k),,τ(E(k)+,)=E(k),+,ρ(E(k)+,+)=E(k),,ρ(E(k)+,)=E(k),+,\begin{array}[]{l}\tau(E(k)^{+,+})=E(k)^{+,+},\quad\tau(E(k)^{-,-})=E(k)^{-,-},\\ \tau(E(k)^{+,-})=E(k)^{-,+},\\ \rho(E(k)^{+,+})=E(k)^{-,-},\quad\rho(E(k)^{+,-})=E(k)^{-,+},\end{array}

for k=0,1,,mk=0,1,\dots,m.

Thus we have the following.

Corollary 6.2.

Let Rn=R2m+1R_{n}=R_{2m+1} be the dihedral quandle of order n=2m+1n=2m+1. The quotient set D(Rn)D(Rn)/τD(R_{n})\otimes D(R_{n})/\langle\tau\rangle consists of 3(m+1)3(m+1) elements,

(34) {E(k)+,+},{E(k),},{E(k)+,,E(k),+}for k=0,1,,m.\{E(k)^{+,+}\},\{E(k)^{-,-}\},\{E(k)^{+,-},E(k)^{-,+}\}\quad\mbox{for $k=0,1,\dots,m$}.
Corollary 6.3.

Let Rn=R2m+1R_{n}=R_{2m+1} be the dihedral quandle of order n=2m+1n=2m+1. The quotient set D(Rn)D(Rn)/ρD(R_{n})\otimes D(R_{n})/\langle\rho\rangle consists of 2(m+1)2(m+1) (=n+1=n+1) elements,

(35) {E(k)+,+,E(k),},{E(k)+,,E(k),+}for k=0,1,,m.\{E(k)^{+,+},E(k)^{-,-}\},\{E(k)^{+,-},E(k)^{-,+}\}\quad\mbox{for $k=0,1,\dots,m$}.
Corollary 6.4.

Let Rn=R2m+1R_{n}=R_{2m+1} be the dihedral quandle of order n=2m+1n=2m+1. The quotient set D(Rn)D(Rn)/τ,ρD(R_{n})\otimes D(R_{n})/\langle\tau,\rho\rangle consists of 2(m+1)2(m+1) (=n+1=n+1) elements,

(36) {E(k)+,+,E(k),},{E(k)+,,E(k),+}for k=0,1,,m.\{E(k)^{+,+},E(k)^{-,-}\},\{E(k)^{+,-},E(k)^{-,+}\}\quad\mbox{for $k=0,1,\dots,m$}.

When n=3n=3 and m=1m=1, E(k)ϵ,δ=[0,k]ϵ,δE(k)^{\epsilon,\delta}=[0,k]^{\epsilon,\delta} for k=0,1k=0,1 and ϵ,δ{+,}\epsilon,\delta\in\{+,-\} in Examples 3.7 and  4.2.

Theorem 6.5.

Let Rn=R2mR_{n}=R_{2m} be the dihedral quandle of order n=2mn=2m. The tensor product D(Rn)D(Rn)D(R_{n})\otimes D(R_{n}) consists of 8(m+1)8(m+1) (=4n+8=4n+8) elements, E(k)0ϵ,δE(k)_{0}^{\epsilon,\delta} and E(k)1ϵ,δE(k)_{1}^{\epsilon,\delta} for k=0,1,,mk=0,1,\dots,m and ϵ,δ{+,}\epsilon,\delta\in\{+,-\}, where

(37) E(k)0ϵ,δ={(i,i+k)ϵ,δ,(i,ik)ϵ,δiRn,even},E(k)1ϵ,δ={(i,i+k)ϵ,δ,(i,ik)ϵ,δiRn,odd}\begin{array}[]{ccl}E(k)_{0}^{\epsilon,\delta}&=&\{(i,i+k)^{\epsilon,\delta},(i,i-k)^{\epsilon,\delta}\mid i\in R_{n,{\rm even}}\},\\ E(k)_{1}^{\epsilon,\delta}&=&\{(i,i+k)^{\epsilon,\delta},(i,i-k)^{\epsilon,\delta}\mid i\in R_{n,{\rm odd}}\}\\ \end{array}

for k=0,1,,mk=0,1,\dots,m.

Proof.  It is a consequence of Theorems 3.4 and 5.3. ∎

The following corollaries are easily verified and we omit the proofs.

Corollary 6.6.

Let Rn=R2mR_{n}=R_{2m} be the dihedral quandle of order n=2mn=2m. The quotient set D(Rn)D(Rn)/τD(R_{n})\otimes D(R_{n})/\langle\tau\rangle consists of 5(m+1)5(m+1) elements (if mm is odd) or 5m+65m+6 elements (if mm is even), which are listed below:

(38) {E(k)0+,+},{E(k)0,},{E(k)0+,,E(k)0,+}{E(k)1+,+},{E(k)1,},{E(k)1+,,E(k)1,+}\begin{array}[]{lll}\{E(k)_{0}^{+,+}\},&\{E(k)_{0}^{-,-}\},&\{E(k)_{0}^{+,-},E(k)_{0}^{-,+}\}\\ \{E(k)_{1}^{+,+}\},&\{E(k)_{1}^{-,-}\},&\{E(k)_{1}^{+,-},E(k)_{1}^{-,+}\}\\ \end{array}

for k=0,2,4,mk=0,2,4,\dots\leq m, and

(39) {E(k)0+,+,E(k)1+,+},{E(k)0,,E(k)1,},{E(k)0+,,E(k)1,+},{E(k)1+,,E(k)0,+}\begin{array}[]{ll}\{E(k)_{0}^{+,+},E(k)_{1}^{+,+}\},&\{E(k)_{0}^{-,-},E(k)_{1}^{-,-}\},\\ \{E(k)_{0}^{+,-},E(k)_{1}^{-,+}\},&\{E(k)_{1}^{+,-},E(k)_{0}^{-,+}\}\\ \end{array}

for k=1,3,5,mk=1,3,5,\dots\leq m.

Corollary 6.7.

Let Rn=R2mR_{n}=R_{2m} be the dihedral quandle of order n=2mn=2m. The quotient set D(Rn)D(Rn)/ρD(R_{n})\otimes D(R_{n})/\langle\rho\rangle consists of 4(m+1)4(m+1) (=2n+4=2n+4) elements, which are listed below:

(40) {E(k)0+,+,E(k)0,},{E(k)0+,,E(k)0,+},{E(k)1+,+,E(k)1,},{E(k)1+,,E(k)1,+}\begin{array}[]{ll}\{E(k)_{0}^{+,+},E(k)_{0}^{-,-}\},&\{E(k)_{0}^{+,-},E(k)_{0}^{-,+}\},\\ \{E(k)_{1}^{+,+},E(k)_{1}^{-,-}\},&\{E(k)_{1}^{+,-},E(k)_{1}^{-,+}\}\\ \end{array}

for k=0,1,,mk=0,1,\dots,m.

Corollary 6.8.

Let Rn=R2mR_{n}=R_{2m} be the dihedral quandle of order n=2mn=2m. The quotient set D(Rn)D(Rn)/τ,ρD(R_{n})\otimes D(R_{n})/\langle\tau,\rho\rangle consists of 3(m+1)3(m+1) elements (if mm is odd) or 3m+43m+4 elements (if mm is even), which are listed below:

(41) {E(k)0+,+,E(k)0,},{E(k)0+,,E(k)0,+},{E(k)1+,+,E(k)1,},{E(k)1+,,E(k)1,+}\begin{array}[]{ll}\{E(k)_{0}^{+,+},E(k)_{0}^{-,-}\},&\{E(k)_{0}^{+,-},E(k)_{0}^{-,+}\},\\ \{E(k)_{1}^{+,+},E(k)_{1}^{-,-}\},&\{E(k)_{1}^{+,-},E(k)_{1}^{-,+}\}\\ \end{array}

for k=0,2,4,mk=0,2,4,\dots\leq m, and

(42) {E(k)0+,+,E(k)0,,E(k)1+,+,E(k)1,},{E(k)0+,,E(k)0,+,E(k)1+,,E(k)1,+}\begin{array}[]{l}\{E(k)_{0}^{+,+},E(k)_{0}^{-,-},E(k)_{1}^{+,+},E(k)_{1}^{-,-}\},\\ \{E(k)_{0}^{+,-},E(k)_{0}^{-,+},E(k)_{1}^{+,-},E(k)_{1}^{-,+}\}\\ \end{array}

for k=1,3,5,mk=1,3,5,\dots\leq m.

7. 11-handles and chords attached to surface-links

Let FF be a surface-link. A 11-handle attached to FF means an embedding h:[0,1]×B24h:[0,1]\times B^{2}\to\mathbb{R}^{4} with Fh([0,1]×B2)=h({0,1}×B2)F\cap h([0,1]\times B^{2})=h(\{0,1\}\times B^{2}). The restriction of hh to [0,1]×{0}[0,1]\times\{0\} (=[0,1])(=[0,1]) is denoted by ah:[0,1]4a_{h}:[0,1]\to\mathbb{R}^{4} and called the core map.

Definition 7.1.

Two 11-handles hh and hh^{\prime} attached to FF are strongly equivalent (or weakly equivalent, resp.) if there is an ambient isotopy of 4\mathbb{R}^{4} carrying hh to hh^{\prime} as a map (or carrying the image h([0,1]×B2)h([0,1]\times B^{2}) to h([0,1]×B2)h^{\prime}([0,1]\times B^{2}) as a subset of 4\mathbb{R}^{4}, resp.) and keeping FF setwise fixed.

We denote by s(F)\mathcal{H}^{\rm s}(F) (or w(F)\mathcal{H}^{\rm w}(F)) the set of strong (or weak) equivalence classes of 11-handles attached to FF.

Let FF be a surface-link in 4\mathbb{R}^{4}.

  • A chord attached to FF is a path a:[0,1]4a:[0,1]\to\mathbb{R}^{4} such that a(t)Fa(t)\in F if and only if t{0,1}t\in\{0,1\}.

  • An initial root (or a terminal root) of aa is a regular neighborhood in FF of the initial point a(0)a(0) (or the terminal point a(1)a(1)).

  • An initial root orientation (or a terminal root orientation) of aa is an orientation of an initial root (or a terminal root). A root orientaion of aa means a pair of an initial root orientation and a terminal root orientaion of aa.

  • A root-oriented chord is a chord equipped with a root orientation or a pair (a,o)(a,o) of a chord aa and a root orientation oo of aa.

When a chord aa is given, there are 44 possibilities for root orientations (Figure 3).

aaaaaaaa(1)(1)(2)(2)(3)(3)(4)(4)

Figure 3. There are 44 possible root orientations for a given chord aa.
Definition 7.2.

Two chords attached to FF are strongly equivalent or homotopic if they are homotopic as chords attached to FF. Two root-oriented chords attached to FF are strongly equivalent or homotopic if they are homotopic as root-oriented chords attached to FF.

For a chord a:[0,1]4a:[0,1]\to\mathbb{R}^{4} attached to FF, we denote by rev(a):[0,1]4{\rm rev}(a):[0,1]\to\mathbb{R}^{4} the chord defined by rev(a)(t)=a(1t){\rm rev}(a)(t)=a(1-t). For a root orientation oo of a chord aa, we denote by o-o the reversed orientation of oo.

Definition 7.3.

Two chords aa and aa^{\prime} attached to FF are weakly equivalent if aa is homotopic to aa^{\prime} or rev(a){\rm rev}(a^{\prime}). Two root-oriented chords (a,o)(a,o) and (a,o)(a^{\prime},o^{\prime}) attached to FF are weakly equivalent if (a,o)(a,o) is homotopic to (a,o)(a^{\prime},o^{\prime}), (a,o)(a^{\prime},-o^{\prime}), (rev(a),o)({\rm rev}(a^{\prime}),o^{\prime}), or (rev(a),o)({\rm rev}(a^{\prime}),-o^{\prime}).

Let 𝒞s(F)\mathcal{C}^{\rm s}(F) (or 𝒞w(F)\mathcal{C}^{\rm w}(F)) denote the set of strong (or weak) equivalence classes of chords attached to FF, and let 𝒞roots(F)\mathcal{C}_{\rm root}^{\rm s}(F) (or 𝒞rootw(F)\mathcal{C}_{\rm root}^{\rm w}(F)) denote the set of strong (or weak) equivalence classes of root-oriented chords attached to FF.

For a 11-handle hh attached to a surface-link FF, the core map aha_{h} is a chord attached to FF. If hh and hh^{\prime} are strongly (or weakly) equivalent, then aha_{h} and aha_{h^{\prime}} are strongly (or weakly) equivalent. Thus we have maps

(43) ϕ0:s(F)𝒞s(F)andψ0:w(F)𝒞w(F)\phi_{0}:\mathcal{H}^{\rm s}(F)\to\mathcal{C}^{\rm s}(F)\quad\mbox{and}\quad\psi_{0}:\mathcal{H}^{\rm w}(F)\to\mathcal{C}^{\rm w}(F)

sending [h][h] to [ah][a_{h}].

Let FF be a surface-link, hh a 11-handle attached to FF and aha_{h} the core map of hh. Let oho_{h} be the root orientation of aha_{h} which is the reverse of the orientation induced from that of ([0,1]×B2)\partial([0,1]\times B^{2}) by hh. We call (ah,oh)(a_{h},o_{h}) the root-oriented chord associated with hh.

The following theorem is given in [10].

Theorem 7.4 ([10]).

Two 11-handles hh and hh^{\prime} are strongly (or weakly) equivalent if and only if the root-oriented chords (ah,oh)(a_{h},o_{h}) and (ah,oh)(a_{h^{\prime}},o_{h^{\prime}}) associated with hh and hh^{\prime} are strongly (or weakly) equivalent. In other words, the maps

(44) ϕ:s(F)𝒞roots(F)andψ:w(F)𝒞rootw(F)\phi:\mathcal{H}^{\rm s}(F)\to\mathcal{C}_{\rm root}^{\rm s}(F)\quad\mbox{and}\quad\psi:\mathcal{H}^{\rm w}(F)\to\mathcal{C}_{\rm root}^{\rm w}(F)

sending [h][h] to [(ah,oh)][(a_{h},o_{h})] are bijections.

In the rest of this section, let FF be an oriented surface-link, and we put a focus on oriented 11-handles attached to FF.

Definition 7.5.

A 11-handle hh attached to FF is oriented if the orientation of h({0,1}×B2)h(\{0,1\}\times B^{2}) induced from the orientation of ([0,1]×B2)\partial([0,1]\times B^{2}) is opposite to that induced from the orientation of FF.

In other words, when (ah,oh)(a_{h},o_{h}) is the root-oriented chord associated with hh, a 11-handle hh attached to FF is oriented if and only if the root orientation oho_{h} matches the orientation of FF.

Let s(F)ori\mathcal{H}^{\rm s}(F)^{\rm ori} (or w(F)ori\mathcal{H}^{\rm w}(F)^{\rm ori}) denote the subset of s(F)\mathcal{H}^{\rm s}(F) (or w(F)\mathcal{H}^{\rm w}(F)) consisting of the strong (or weak) equivalence classes of oriented 11-handles attached to FF.

The following is observed in [1, 5], cf. [10].

Theorem 7.6 ([1, 5], cf. [10]).

Let FF be an oriented surface-link. Two oriented 11-handles hh and hh^{\prime} attached to FF are strongly (or weakly) equivalent if and only if their chords aha_{h} and aha_{h^{\prime}} are strongly (or weakly) equivalence. In other words, the restriction maps of ϕ0\phi_{0} and ψ0\psi_{0} to s(F)ori\mathcal{H}^{\rm s}(F)^{\rm ori} and w(F)ori\mathcal{H}^{\rm w}(F)^{\rm ori},

(45) ϕ0ori:s(F)ori𝒞s(F)andψ0ori:w(F)ori𝒞w(F),\phi_{0}^{\rm ori}:\mathcal{H}^{\rm s}(F)^{\rm ori}\to\mathcal{C}^{\rm s}(F)\quad\mbox{and}\quad\psi_{0}^{\rm ori}:\mathcal{H}^{\rm w}(F)^{\rm ori}\to\mathcal{C}^{\rm w}(F),

sending [h][h] to [ah][a_{h}], are bijections.

Remark 7.7.

For an oriented surface-link FF, let 𝒞roots(F)ori\mathcal{C}_{\rm root}^{\rm s}(F)^{\rm ori} (or 𝒞rootw(F)ori\mathcal{C}_{\rm root}^{\rm w}(F)^{\rm ori}) denote the subset of 𝒞roots(F)\mathcal{C}_{\rm root}^{\rm s}(F) (or 𝒞rootw(F)\mathcal{C}_{\rm root}^{\rm w}(F)) such that the root orientations match the orientation of FF. There are natural bijections

(46) 𝒞s(F)𝒞roots(F)oriand𝒞w(F)𝒞rootw(F)ori\mathcal{C}^{\rm s}(F)\to\mathcal{C}_{\rm root}^{\rm s}(F)^{\rm ori}\quad\mbox{and}\quad\mathcal{C}^{\rm w}(F)\to\mathcal{C}_{\rm root}^{\rm w}(F)^{\rm ori}

sending [a][a] to [(a,oF)][(a,o_{F})] where oFo_{F} is the root orientation of aa which matches the orientation of FF. Then the bijections ϕ0ori\phi_{0}^{\rm ori} and ψ0ori\psi_{0}^{\rm ori} are regarded as the bijections

(47) ϕori:s(F)ori𝒞roots(F)oriandψori:w(F)ori𝒞rootw(F)ori,\phi^{\rm ori}:\mathcal{H}^{\rm s}(F)^{\rm ori}\to\mathcal{C}_{\rm root}^{\rm s}(F)^{\rm ori}\quad\mbox{and}\quad\psi^{\rm ori}:\mathcal{H}^{\rm w}(F)^{\rm ori}\to\mathcal{C}_{\rm root}^{\rm w}(F)^{\rm ori},

which are restrictions of the bijections ϕ\phi and ψ\psi in Theorem 7.4.

8. 11-handles and chords up to strong equivalence

In this section, we consider 11-handles and chords up to strong equivalence. For a surface-link FF, let Q~(F)\widetilde{Q}(F) be the full knot quandle of FF.

Let (D,a)(D,a) be a pair of an oriented meridian disk DD of FF and a path a:[0,1]4a:[0,1]\to\mathbb{R}^{4} in the knot exterior connecting DD and the base point, say qq. Let a~:[0,1]4\tilde{a}:[0,1]\to\mathbb{R}^{4} be a path which is the concatenation of the path aa and a path in DD starting at the center of DD and terminating at a(0)a(0). Then a~:[0,1]4\tilde{a}:[0,1]\to\mathbb{R}^{4} is a path in 4\mathbb{R}^{4} such that a~(t)F\tilde{a}(t)\in F if and only if t=0t=0 and a~(1)=q\tilde{a}(1)=q. We call a~\tilde{a} an extended path of aa. The initial point of a~\tilde{a} is a point of FF. Using the orientation of DD and the standard orientation of 4\mathbb{R}^{4}, we can give an orientation of a regular neighborhood in FF of the initial point of a~\tilde{a}, which we call the root orientation determined from the orientation of DD and we denote it by o(D,a)o_{(D,a)}.

We define a map μ:Q~(F)Q~(F)𝒞roots(F)\mu:\widetilde{Q}(F)\otimes\widetilde{Q}(F)\to\mathcal{C}_{\rm root}^{\rm s}(F).

Let x1=[(D1,a1)]x_{1}=[(D_{1},a_{1})] and x2=[(D2,a2)]x_{2}=[(D_{2},a_{2})] be elements of Q~(F)\widetilde{Q}(F). Let a1~\tilde{a_{1}} and a2~\tilde{a_{2}} be extended paths of a1a_{1} and a2a_{2}, and o(D1,a1)o_{(D_{1},a_{1})} and o(D2,a2)o_{(D_{2},a_{2})} be the root orientations determined from the orientations of D1D_{1} and D2D_{2}. The concatenation of a1~\tilde{a_{1}} and a2~1\tilde{a_{2}}^{-1} is a chord attached to FF. We denote it by a(x1,x2)a(x_{1},x_{2}) and call it the chord associated with (x1,x2)(x_{1},x_{2}). We choose a root orientation of a(x1,x2)a(x_{1},x_{2}) such that the initial root orientation is o(D1,a1)o_{(D_{1},a_{1})} and the terminal root orientation is o(D2,a2)o_{(D_{2},a_{2})}. Then the strong equivalence class of the root-oriented chord is uniquely determined from x1=[(D1,a1)]x_{1}=[(D_{1},a_{1})] and x2=[(D2,a2)]x_{2}=[(D_{2},a_{2})]. We denote this root-oriented chord by (a(x1,x2),o(x1,x2))(a(x_{1},x_{2}),o(x_{1},x_{2})) and call it the root-oriented chord associated with (x1,x2)(x_{1},x_{2}).

Theorem 8.1.

Let

(48) Q~(F)×Q~(F)𝒞roots(F)\widetilde{Q}(F)\times\widetilde{Q}(F)\to\mathcal{C}_{\rm root}^{\rm s}(F)

be the map sending (x1,x2)Q~(F)×Q~(F)(x_{1},x_{2})\in\widetilde{Q}(F)\times\widetilde{Q}(F) to the strong equivalence class of (a(x1,x2),o(x1,x2))(a(x_{1},x_{2}),o(x_{1},x_{2})). It induces a well-defined and bijective map

(49) μ:Q~(F)Q~(F)𝒞roots(F).\mu:\widetilde{Q}(F)\otimes\widetilde{Q}(F)\to\mathcal{C}_{\rm root}^{\rm s}(F).

Proof.  Consider (x1g,x2g)(x_{1}\cdot g,x_{2}\cdot g) for gF(Q~(F))g\in F(\widetilde{Q}(F)). Then x1g=[(D1,a1γ)]x_{1}\cdot g=[(D_{1},a_{1}\gamma)] and x2g=[(D2,a2γ)]x_{2}\cdot g=[(D_{2},a_{2}\gamma)] for some loop γ\gamma in 4\mathbb{R}^{4} such that γ\gamma misses FF and γ(0)=γ(1)=q\gamma(0)=\gamma(1)=q. (This is seen from the definition of the operations \ast and ¯\overline{\ast} for the knot symmetric quandle Q~(F)\widetilde{Q}(F). See Figure 1.) The chord associated with (x1g,x2g)(x_{1}\cdot g,x_{2}\cdot g) is a1~γγ1a2~1\tilde{a_{1}}\gamma\gamma^{-1}\tilde{a_{2}}^{-1} which is homotopic to a1~a2~1\tilde{a_{1}}\tilde{a_{2}}^{-1} without changing D1D_{1} and D2D_{2}. Thus (a(x1g,x2g),o(x1g,x2g))(a(x_{1}\cdot g,x_{2}\cdot g),o(x_{1}\cdot g,x_{2}\cdot g)) is strongly equivalent to (a(x1,x2),o(x1,x2))(a(x_{1},x_{2}),o(x_{1},x_{2})). Hence, μ\mu is well-defined.

Let (a,o)(a,o) be a root-oriented chord attached to FF. Let D1D_{1} (or D2D_{2}) be an oriented meridian disk of FF over the initial point a(0)a(0) (or the terminal point a(1)a(1)) such that the orientation matches the root orientation oo. Moving aa up to homotopy, we change aa such that a(1/2)=qa(1/2)=q. Then aa is the associated chord with (x1,x2)(x_{1},x_{2}) for some (x1,x2)Q~(F)×Q~(F)(x_{1},x_{2})\in\widetilde{Q}(F)\times\widetilde{Q}(F). Thus, μ\mu is surjective.

Suppose that [(a(x1,x2),o(x1,x2))]=[(a(x1,x2),o(x1,x2))][(a(x_{1},x_{2}),o(x_{1},x_{2}))]=[(a(x^{\prime}_{1},x^{\prime}_{2}),o(x^{\prime}_{1},x^{\prime}_{2}))]. Let αs:[0,1]4\alpha_{s}:[0,1]\to\mathbb{R}^{4} (s[0,1])(s\in[0,1]) be a homotopy of chords attached to FF such that α0=a(x1,x2)\alpha_{0}=a(x_{1},x_{2}) and α1=a(x1,x2)\alpha_{1}=a(x^{\prime}_{1},x^{\prime}_{2}). Let β:[0,1]4\beta:[0,1]\to\mathbb{R}^{4} be the path defined by β(s)=αs(1/2)\beta(s)=\alpha_{s}(1/2) for s[0,1]s\in[0,1]. Note that β\beta misses FF and β(0)=β(1)=q\beta(0)=\beta(1)=q. Then a1~β\tilde{a_{1}}\beta is homotopic to a1~\tilde{a^{\prime}_{1}} through paths in 4\mathbb{R}^{4} such that the paths intersect with FF at their initial points only and the teminal points are qq. This implies that x1g=x1Q~(F)x_{1}\cdot g=x^{\prime}_{1}\in\widetilde{Q}(F) where gg is an element of F(Q~(F))F(\widetilde{Q}(F)) which presents [β]π1(4F,q)=As(Q~(F),ρ)[\beta]\in\pi_{1}(\mathbb{R}^{4}\setminus F,q)=As(\widetilde{Q}(F),\rho). (Note that there is a natural projection F(Q~(F))As(Q~(F),ρ)F(\widetilde{Q}(F))\to As(\widetilde{Q}(F),\rho) and As(Q~(F),ρ)As(\widetilde{Q}(F),\rho) is identified with G(F)=π1(4F,q)G(F)=\pi_{1}(\mathbb{R}^{4}\setminus F,q) in the sense of Theorem 4.5 of [9] (Remark 2.8), which is analogous to the fact given in [3] that As(Q(F))As(Q(F)) is identified with G(F)G(F).) Similarly, we have x2g=x2x_{2}\cdot g=x^{\prime}_{2} for the same gg. Thus [(x1,x2)]=[(x1,x2)][(x_{1},x_{2})]=[(x^{\prime}_{1},x^{\prime}_{2})] in Q~(F)Q~(F)\widetilde{Q}(F)\otimes\widetilde{Q}(F). ∎

Combining the bijection ϕ:s(F)𝒞roots(F)\phi:\mathcal{H}^{\rm s}(F)\to\mathcal{C}_{\rm root}^{\rm s}(F) in Theorem 7.4 and the inverse map of μ\mu in Theorem 8.1, we obtain a bijection as follows.

Theorem 8.2.

Let FF be a surface-link. The map

(50) Φ=μ1ϕ:s(F)Q~(F)Q~(F)\Phi=\mu^{-1}\circ\phi:\mathcal{H}^{\rm s}(F)\to\widetilde{Q}(F)\otimes\widetilde{Q}(F)

is a bijection.

Proof.  It follows from Theorems 7.4 and 8.1. ∎

This theorem gives the former assertion of Theorem 1.1.

We consider the case where FF is an oriented surface-link.

Let x1=[(D1,a1)]x_{1}=[(D_{1},a_{1})] and x2=[(D2,a2)]x_{2}=[(D_{2},a_{2})] be elements of Q(F)Q(F). Let a1~\tilde{a_{1}} and a2~\tilde{a_{2}} be extended paths of a1a_{1} and a2a_{2}. The concatenation of a1~\tilde{a_{1}} and a2~1\tilde{a_{2}}^{-1} is a chord attached to FF. We denote it by a(x1,x2)a(x_{1},x_{2}) and call it the chord associated with (x1,x2)(x_{1},x_{2}).

Theorem 8.3.

Let

(51) Q(F)×Q(F)𝒞s(F).Q(F)\times Q(F)\to\mathcal{C}^{\rm s}(F).

be the map sending (x1,x2)(x_{1},x_{2}) to the strong equivalence class of a(x1,x2)a(x_{1},x_{2}). It induces a well-defined and bijective map

(52) μ0ori:Q(F)Q(F)𝒞s(F).\mu_{0}^{\rm ori}:Q(F)\otimes Q(F)\to\mathcal{C}^{\rm s}(F).

Proof.  It is proved by a similar argument to the proof of Theorem 8.1 by forgetting the root orientations. ∎

Combining the bijection ϕ0ori:s(F)ori𝒞s(F)\phi_{0}^{\rm ori}:\mathcal{H}^{\rm s}(F)^{\rm ori}\to\mathcal{C}^{\rm s}(F) in Theorem 7.6 and the inverse map of μ0ori\mu_{0}^{\rm ori}, we have a bijection as follows.

Theorem 8.4.

Let FF be an oriented surface-link. The map

(53) Φ0ori=(μ0ori)1ϕ0ori:s(F)oriQ(F)Q(F)\Phi_{0}^{\rm ori}=(\mu_{0}^{\rm ori})^{-1}\circ\phi_{0}^{\rm ori}:\mathcal{H}^{\rm s}(F)^{\rm ori}\to Q(F)\otimes Q(F)

is a bijection.

Proof.  It follows from Theorems 7.6 and 8.3. ∎

This theorem gives the former assertion of Theorem 1.2.

Remark 8.5.

Let FF be an oriented surface-link. There is a commutative diagram as follows:

(54) s(F)oriϕ0ori𝒞s(F)(μ0ori)1Q(F)Q(F) incl.     incl.  s(F)ϕ𝒞roots(F)μ1Q~(F)Q~(F),\begin{array}[]{ccccc}\mathcal{H}^{\rm s}(F)^{\rm ori}&\smash{\mathop{\hbox to36.98866pt{\rightarrowfill}}\limits^{\phi_{0}^{\rm ori}}}&\mathcal{C}^{\rm s}(F)&\smash{\mathop{\hbox to36.98866pt{\rightarrowfill}}\limits^{(\mu_{0}^{\rm ori})^{-1}}}&Q(F)\otimes Q(F)\\ \Big{\downarrow}\hbox to0.0pt{\hss$\vbox{\hbox{$\scriptstyle\rm incl.\,$}}$ }&&\Big{\downarrow}\hbox to0.0pt{\hss$\vbox{\hbox{$\scriptstyle\,$}}$ }&&\Big{\downarrow}\hbox to0.0pt{$\vbox{\hbox{$\scriptstyle\rm incl.\,$}}$ \hss}\\ \mathcal{H}^{\rm s}(F)&\smash{\mathop{\hbox to36.98866pt{\rightarrowfill}}\limits^{\phi}}&\mathcal{C}_{\rm root}^{\rm s}(F)&\smash{\mathop{\hbox to36.98866pt{\rightarrowfill}}\limits^{\mu^{-1}}}&\widetilde{Q}(F)\otimes\widetilde{Q}(F),\end{array}

where the first and the third vertical maps are inclusion maps, and the second vertical map 𝒞s(F)𝒞roots(F)\mathcal{C}^{\rm s}(F)\to\mathcal{C}_{\rm root}^{\rm s}(F) is an injective map sending [a][a] to [(a,oF)][(a,o_{F})] where oFo_{F} is the root orientation of aa which matches the orientation of FF. All horizontal maps are bijections.

9. 11-handles and chords up to weak equivalence

In this section, we consider 11-handles and chords up to weak equivalence.

Let FF be a surface-link and let (Q~(F),ρ)(\widetilde{Q}(F),\rho) be the knot symmetric quandle. Let τ\tau and ρ\rho be involutions of Q~(F)Q~(F)\widetilde{Q}(F)\otimes\widetilde{Q}(F) with τ[x1,x2]=[x2,x1]\tau[x_{1},x_{2}]=[x_{2},x_{1}] and ρ[x1,x2]=[ρ(x1),ρ(x2)]\rho[x_{1},x_{2}]=[\rho(x_{1}),\rho(x_{2})].

Theorem 9.1.

The bijection μ:Q~(F)Q~(F)𝒞roots(F)\mu:\widetilde{Q}(F)\otimes\widetilde{Q}(F)\to\mathcal{C}_{\rm root}^{\rm s}(F) in Theorem 8.1 induces a bijection

(55) μτ,ρ:Q~(F)Q~(F)/τ,ρ𝒞rootw(F).\mu^{\langle\tau,\rho\rangle}:\widetilde{Q}(F)\otimes\widetilde{Q}(F)/\langle\tau,\rho\rangle\to\mathcal{C}_{\rm root}^{\rm w}(F).

Proof.  Let rev1{\rm rev}_{1} and rev2{\rm rev}_{2} be involutions of 𝒞roots(F)\mathcal{C}_{\rm root}^{\rm s}(F) with rev1([(a,o)])=[(rev(a),o)]{\rm rev}_{1}([(a,o)])=[({\rm rev}(a),o)] and rev2([(a,o)])=[(a,o)]{\rm rev}_{2}([(a,o)])=[(a,-o)]. Note that the quotient of 𝒞roots(F)\mathcal{C}_{\rm root}^{\rm s}(F) by the equivalence relation generated by rev1{\rm rev}_{1} and rev2{\rm rev}_{2} is 𝒞rootw(F)\mathcal{C}_{\rm root}^{\rm w}(F). Since μτ=rev1μ\mu\circ\tau={\rm rev}_{1}\circ\mu and μρ=rev2μ\mu\circ\rho={\rm rev}_{2}\circ\mu, we have the result. ∎

Combining the bijection ψ:w(F)𝒞rootw(F)\psi:\mathcal{H}^{\rm w}(F)\to\mathcal{C}_{\rm root}^{\rm w}(F) in Theorem 7.4 and the inverse map of μτ,ρ\mu^{\langle\tau,\rho\rangle}, we have a bijection as follows.

Theorem 9.2.

Let FF be a surface-link. The map

(56) Ψ=(μτ,ρ)1ψ:w(F)Q~(F)Q~(F)/τ,ρ\Psi=(\mu^{\langle\tau,\rho\rangle})^{-1}\circ\psi:\mathcal{H}^{\rm w}(F)\to\widetilde{Q}(F)\otimes\widetilde{Q}(F)/\langle\tau,\rho\rangle

is a bijection.

Proof.  It follows from Theorems 7.4 and 9.1. ∎

This theorem give the latter assertion of Theorem 1.1.

Let FF be an oriented surface-link. Let τ\tau be the involution of Q(F)Q(F)Q(F)\otimes Q(F) with τ[x1,x2]=[x2,x1]\tau[x_{1},x_{2}]=[x_{2},x_{1}].

Theorem 9.3.

The bijection μ0ori:Q(F)Q(F)𝒞s(F)\mu_{0}^{\rm ori}:Q(F)\otimes Q(F)\to\mathcal{C}^{\rm s}(F) induces a bijection

(57) μ0τ:Q(F)Q(F)/τ𝒞w(F).\mu_{0}^{\langle\tau\rangle}:Q(F)\otimes Q(F)/{\langle\tau\rangle}\to\mathcal{C}^{\rm w}(F).

Proof.  Let rev:𝒞s(F)𝒞s(F){\rm rev}:\mathcal{C}^{\rm s}(F)\to\mathcal{C}^{\rm s}(F) be the involution of Q(F)Q(F) with rev([a])=[rev(a)]{\rm rev}([a])=[{\rm rev}(a)]. The quotient of 𝒞s(F)\mathcal{C}^{\rm s}(F) by the equivalence relation generated by rev{\rm rev} is 𝒞w(F)\mathcal{C}^{\rm w}(F). Since μ0τ=revμ0\mu_{0}\circ\tau={\rm rev}\circ\mu_{0}, we have the result. ∎

Combining the bijection ψ0ori:w(F)ori𝒞w(F)\psi_{0}^{\rm ori}:\mathcal{H}^{\rm w}(F)^{\rm ori}\to\mathcal{C}^{\rm w}(F) in Theorem 7.6 and the inverse map of μ0τ\mu_{0}^{\langle\tau\rangle}, we have a bijection as follows.

Theorem 9.4.

Let FF be an oriented surface-link. The map

(58) Ψ0ori=(μ0τ)1ψ0ori:w(F)oriQ(F)Q(F)/τ\Psi_{0}^{\rm ori}=(\mu_{0}^{\langle\tau\rangle})^{-1}\circ\psi_{0}^{\rm ori}:\mathcal{H}^{\rm w}(F)^{\rm ori}\to Q(F)\otimes Q(F)/\langle\tau\rangle

is a bijection.

Proof.  It follows from Theorems 7.4 and 9.3. ∎

This theorem give the latter assertion of Theorem 1.2.

10. Examples and invariants of 11-handles

Example 10.1.

Let T(2,n)T(2,n) be a torus knot of type (2,n)(2,n) with n=2m+1n=2m+1, and let FF be the 22-twist spun T(2,n)T(2,n). The knot quandle Q(F)Q(F) has a presentation

(59) a,bb=a(ba)m,ba2=b,\langle a,\,b\mid b=a^{(ba)^{m}},\,b^{a^{2}}=b\rangle,

which is isomorphic to the dihedral quandle RnR_{n} of order nn (cf. [15]), where we follow the convention of quandle operation and presentation due to [3] (cf. [11]). The full knot quandle Q~(F)\widetilde{Q}(F), with the standard involution as a knot symmetric quandle, is isomorphic to the symmetric double of Q(F)Q(F) (Remark 2.7) and hence to the symmetric double D(Rn)D(R_{n}). Fix an isomorphism from Q(F)Q(F) to RnR_{n} and identify Q(F)Q(F) with RnR_{n}. Then

(60) Q(F)Q(F)=RnRnandQ~(F)Q~(F)=D(Rn)D(Rn).Q(F)\otimes Q(F)=R_{n}\otimes R_{n}\quad\mbox{and}\quad\widetilde{Q}(F)\otimes\widetilde{Q}(F)=D(R_{n})\otimes D(R_{n}).

By Theorems 8.4 and 8.2, we obtain bijections s(F)oriRnRn\mathcal{H}^{\rm s}(F)^{\rm ori}\to R_{n}\otimes R_{n} and s(F)D(Rn)D(Rn)\mathcal{H}^{\rm s}(F)\to D(R_{n})\otimes D(R_{n}). Since we have Theorems 5.1 and 6.1, they provide us complete classifications of oriented 11-handles and all 11-handles attached to FF up to strong equivalence. In particular, we see that #s(F)ori=m+1\#\mathcal{H}^{\rm s}(F)^{\rm ori}=m+1 and #s(F)=2n+2\#\mathcal{H}^{\rm s}(F)=2n+2.

By Theorems 9.4 and 9.2, we obtain bijections w(F)oriRnRn/τ\mathcal{H}^{\rm w}(F)^{\rm ori}\to R_{n}\otimes R_{n}/\langle\tau\rangle and w(F)D(Rn)D(Rn)/τ,ρ\mathcal{H}^{\rm w}(F)\to D(R_{n})\otimes D(R_{n})/\langle\tau,\rho\rangle. Since we have Corollaries 5.2 and 6.4, they provide us complete classifications of oriented 11-handles and all 11-handles up to weak equivalence. In particular, we have #w(F)ori=m+1\#\mathcal{H}^{\rm w}(F)^{\rm ori}=m+1 and #w(F)=n+1\#\mathcal{H}^{\rm w}(F)=n+1.

Example 10.2.

Let FF be a knotted projective plane which is the connected sum of the 22-twist spun T(2,n)T(2,n) with n=2m+1n=2m+1 and a (positive or negative) standard projective plane in 4\mathbb{R}^{4}. The full knot quandle Q~(F)\widetilde{Q}(F) is isomorphic to the dihedral quandle RnR_{n} and the standard involution ρ\rho as a knot symmetric quandle is the identity map (cf. [9, 15]). Fix an identification of Q~(F)\widetilde{Q}(F) and RnR_{n}, and we have

(61) Q~(F)Q~(F)=Rn×Rn.\widetilde{Q}(F)\otimes\widetilde{Q}(F)=R_{n}\times R_{n}.

By Theorem 8.2, we obtain a bijection s(F)RnRn\mathcal{H}^{\rm s}(F)\to R_{n}\otimes R_{n}, which provides us a complete classification of 11-handles attached to FF up to strong equivalence. In particular, we have #s(F)=m+1\#\mathcal{H}^{\rm s}(F)=m+1.

By Theorem 9.2, we obtain a bijection w(F)RnRn/τ,ρ=RnRn/τ\mathcal{H}^{\rm w}(F)\to R_{n}\otimes R_{n}/\langle\tau,\rho\rangle=R_{n}\otimes R_{n}/\langle\tau\rangle, which provides us a complete classification of 11-handles attached to FF up to weak equivalence. In particular, we have #w(F)=m+1\#\mathcal{H}^{\rm w}(F)=m+1.

For a surface-link such that Q~(F)\widetilde{Q}(F) (or Q(F)Q(F)) is infinite or has a large cardinality, it is difficult to compute or list the elements of the tensor product Q~(F)Q~(F)\widetilde{Q}(F)\otimes\widetilde{Q}(F) (or Q(F)×Q(F)Q(F)\times Q(F)) and its quotient. In such a case, it is useful to construct an invariant.

Let FF be a surface-link and let XX be a (finite) quandle. Let f:Q~(F)Xf:\widetilde{Q}(F)\to X be a quandle homomorphism. By Lemma 4.3, it induces a map

(62) ff:Q~(F)Q~(F)XX,[x1,x2][f(x1),f(x2)].f\otimes f:\widetilde{Q}(F)\otimes\widetilde{Q}(F)\to X\otimes X,\quad[x_{1},x_{2}]\mapsto[f(x_{1}),f(x_{2})].

Combining the bijection in Theorem 8.2, we have a map

(63) Is(f)=(ff)Φ:s(F)XX,[h](ff)[Φ([h])].{\rm I}^{\rm s}(f)=(f\otimes f)\circ\Phi:\mathcal{H}^{\rm s}(F)\to X\otimes X,\quad[h]\mapsto(f\otimes f)[\Phi([h])].

In this way, for each quandle homomorphism f:Q~(F)Xf:\widetilde{Q}(F)\to X, we obtain an invariant Is(f){\rm I}^{\rm s}(f) of strong equivalence classes of 11-handles attached to FF valued in XXX\otimes X.

When FF is an oriented surface-link, for any quandle homomorphism f:Q(F)Xf:Q(F)\to X, we obtain an invariant Ioris(f)=(ff)Φ0ori{\rm I}^{\rm s}_{\rm ori}(f)=(f\otimes f)\circ\Phi_{0}^{\rm ori} of strong equivalence classes of oriented 11-handles attached to FF valued in XXX\otimes X.

Let (X,ρ)(X,\rho) be a symmetric quandle. Let f:(Q~(F),ρ)(X,ρ)f:(\widetilde{Q}(F),\rho)\to(X,\rho) be a symmetric quandle homomorphism to a (finite) symmetric quandle (X,ρ)(X,\rho). By Theorem 9.2, we have a map

(64) Iw(f)=(ff)Ψ:w(F)XX/τ,ρ,[h](ff)[Ψ([h])].{\rm I}^{\rm w}(f)=(f\otimes f)\circ\Psi:\mathcal{H}^{\rm w}(F)\to X\otimes X/\langle\tau,\rho\rangle,\quad[h]\mapsto(f\otimes f)[\Psi([h])].

In this way, for each symmetric quandle homomorphism f:(Q~(F),ρ)(X,ρ)f:(\widetilde{Q}(F),\rho)\to(X,\rho), we obtain an invariant Iw(f){\rm I}^{\rm w}(f) of weak equivalence classes of 11-handles attached to FF valued in XX/τ,ρX\otimes X/\langle\tau,\rho\rangle.

When FF is an oriented surface-link, for any quandle homomorphism f:Q(F)Xf:Q(F)\to X, we obtain an invariant Ioriw(f)=(ff)Ψ0ori{\rm I}^{\rm w}_{\rm ori}(f)=(f\otimes f)\circ\Psi_{0}^{\rm ori} of weak equivalence classes of oriented 11-handles attached to FF valued in XX/τX\otimes X/\langle\tau\rangle.

Now we have Theorems 1.3 and 1.4.

The dihedral quandle RnR_{n} and the symmetric double (D(X),ρ)(D(X),\rho) can be used to construct such an invariant. Refer to [4, 13, 16] for other finite quandles with small orders. For such a small quandle, one can compute the tensor product and use it to construct an invariant of 11-handles.

References

  • [1] J. Boyle, Classifying 11-handles attached to knotted surfaces, Trans. Amer. Math. Soc. 306 (1988), 475–487.
  • [2] S. Carter, S. Kamada and M. Saito, Surfaces in 44-Space, Encyclopaedia of Mathematical Sciences, Vol. 142, Low-Dimensional Topology, III. Springer-Verlag, Berlin Heidelberg New York, 2004.
  • [3] R. Fenn and C. Rourke, Racks and links in codimension two, J. Knot Theory Ramifications 1 (1992), 343–406.
  • [4] C. Hayashi, Canonical forms for operation tables of finite connected quandles, Comm. Algebra 41 (2013), 3340–3349.
  • [5] F. Hosokawa and A. Kawauchi, Proposals for unknotted surfaces in four-spaces, Osaka J. Math. 16 (1979), 233–248.
  • [6] D. Joyce, A classification invariant of knots, the knot quandle, J. Pure Appl. Algebra 23 (1982), 37–65.
  • [7] S. Kamada, Non-orientable surfaces in 44-space, Osaka J. Math. 26 (1989), 367–385.
  • [8] S. Kamada, Quandles with good involutions, their homologies and knot invariants, in: Intelligence of Low Dimensional Topology 2006, Eds. J. S. Carter et. al., pp. 101–108, World Scientific Publishing Co., 2007.
  • [9] S. Kamada, Quandles and symmetric quandles for higher dimensional knots, Knots in Poland III – Part III, Banach Center Publications, 103 (2014), 145–158.
  • [10] S. Kamada, Cords and 1-handles attached to surface-knots, Bol. Soc. Mat. Mex., 20 (2014), no. 2, 595–609.
  • [11] S. Kamada, Surface-Knots in 4-Space, Springer Monographs in Mathematics, Springer, 2017
  • [12] S. Kamada and K. Oshiro, Homology groups of symmetric quandles and cocycle invariants of links and surface-links, Trans. Amer. Math. Soc. 362 (2010), no. 10, 5501–5527.
  • [13] P. Lopes, D. Roseman, On finite racks and quandles, Comm. Algebra 34 (2006), 371–406.
  • [14] S. Matveev, Distributive groupoids in knot theory, Mat. Sb. (N.S.) 119 (161) (1982) 78–88, 160; English translation: Math. USSR-Sb. 47 (1984), 73–83.
  • [15] S. Satoh, Surface diagrams of twist-spun 22-knots, J. Knot Theory Ramifications 11 (2002), 413–430.
  • [16] L. Vendramin, On the classification of quandles of low order, J. Knot Theory Ramifications 21 (2012), 1250088 (10 pages).