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

ON INEQUALITIES BETWEEN UNKNOTTING NUMBERS AND CROSSING NUMBERS OF SPATIAL EMBEDDINGS OF TRIVIALIZABLE GRAPHS AND HANDLEBODY-KNOTS

YUTA AKIMOTO
Abstract.

We study relations between unknotting number and crossing number of a spatial embedding of a handcuff-graph and a theta curve. It is well known that for any non-trivial knot KK twice the unknotting number of KK is less than or equal to the crossing number of KK minus one. We show that this is extended to handlebody-knots. We also characterize the handlebody-knots which satisfy the equality.

Key words and phrases:
spatial graph, handcuff-graph, theta curve, handlebody-knot, unknotting number, crossing number

1. INTRODUCTION

Let LL be a link in the 3-dimensional Euclidean space 3\mathbb{R}^{3}. The unknotting number u(L)u(L) is the minimal number of crossing changes (Fig. 1) from LL to a trivial link. The crossing number c(L)c(L) is the minimal number of crossing points among all regular diagrams of LL. It is well-known that u(L)u(L) is less than or equal to half of c(L)c(L) (see for example [12]). In [12] Taniyama characterized the links which satisfy the equality as follows.

Refer to caption
Fig. 1.
Theorem 1.1.

[12, Theorem 1.5(2)1.5\,(2)] Let L be a μ\mu-component link that satisfies the equality u(L)=c(L)2u(L)=\dfrac{c(L)}{2}. Then LL has a diagram D=γ1γμD=\gamma_{1}\cup\cdots\cup\gamma_{\mu} such that each γi\gamma_{i} is a simple closed curve on 2\mathbb{R}^{2} and for each pair i,ji,\ j, the subdiagram γiγj\gamma_{i}\cup\gamma_{j} is an alternating diagram or a diagram without crossings.

In [1] Taniyama and the author showed that this inequality is not extended to spatial embeddings of planar graphs and this inequality is extended to spatial embeddings of trivializable planar graphs. Namely for any spatial embedding ff of a trivializable planar graph, u(f)u(f) is less than or equal to half of c(f)c(f). For example, a handcuff-graph and a theta-curve as illustrated in Fig. 2 are trivializable. We characterize the spatial embeddings of a handcuff-graph or a theta curve which satisfy the equality as follows.

Refer to caption handcuff-graph Refer to caption theta curve
Fig. 2.
Theorem 1.2.

Let GG be a handcuff-graph and let ff be a spatial embedding of GG. Then ff satisfies the equality u(f)=c(f)2u(f)=\dfrac{c(f)}{2} if and only if ff has a diagram DD with the following conditions :
(1)(1)\,Each edge of DD has no self-crossings.
(2)(2)\,All crossings of DD are crossings between two loops.
(3)(3)\,Two loops of DD form an alternating diagram or a diagram without crossings.

Theorem 1.3.

Let GG be a theta curve and let ff be a spatial embedding of GG. Then ff satisfies the equality u(f)=c(f)2u(f)=\dfrac{c(f)}{2} if and only if ff is trivial.

We note that unknotting numbers of spatial embeddings of a theta curve is studied in [2].

A handlebody-knot is an embedded handlebody in the 3-dimensional Euclidean space 3\mathbb{R}^{3}, which is introduced by Ishii in [5]. Two handlebody-knots H1H_{1} and H2H_{2} are equivalent if there is an orientation-preserving homeomorphism hh of 3\mathbb{R}^{3} with h(H1)=H2h(H_{1})=H_{2}. A spine of a handlebody-knot HH is a spatial graph whose regular neightborhood is HH. In this paper, we assume that spines have no degree 1 verticies. Any handlebody-knot HH can be represented by a spatial trivalent graph that is a spine of HH. In particular, genus 2 handlebody-knot can be represented by a spatial embedding of a handcuff-graph or a theta curve. A crossing change of a handlebody-knot HH is that of a spatial trivalent graph representing HH. In [6] Iwaliri showed that a crossing change of a handlebody-knot is an unknotting operation and give lower bounds of the unknotting numbers for handlebody-knots by the numbers of some finite Alexander quandle colorings.

We have the following well-known relation between unknotting number and crossing number of classical knots.

Proposition 1.4.

Let KK be a nontrivial knot. Then u(K)c(K)12u(K)\leq\dfrac{c(K)-1}{2}.

In [12] Taniyama characterized the knots which satisfy the equality as follows.

Theorem 1.5.

[12, Theorem1.4 (2)] Let KK be a nontrivial knot that satisfies the equality u(K)=c(K)12u(K)=\dfrac{c(K)-1}{2}. Then KK is a (2,p)(2,\ p)-torus knot for some odd number p±1p\neq\pm 1.

In this paper, as an extension of Proposition 1.4, we show the following theorem.

Theorem 1.6.

Let HH be a non-trivial handlebody-knot. Then u(H)c(H)12u(H)\leq\dfrac{c(H)-1}{2}.

The spine of genus 1 handlebody-knot is a classical knot. Therefore Theorem 1.6 is an extention of Proposition 1.4. It follows from Theorem 1.2 and Theorem 1.3 that for any non-trivial handlebody-knot HH with genus 2 twice the unknotting number of HH is less than or equal to the crossing number of HH minus one  (see section 4).

It follows from Theorem 1.5 that genus 1 handlebody-knot HH with u(H)=c(H)12u(H)=\dfrac{c(H)-1}{2} is a regular neighborhood of a (2,p)(2,\ p)-torus knot. We also characterize genus n2n\geq 2 handlebody-knots which satisfy the equality as follows.

Theorem 1.7.

Let n2n\geq 2 and let HH be a nontrivial genus nn handlebody-knot that satisfies the equality u(H)=c(H)12u(H)=\dfrac{c(H)-1}{2}. Then HH is a handlebody-knot represented by D3D_{3} or D3D_{-3} illustrated in Fig. 3.

Refer to caption Refer to caption
Fig. 3.

This paper consists of five sections. In section 2 we review trivializability of planar graphs and inequalities between unknotting numbers and crossing numbers of spatial embeddings of planar graphs. In section 3 we introduce unknotting number of handlebody-knots. In section 4 we give proofs of Theorem 1.2 and Theorem 1.3. In section 5 we give proofs of Theorem 1.6 and Theorem 1.7.

2. UNKNOTTING NUMBERS AND CROSSING NUMBERS OF SPATIAL EMBEDDINGS OF PLANAR GRAPHS

Let GG be a planar graph. A spatial embedding of GG is an embedding f:G3f:G\rightarrow\mathbb{R}^{3}. Its image f(G)f(G) is said to be a spatial graph. Let π:32\pi:\mathbb{R}^{3}\rightarrow\mathbb{R}^{2} be a natural projection defined by π(x,y,z)=(x,y)\pi(x,y,z)=(x,y). Let SE(G)SE(G) be the set of all spatial embeddings of GG. A regular projection of GG is a continuous map f~:G2\tilde{f}:G\rightarrow\mathbb{R}^{2} whose double points are only finitely many transversal double points. Such a double point is said to be a crossing point or simply a crossing. If we give over/under informations at each crossing points of a regular projection f~\tilde{f} of GG, then f~\tilde{f} together with the over/under informations represents a spatial embedding f:G3f:G\rightarrow\mathbb{R}^{3} such that f~=πf\tilde{f}=\pi\circ f. Such a regular projection together with the over/under informations is said to be a diagram of f(G)f(G). Then we say that ff is obtained from f~\tilde{f}. We also call f~\tilde{f} a regular projection of f(G)f(G). For a diagram DD of a spatial embedding, the set of all crossings of DD is denoted by C(D)\textit{C}(D). The number of crossings of DD is denoted by c(D)=|C(D)|c(D)=|\textit{C}(D)|.

An element fSE(G)f\in SE(G) is said to be trivial, if it is ambient isotopic to tSE(G)t\in SE(G) such that t(G)2t(G)\subset\mathbb{R}^{2}. Any spatial embedding of a planar graph can be transformed into trivial one by crossing changes. Therefore unknotting number is naturally extended to spatial embeddings of planar graphs as follows. For fSE(G)f\in SE(G), the unknotting number u(f)u(f) is defined to be the minimal number of crossing changes from ff to a trivial embedding of GG. The crossing number c(f)c(f) is defined to be the minimal number of crossing points among all diagrams of spatial embeddings that are ambient isotopic to ff.

For any link LL, LL satisfies the inequality u(L)c(L)2u(L)\leq\dfrac{c(L)}{2}. But this is not extended for spatial embeddings of planar graph, namely there are a planar graph GG and a spatial embedding ff of GG such that u(f)>c(f)2u(f)>\dfrac{c(f)}{2}. Let P3P_{3} the cube graph and f3SE(P3)f_{3}\in SE(P_{3}) a spatial embedding of P3P_{3} as illustrated in Fig. 4. The spatial graph f3(P3)f_{3}(P_{3}) contains three Hopf-links and one crossing change of edges of f3(P3)f_{3}(P_{3}) unknot at most two of them (See Fig. 5). Then u(f3)2u(f_{3})\geq 2. Since f3(P3)f_{3}(P_{3}) contains a trefoil whose crossing number is 3, c(f3)=3c(f_{3})=3 and u(f3)>c(f3)2u(f_{3})>\dfrac{c(f_{3})}{2} [1].

Refer to caption
Fig. 4.
Refer to caption
Fig. 5.

Now we review the reason why it happens for some planar graphs. The key point of the proof of u(L)c(L)2u(L)\leq\dfrac{c(L)}{2} for a link LL is that any link diagram can be transformed into a trivial link diagram by changing over/under informations at some crossings of the diagram. Let DD be a minimal crossing diagram of LL. Let AA be a subset of C(D)C(D) such that changing over/under informations at all crossings in AA turns DD to a diagram T1T_{1} of a trivial link. Let T2T_{2} be a diagram that is obtained from T1T_{1} by changing over/under informations at all crossings. A mirror image of a trivial link is also trivial. Thus T2T_{2} is a diagram of a trivial link. Note that T2T_{2} is obtained from DD by changing over/under informations at all crossings in C(D)AC(D)-A. Therefore we have

u(L)u(D)min{|A|,|C(D)A|}c(D)2=c(L)2u(L)\leq u(D)\leq\min\{|A|,|C(D)-A|\}\leq\dfrac{c(D)}{2}=\dfrac{c(L)}{2}

On the other hand, all diagrams obtained from πf3(P3)\pi\circ f_{3}(P_{3}) (Fig. 6) represent non-trivial spatial graphs since each of the spatial graphs obtained from these diagrams contains at least one Hopf-link. A regular projection f~\tilde{f} of a planar graph GG is said to be a knotted projection [11], if all spatial embeddings of GG which can be obtained from f~\tilde{f} are non-trivial.

Refer to caption

πf3(P3)\pi\circ f_{3}(P_{3})

Fig. 6.

A planar graph is said to be trivializable if it has no knotted projections. In [11] Taniyama gave a class of trivializable graphs. In [9] Sugiura and Suzuki extended the class. In [10] Tamura gave another class of trivializable graphs.

For a spatial embedding of a trivializable planar graph, the same argument as for a link works, and we have the following proposition.

Proposition 2.1.

[1] Let GG be a trivializable planar graph and f:G3f:G\rightarrow\mathbb{R}^{3} a spatial embedding of GG. Then u(f)c(f)2u(f)\leq\dfrac{c(f)}{2}.

3. UNKNOTTING NUMBERS AND CROSSING NUMBERS OF HANDLEBODY-KNOTS

We review that crossing change of a handlebody-knot is an unknotting operation [6].

A diagram of a handlebody-knot HH is that of a spatial trivalent graph representing HH. In [5], Ishii gave a list of fundamental moves among diagrams of handlebody-knots, which is called R1-6 moves illustrated in Fig. 7. Ishii showed that two handlebody-knots are equivalent if and only if their representing diagrams are related by a finite sequence of R1-6 moves. Note that R6-move is also called IH-move.

Refer to caption Refer to caption Refer to caption
Refer to caption Refer to caption Refer to caption
Fig. 7.

A crossing change of a handlebody-knot HH is that of a spatial trivalent graph representing HH. This move can be realized by switching two tubes illustrated in Fig. 8. A genus nn handlebody-knot is trivial if it is equivalent to a handlebody-knot represented by a diagram illustrated in Fig. 9.

Refer to caption
Fig. 8.
Refer to caption
Fig. 9.

Let TnT_{n} be the trivalent graph whose image is illustrated in Fig. 9. Any handlebody-knot is represented by a diagram of a spatial embedding of TnT_{n} since a genus nn handlebody has TnT_{n} as a spine. Note that TnT_{n} is a trivializable graph [9]. Namely, any diagram DD of a spatial embedding of TnT_{n} can be changed to a trivial spatial graph diagram by changing over/under informations at some crossings of DD. Then we have the following proposition.

Proposition 3.1.

[6, Proposition 2.1] Any handlebody-knot can be transformed into trivial one by crossing changes.

Therefore unknotting number is naturally extended to handlebody-knots as follows. For a handlebody-knot HH, the unknotting number u(H)u(H) is the minimal number of crossing changes needed to obtain a trivial handlebody-knot from HH. The crossing number c(H)c(H) is the minimal number of crossing points among all diagrams of handlebody-knots that are equivalent to HH.

By the proof of [8, Proposition 3.1] we see that any diagram DD of a spatial graph can be transformed into a diagram of a spatial graph whose neighborhood are ambient isotopic to a neighborhood of a trivial bouquet by changing over/under informations at some crossings of DD. Therefore, in [6], Iwakiri also showed that Proposition 3.1 can be refined to the strong statement as follows.

Proposition 3.2.

[6] Any handlebody-knot diagram can be transformed into a trivial handlebody-knot diagram by changing over/under informations at some crossings of the diagram.

For a handlebody-knot diagram DD, the unknotting number u(D)u(D) is the minimal number of changing over/under informations at crossings of DD needed to obtain a trivial handlebody-knot diagram. Same as Proposition 2.1 we have u(D)c(D)2u(D)\leq\dfrac{c(D)}{2} and u(H)c(H)2u(H)\leq\dfrac{c(H)}{2}.

In section 5, we show that a handlebody-knot HH satisfies u(H)=c(H)2u(H)=\dfrac{c(H)}{2} if and only if HH is trivial (Theorem 1.6). Then it is natural to ask when handlebody-knots satisfy the equality u(H)=c(H)12u(H)=\dfrac{c(H)-1}{2}. Let HHH1H_{1} and H2H_{2} be handlebody-knots in 3\mathbb{R}^{3} and let SS be a 22-sphere in 3\mathbb{R}^{3}. Suppose that HS=H1H2H\cap S=H_{1}\cap H_{2} is a 22-disk and H=H1H2H=H_{1}\cup H_{2}. Then HH is said to be a disk sum of H1H_{1} and H2H_{2} and denoted by H=H1#H2H=H_{1}\#H_{2}. In [12] Taniyama showed that if a classical knot KK satisfies u(K)=c(K)12u(K)=\dfrac{c(K)-1}{2} then KK is a (2,p)(2,\ p)-torus knot for some odd number p±1p\neq\pm 1 (Theorem 1.5). Therefore the handlebody-knots illustrated in Fig. 10 may satisfy the equality. But by the following proposition only two of these handlebody-knots satisfy the equality.

Proposition 3.3.

Let n2n\geq 2 and let HH be a genus nn handlebody-knot such that H=K#On1H=K\ \#\ O_{n-1}, where KK is a genus 11 handlebody-knot whose spine is a 22-bridge knot and On1O_{n-1} is a genus n1n-1 trivial handlebody-knot. Then u(H)=1u(H)=1.

Refer to caption D3D_{3} Refer to caption D5D_{5} Refer to caption D7D_{7}
Refer to caption D3D_{-3} Refer to caption D5D_{-5} Refer to caption D7D_{-7}
Fig. 10.
Proof.

Let KK^{\prime} be the spine of KK. Let HH^{\prime} be the handlebody-knot obtained from K#On1K\#O_{n-1} by one crossing change as illustrated in the left of Fig. 11. By [7, Proposition 3.1] we see that the tunnel τ\tau for KK^{\prime} as illustrated in the right of Fig. 11 is an unknotting tunnel. Therefore the genus 2 handlebody-knot represented by the right of Fig. 11 is trivial. Since a disk sum of two trivial handlebody-knots is trivial, HH^{\prime} is also trivial.

Refer to caption b1,b2,bn: 2b_{1},\ b_{2},\ \cdots b_{n}:\ 2-braids Refer to caption
Fig. 11.

4. PROOFS OF THEOREM 1.2 AND THEOREM 1.3

Let DD be a diagram of a spatial graph f(G)f(G) and let HH be a subgraph of GG. Then the diagram of f(H)f(H) that is contained in DD is said to be a subdiagram of DD. For subdiagrams A,BA,\ B of a diagram DD, let c(A)c(A) be the number of all crossings on AA among the crossings of DD and let c(A,B)c(A,\,B) be the number of all crossings between AA and BB.

Lemma 4.1.

Let GG be a trivializable graph and let ff be a spatial embedding of GG. Let DD be a diagram of f(G)f(G). If DD has a self-crossing, then u(D)c(D)12u(D)\leq\dfrac{c(D)-1}{2}.

Proof.

Let PP be a self-crossing of DD. By smoothing DD at PP, we have a diagram DD^{\prime} such that one of the components of DD^{\prime} represents a knot (see Fig. 12). Let γ1\gamma_{1} be a component of DD^{\prime} that represents a knot and let γ2\gamma_{2} be the other component of DD^{\prime}.

Refer to caption DD Refer to caption DD^{\prime}
Fig. 12.

If we change some crossings on γ1\gamma_{1} so that the part γ1\gamma_{1} is over other component of DD and itself unknotted then we have a spatial embedding that has a diagram γ2\gamma_{2}. Also we may change some crossings on γ1\gamma_{1} so that the part γ1\gamma_{1} is under the other component of DD and itself unknotted. Note that we can choose these two crossing changes complementary on the crossings on γ1\gamma_{1}. We choose one of them that have no more crossing changes than the other. Thus by changing no more than c(D)c(γ2)12\dfrac{c(D)-c(\gamma_{2})-1}{2} crossings of DD we have a spatial embedding that has a diagram γ2\gamma_{2}. Note that the key point here is that we do not need to change the crossing cc. Since γ2\gamma_{2} is also a diagram of a spatial embedding of a trivializable graph, we have u(γ2)c(γ2)2u(\gamma_{2})\leq\dfrac{c(\gamma_{2})}{2}. Therefore we have

u(D)u(γ2)+c(D)c(γ2)12c(γ2)+(c(D)c(γ2)1)2=c(D)12u(D)\leq u(\gamma_{2})+\dfrac{c(D)-c(\gamma_{2})-1}{2}\leq\dfrac{c(\gamma_{2})+(\ c(D)-c(\gamma_{2})-1\ )}{2}=\dfrac{c(D)-1}{2}

Lemma 4.2.

Let GG be a trivializable planar graph and let ff be a spatial embedding of GG such that u(f)=c(f)2u(f)=\dfrac{c(f)}{2}. Let DD be a minimal crossing diagram of f(G)f(G). Then u(D)=c(D)2u(D)=\dfrac{c(D)}{2}.

Proof.

It is sufficient to show that u(D)c(D)2u(D)\geq\dfrac{c(D)}{2}. Since u(f)u(D)u(f)\leq u(D) and c(f)=c(D)c(f)=c(D) we have

u(D)u(f)=c(f)2=c(D)2u(D)\geq u(f)=\dfrac{c(f)}{2}=\dfrac{c(D)}{2}

Lemma 4.3.

Let DD be a diagram of a spatial embedding of a handcuff-graph such that u(D)=c(D)2u(D)=\dfrac{c(D)}{2}. Then DD satisfies the following conditions :
(1)(1)\,Each edge of DD has no self-crossings.
(2)(2)\,All crossings of DD are crossings between two loops.
(3)(3)\,Two loops of DD form an alternating diagram or a diagram without crossings.

Proof.

By Lemma 4.1 DD satisfies (1)(1). In the following we show that DD satisfies (2)(2) and (3)(3).

Let γ1\gamma_{1} and γ2\gamma_{2} be two loops of DD and let ee be the edge of DD that is not γi\gamma_{i} (i=1, 2i=1,\ 2). If we change some crossings on γ2\gamma_{2} so that the part γ2\gamma_{2} is over Dγ2D-\gamma_{2} of DD then we have a diagram of a trivial spatial embedding of GG since γi\gamma_{i} is a simple closed curve on 2\mathbb{R}^{2} (i=1, 2i=1,\ 2). See for example Fig. 13. Also we may change some crossings on γ2\gamma_{2} that the part γ2\gamma_{2} is under Dγ2D-\gamma_{2} of D and itself unknotted. Note that these two crossing changes are complementary on the crossings on γ2\gamma_{2}. We choose one of them that have no more crossing changes than the other. Thus by changing no more than c(D)c(γ1,e)2\dfrac{c(D)-c(\gamma_{1},\ e)}{2} crossings of DD we have a trivial diagram and u(D)c(D)c(γ1,e)2u(D)\leq\dfrac{c(D)-c(\gamma_{1},\ e)}{2}. The key point here is that we do not need to change crossings between γ1\gamma_{1} and ee. Since u(D)=c(D)2u(D)=\dfrac{c(D)}{2} we have c(γ1,e)=0c(\gamma_{1},\ e)=0. Similarly we have c(γ2,e)=0c(\gamma_{2},\ e)=0. Therefore DD satisfies (2)(2).

Refer to caption Refer to caption
Fig. 13.

Suppose that γ1γ2\gamma_{1}\cup\gamma_{2} is not an alternating diagram. Then we may suppose without loss of generality that there is an arc α\alpha of γ1\gamma_{1} disjoint from ee such that αγ2=α={c1,c2}\alpha\cap\gamma_{2}=\partial\alpha=\{c_{1},\ c_{2}\} and γ1\gamma_{1} is over γ2\gamma_{2} at c1c_{1} and c2c_{2}. See Fig. 14 .

Refer to caption
Fig. 14.

Let AA be the set of all crossings of DD at which γ1\gamma_{1} is under γ2\gamma_{2}. Let B=C(D)\(A{c1.c2})B=C(D)\backslash(A\cup\{c_{1}.\ c_{2}\}). Then by the height function argument first used in [11] we see that changing all crossings in AA (resp. BB) produce a trivial spatial embedding. See Fig. 15.

Refer to caption Refer to caption
Fig. 15.

Therefore we have

u(D)min{|A|,|B|}c(D)22u(D)\leq\min\{|A|,\ |B|\}\leq\dfrac{c(D)-2}{2}

This is contradicts to the equation u(D)=c(D)2u(D)=\dfrac{c(D)}{2}. Thus DD satisfies (3)(3) as desired. ∎

Proof of Theorem1.2
First, we show that if there exists a diagram DD of f(G)f(G) satisfying (1)(1), (2)(2) and (3)(3), then u(f)=c(f)2u(f)=\dfrac{c(f)}{2}. We may suppose that c(D)>0c(D)>0. Let L=l1l2L=l_{1}\cup l_{2} be a 22-component link represented by two loops of DD. See for example Fig. 16. Since the diagram of LL consists of two simple closed curves and it is alternating, we see that twice the absolute value of the linking number 2|lk(l1,l2)|2|\,lk(l_{1},l_{2})\,| is equal to C(D)C(D). Therefore we have

u(f)u(L)|lk(l1,l2)|=c(D)2=c(f)2u(f)\geq u(L)\geq|\,lk(l_{1},l_{2})\,|=\dfrac{c(D)}{2}=\dfrac{c(f)}{2}

By Proposition 2.1 we have u(f)=c(f)2u(f)=\dfrac{c(f)}{2}.

Refer to caption
Fig. 16.

Let ff be a spatial embedding of GG such that u(f)=c(f)2u(f)=\dfrac{c(f)}{2} and let DD be a minimal crossing diagram of f(G)f(G). By Lemma 4.2 we have u(D)=c(D)2u(D)=\dfrac{c(D)}{2}. By Lemma 4.3 DD satisfies (1)(1), (2)(2) and (3)(3) as desired. ∎

Lemma 4.4.

Let GG be a theta curve. Let DD be a diagram of a spatial embedding of GG such that u(D)=c(D)2u(D)=\dfrac{c(D)}{2}. Then c(D)=0c(D)=0.

Proof.

By Lemma 4.1 we may suppose that each edge of DD has no self-crossings. Suppose that c(D)>0c(D)>0. Then there exists a crossing cc on DD between two edges. Let f~:G2\tilde{f}:G\rightarrow\mathbb{R}^{2} be a regular projection of GG where DD is obtained from f~(G)\tilde{f}(G). Let vv and uu be two vertices of GG. Let GG^{\prime} be the graph obtained by adding 22 verticies v1,v1v_{1},\ v_{1}^{\prime} to GG such that f~(v1)=f~(v1)=c\tilde{f}(v_{1})=\tilde{f}(v_{1}^{\prime})=c and v1v_{1} (resp. v1v_{1}^{\prime}) is contained in the over-arc (resp. the under-arc) at cc. Let PP be the path from vv to uu that contains v1v_{1}. We fix a spanning tree TT of GG^{\prime} that contains PP (see for example Fig. 17). Let h:Gh:G^{\prime}\rightarrow\mathbb{R} be a continuous function with the following properties :
(1)(1)\,For each vertex tt of GG^{\prime}, h(t)=dT(t,v)h(t)=-d_{T}(t,\ v). Here dT(t,v)d_{T}(t,v) be the number of edges of the path in TT joining tt and vv.
(2)h|e(2)\,h|_{e} is injective for each edge ee of GG^{\prime}

Refer to caption DD Refer to caption GG^{\prime}
Fig. 17.

We can deform hh slightly so that h(v1)>h(v1)h(v_{1})>h(v_{1}^{\prime}) since dT(v,v1)=dT(v,v1)=1d_{T}(v,v_{1})=d_{T}(v,v_{1}^{\prime})=1. Then we give over/under information to f~\tilde{f} to produce a spatial embedding f:G3=2×f:G\rightarrow\mathbb{R}^{3}=\mathbb{R}^{2}\times\mathbb{R} such that p1f=f~p_{1}\circ f=\tilde{f} and p2f=hp_{2}\circ f=h, where p1p_{1} (resp. p2)p_{2}) denotes the projection of 3\mathbb{R}^{3} to the first factor (respectively to the second factor) of 2×\mathbb{R}^{2}\times\mathbb{R}. Let Π:32\Pi:\mathbb{R}^{3}\rightarrow\mathbb{R}^{2} be a projection defined by Π(x,y,z)=(x,z)\Pi(x,y,z)=(x,z). We deform ff slightly by an ambient isotopy if necessary so that Πf\Pi\circ f is a regular projection. Then we can eliminate all crossings of Πf\Pi\circ f by eliminating the crossing nearest to vv repeatedly (see Fig. 18). Therefore ff is trivial.

Refer to caption
Fig. 18.

Let DD^{\prime} be the diagram of f(G)f(G) where DD^{\prime} is obtained from f~(G)\tilde{f}(G). We note that DD and DD^{\prime} are deformed into each other by changing over/under informations of all crossing points without changing over/under informations of cc. Let D′′D^{\prime\prime} be the diagram that is obtained from DD^{\prime} by changing over/under informations of all crossing points with the exception of cc (see for example Fig. 19). Let h:Gh^{\prime}:G^{\prime}\rightarrow\mathbb{R} be a continuous function such that h=hh^{\prime}=-h. We can deform hh^{\prime} slightly so that h(v1)>h(v1)h^{\prime}(v_{1})>h^{\prime}(v_{1}^{\prime}). Then D′′D^{\prime\prime} is the diagram of a spatial embedding of f:G3=2×f^{\prime}:G\rightarrow\mathbb{R}^{3}=\mathbb{R}^{2}\times\mathbb{R} such that p1f=f~p_{1}\circ f^{\prime}=\tilde{f} and p2f=hp_{2}\circ f^{\prime}=h^{\prime}. Same as the case of f(G)f(G), ff^{\prime} is also trivial.

Refer to caption DD^{\prime} Refer to caption D′′D^{\prime\prime}
Fig. 19.

Let AA be a subset of C(D)C(D) such that changing all crossings in AA turns DD to DD^{\prime}. We note that changing all crossings in (C(D){c})A(\ C(D)-\{c\}\ )-A turns DD to D′′D^{\prime\prime}. Therefore we have

u(D)min{|A|,|(C(D){c})A|}c(D)12u(D)\leq\min\{|A|,|(\ C(D)-\{c\}\ )-A|\}\leq\dfrac{c(D)-1}{2}

This is contradicts to the equation u(D)=c(D)2u(D)=\dfrac{c(D)}{2}. Therefore we have c(D)=0c(D)=0 and DD is a diagram of a trivial theta curve. ∎

Proof of Theorem1.3
Let ff be a spatial embedding of GG such that u(f)=c(f)2u(f)=\dfrac{c(f)}{2} and let DD be a minimal crossing diagram of f(G)f(G). By Lemma 4.2 we have u(D)=c(D)2u(D)=\dfrac{c(D)}{2}. By Lemma 4.4 we see that ff is trivial.∎

Remark 4.5.

We can prove Theorem 1.6 in the case of genus 2 by observing Lemma 4.3 and Lemma 4.4. Let DD be a minimal crossing diagram of a non-trivial genus 2 handlebody-knot HH. Then DD is also a diagram of a spatial embedding of a handcuff-graph or a theta curve.

In the case DD is a diagram of a spatial handcuff-graph, by Lemma 4.3 all crossings of DD are between two loops or u(D)c(D)12u(D)\leq\dfrac{c(D)-1}{2}. In the case all crossings of DD are between two loops, by one IH-move on the edge that is not a loop we have a diagram DD^{\prime} of HH such that c(D)=c(D)=c(H)c(D^{\prime})=c(D)=c(H) and DD^{\prime} is also a diagram of a spatial theta curve (see Fig. 20). By Lemma 4.4 we have u(D)c(D)12u(D^{\prime})\leq\dfrac{c(D^{\prime})-1}{2}.

In the case DD is a diagram of a spatial theta curve, by Lemma 4.4 we have u(D)c(D)12u(D)\leq\dfrac{c(D)-1}{2}. In the both cases we have u(H)c(H)12u(H)\leq\dfrac{c(H)-1}{2}.

Refer to caption DD Refer to caption DD^{\prime}
Fig. 20.

5. PROOFS OF THEOREM 1.6 AND THEOREM 1.7

In this section we prove Theorem 1.6 and Theorem 1.7. In the following we give an inequality between unknotting number and crossing number by an observation of subdivided graph.

Let f~:G2\tilde{f}:G\rightarrow\mathbb{R}^{2} be a regular projection of a graph GG. Let c1,c2,,ckc_{1},\ c_{2},\ \cdots,\ c_{k} be crossing points of f~(G)\tilde{f}(G). A subdivided graph of GG at {c1,c2,,ck}\{c_{1},\ c_{2},\ \cdots,\ c_{k}\} is a graph obtained by adding 2k2k vertices v1,v1,v2,v2,,vk,vkv_{1},\ v_{1}^{\prime},\ v_{2},\ v_{2}^{\prime},\ \cdots,v_{k},\ v_{k}^{\prime} to GG such that f~(vi)=f~(vi)=ci\tilde{f}(v_{i})=\tilde{f}(v_{i}^{\prime})=c_{i} and viv_{i} (resp. viv_{i}^{\prime}) is contained in the over-arc (resp. the under-arc) at cic_{i}. Then we say that viv_{i} (resp. vi)v_{i}^{\prime}) is an over-vertex (resp. under-vertex) at cic_{i}. Let TT be a spanning tree of GG^{\prime}. For any two vertices vv and uu of GG^{\prime}, let dT(v,u)d_{T}(v,u) be the number of edges of the path in TT joining vv and uu.

Lemma 5.1.

Let DD be a diagram of a nontrivial handlebody-knot HH. Let f~:G2\tilde{f}:G\rightarrow\mathbb{R}^{2} be a regular projection of a connected trivalent graph GG where DD is obtained from f~(G)\tilde{f}(G). Let c1,c2,,ckc_{1},\ c_{2},\ \cdots,\ c_{k} be crossing points of f~(G)\tilde{f}(G). Let GG^{\prime} be the subdivided graph of GG at {c1,c2,,ck}\{c_{1},\ c_{2},\ \cdots,\ c_{k}\}. Let v1,v1,v2,v2,,vk,vkv_{1},\ v_{1}^{\prime},\ v_{2},\ v_{2}^{\prime},\ \cdots,v_{k},\ v_{k}^{\prime} be vertices of GG^{\prime} such that viv_{i} (resp. viv_{i}^{\prime}) is an over-vertex (resp. under-vertex) at ci(i=1,2,,k)c_{i}\ (i=1,2,\cdots,k). If there exists a vertex vv of GG^{\prime} and a spanning tree TT of GG^{\prime} such that dT(v,vi)=dT(v,vi)d_{T}(v,v_{i})=d_{T}(v,v_{i}^{\prime}) for all i{1,2,,k}i\in\{1,2,\cdots,k\}, then u(D)c(D)k2u(D)\leq\dfrac{c(D)-k}{2}.

Proof.

The proof is analogous to the proof of [8, Proposition 3.2]. We fix a vertex vv of GG^{\prime} and a spanning tree TT of GG^{\prime} such that dT(v,vi)=dT(v,vi)d_{T}(v,v_{i})=d_{T}(v,v_{i}^{\prime}) for all i{1,2,,k}i\in\{1,2,\cdots,k\} (see Fig. 21). Let h:Gh:G^{\prime}\rightarrow\mathbb{R} be a continuous function with the following properties :

Refer to caption Refer to caption
Fig. 21.

(1)(1)\,For each vertex uu of TT, h|T(u)=dT(v,u)h|_{T}(u)=-d_{T}(v,u).
(2)h|e(2)\,h|_{e} is injective for each edge ee of TT.
(3)(3)\,Each edge of GTG^{\prime}-T has exactly one minimum point of hh.
We can deform hh slightly so that h(vi)>h(vi)(i=1,2,,k)h(v_{i})>h(v_{i}^{\prime})\ (i=1,2,\cdots,k) since dT(v,vi)=dT(v,vi)(i=1,2,,k)d_{T}(v,v_{i})=d_{T}(v,v_{i}^{\prime})\ (i=1,2,\cdots,k). Then we give over/under informations to f~\tilde{f} to produce a spatial embedding f:G3=2×f:G\rightarrow\mathbb{R}^{3}=\mathbb{R}^{2}\times\mathbb{R} such that p1f=f~p_{1}\circ f=\tilde{f} and p2f=hp_{2}\circ f=h, where p1p_{1} (respectively p2)p_{2}) denotes the projection of 3\mathbb{R}^{3} to the first factor (respectively to the second factor) of 2×\mathbb{R}^{2}\times\mathbb{R}.

Let DD^{\prime} be the diagram of f(G)f(G) where DD^{\prime} is obtained from f~(G)\tilde{f}(G). We note that DD and DD^{\prime} are deformed into each other by changing over/under informations of crossing points without changing over/under informations of c1,c2,,ckc_{1},\ c_{2},\ \cdots,\ c_{k} (see for example Fig. 22). Since we obtain a bouquet as in Fig. 23 which is trivial by contracting spatial edges of f(T)f(T), DD^{\prime} is a diagram of a trivial handlebody-knot.

Refer to caption Refer to caption
Fig. 22.
Refer to caption Refer to caption
Fig. 23.

Let D′′D^{\prime\prime} be the diagram that is obtained from DD^{\prime} by changing over/under informations of all crossing points with the exception of c1,c2,,ckc_{1},\ c_{2},\ \cdots,\ c_{k}. Let h:Gh^{\prime}:G^{\prime}\rightarrow\mathbb{R} be a continuous function such that h=hh^{\prime}=-h. We can deform hh^{\prime} slightly so that h(vi)>h(vi)(i=1,2,,k)h^{\prime}(v_{i})>h^{\prime}(v_{i}^{\prime})\ (i=1,2,\cdots,k)  (see for example Fig. 24). Then D′′D^{\prime\prime} is the diagram of a spatial embedding of f:G3=2×f^{\prime}:G\rightarrow\mathbb{R}^{3}=\mathbb{R}^{2}\times\mathbb{R} such that p1f=f~p_{1}\circ f^{\prime}=\tilde{f} and p2f=hp_{2}\circ f^{\prime}=h^{\prime}. Thus we obtain a trivial bouquet by contracting spatial edges of f(T)f^{\prime}(T) and D′′D^{\prime\prime} is a diagram of a trivial handlebody-knot.

Refer to caption Refer to caption
Fig. 24.

Let AA be a subset of C(D)C(D) such that changing all crossings in AA turns DD to DD^{\prime}. We note that changing all crossings in (C(D){c1,c2,,ck})A(\ C(D)-\{c_{1},\ c_{2},\ \cdots,\ c_{k}\}\ )-A turns DD to D′′D^{\prime\prime}. Therefore we have

u(D)min{|A|,|(C(D){c1,c2,,ck})A|}c(D)k2u(D)\leq\min\{|A|,|(\ C(D)-\{c_{1},\ c_{2},\ \cdots,\ c_{k}\}\ )-A|\}\leq\dfrac{c(D)-k}{2}

Proof of Theorem 1.6
Let DD be a minimal crossing diagram of a nontrivial handlebody-knot HH. Let f~:G2\tilde{f}:G\rightarrow\mathbb{R}^{2} be a regular projection of a connected trivalent graph GG where DD is obtained from f~(G)\tilde{f}(G). Let c1c_{1} be a crossing point of f~(G)\tilde{f}(G). Let GG^{\prime} be the subdivided graph of GG at {c1}\{c_{1}\}. Let v1v_{1} (resp. v1v_{1}^{\prime}) be a vertex of GG^{\prime} such that v1(resp.v1)v_{1}\ (resp.\ v_{1}^{\prime}) is over-vertex (resp. under-vertex) at c1c_{1}. Let TT be a spanning tree of GG^{\prime} containing v1v_{1} and v1v_{1}^{\prime}.

Refer to caption Refer to caption
Fig. 25.

By subdividing TT if necessary, we can choose a vertex vv of TT such that dT(v,v1)=dT(v,v1)d_{T}(v,v_{1})=d_{T}(v,v_{1}^{\prime}) since there exists a path in TT joining v1v_{1} and v1v_{1}^{\prime} (see for example Fig. 25). Then by Lemma 5.1 we have u(D)c(D)12u(D)\leq\dfrac{c(D)-1}{2}. Since u(H)u(D)u(H)\leq u(D) and c(D)=c(H)c(D)=c(H) we have u(H)c(H)12u(H)\leq\dfrac{c(H)-1}{2}. ∎

Lemma 5.2.

Let DD be a minimal crossing diagram of a nontrivial handlebody-knot HH that satisfies the equality u(D)=c(D)12u(D)=\dfrac{c(D)-1}{2}. Let γ\gamma be a cycle of DD that has at least one crossing of DD. Then the following (1)(1) and (2)(2) holds.
(1)(1)\,All crossings of DD are self-crossings of γ\gamma.
(2)(2)\,There exists an odd number p±1p\neq\pm 1 such that γ\gamma is a reduced alternating diagram of a (2,p)(2,\ p)-torus knot.

Proof.

Suppose that γ\gamma has just one crossing. Suppose that γ\gamma itself is a simple closed curve on 2\mathbb{R}^{2}. Then we have a diagram DD^{\prime} of HH as illustrated in Fig. 26 such that c(D)=c(D)1c(D^{\prime})=c(D)-1. Suppose that γ\gamma is not a simple closed curve on 2\mathbb{R}^{2} and γ\gamma has exactly one crossing of DD. Then by a similar deformation we have a diagram DD^{\prime} of HH with c(D)=c(D)1c(D^{\prime})=c(D)-1. This contradicts that DD is a minimal crossing diagram. Thus γ\gamma has at least two crossings of DD.

Refer to caption DD Refer to caption Refer to caption DD^{\prime}
Fig. 26.

Let f~\tilde{f} be a regular projection of a trivalent graph GG where DD is obtained from f~(G)\tilde{f}(G). First, we show that if (1)(1) does not hold, then u(D)c(D)22u(D)\leq\dfrac{c(D)-2}{2}. We will show this claim step by step as follows.
 
Subclaim 1. If one of the crossings on γ\gamma, say c1c_{1}, is a crossing between γ\gamma and DγD-\gamma and another crossing on γ\gamma, say c2c_{2}, is a self-crossing of γ\gamma, then u(D)c(D)22u(D)\leq\dfrac{c(D)-2}{2}.

Proof.

Let GG^{\prime} be the subdivided graph of GG at {c1,c2}\{c_{1},\ c_{2}\}. Let viv_{i} (resp. vi)v_{i}^{\prime}) be the over-vertex (resp. under-vertex) at cic_{i} (i=1, 2)(i=1,\ 2). Then GG^{\prime} is the graph as illustrated in Fig. 27 (a)(a) or (b)(b).

Refer to caption
aa
Refer to caption
bb
Fig. 27.
Refer to caption
Fig. 28.

By subdividing if necessary, we can choose a spanning tree TT of GG^{\prime} and a vertex vv of TT such that dT(v,vi)=dT(v,vi)(i=1,2)d_{T}(v,v_{i})=d_{T}(v,v_{i}^{\prime})\ (i=1,2). A choice of TT and vv for the case of Fig. 27 (a)(a) is illustrated in Fig. 28. By Lemma 5.1 we have u(D)c(D)22u(D)\leq\dfrac{c(D)-2}{2}. ∎

Subclaim 2. If two of crossings on γ\gamma, say c1c_{1} and c2c_{2}, are crossings between γ\gamma and DγD-\gamma then u(D)c(D)22u(D)\leq\dfrac{c(D)-2}{2}.

Proof.

Let GG^{\prime} be the subdivided graph of GG at {c1,c2}\{c_{1},\ c_{2}\}. Let viv_{i} (resp. vi)v_{i}^{\prime}) be the over-vertex (resp. under-vertex) at cic_{i} (i=1, 2)(i=1,\ 2). Then GG^{\prime} is one of the graphs as illustrated in Fig. 29.

Refer to caption
aa
Refer to caption
bb
Refer to caption
cc
Refer to caption
dd
Fig. 29.
Refer to caption
Fig. 30.

By subdividing if necessary, we can choose a spanning tree TT of GG^{\prime} containing v1,v1,v2,v2v_{1},\,v_{1}^{\prime},\,v_{2},\,v_{2}^{\prime} and a vertex vv of TT such that dT(v,vi)=dT(v,vi)(i=1,2)d_{T}(v,v_{i})=d_{T}(v,v_{i}^{\prime})\ (i=1,2). A choice of TT and vv for the case of Fig. 29 (a)(a) is illustrated in Fig. 30. By Lemma 5.1 we have u(D)c(D)22u(D)\leq\dfrac{c(D)-2}{2}. ∎

Subclaim 3. If there exists a self-crossing of DγD-\gamma, say c1c_{1}, then u(D)c(D)22u(D)\leq\dfrac{c(D)-2}{2}.

Proof.

By Subclaim 2 we may assume that γ\gamma has a self-crossing, say c2c_{2}. Let GG^{\prime} be the subdivided graph of GG at {c1,c2}\{c_{1},\ c_{2}\}. Let viv_{i} (resp. vi)v_{i}^{\prime}) be the over-vertex (resp. under-vertex) at cic_{i} (i=1, 2)(i=1,\ 2). Then GG^{\prime} is one of the graphs as illustrated in Fig. 31.

Refer to caption
aa
Refer to caption
bb
Refer to caption
cc
Refer to caption
dd
Fig. 31.
Refer to caption
Fig. 32.

By subdividing if necessary, we can choose a spanning tree TT of GG^{\prime} containing v1,v1,v2,v2v_{1},\,v_{1}^{\prime},\,v_{2},\,v_{2}^{\prime} and a vertex vv of TT such that dT(v,vi)=dT(v,vi)(i=1,2)d_{T}(v,v_{i})=d_{T}(v,v_{i}^{\prime})\ (i=1,2). A choice of TT and vv for the case Fig. 31 (a)(a) is illustrated in Fig. 32. By Lemma 5.1 we have u(D)c(D)22u(D)\leq\dfrac{c(D)-2}{2}. ∎

From the above we see that γ\gamma satisfies (1)(1).
 
Subclaim 4. If γ\gamma is not obtained from a standard projection of a (2,p)(2,\ p)-torus knot for any odd number p>1p>1, then u(D)c(D)22u(D)\leq\dfrac{c(D)-2}{2}.

Proof.

Let GG^{\prime} be the subdivided graph of GG at C(D)C(D) and let Γ\Gamma be a cycle of GG^{\prime} such that γ\gamma is obtained from f~(Γ)\tilde{f}(\Gamma). Note that if f~(Γ)\tilde{f}(\Gamma) is a standard projection of a (2,p)(2,\ p)-torus knot for some odd number p±1p\neq\pm 1 as the case p=5p=5 is illustrated in the left of Fig. 33, then any pair of crossings on γ\gamma are not parallel. Namely Γ\Gamma is as illustrated in the right of Fig. 33. It follows from [3, Theorem 1] that the converse is also true (see also [4, Proof of Theorem 1.11]).

Refer to caption Refer to caption
Fig. 33.

Therefore there are two crossings c1c_{1}, c2c_{2} of γ\gamma such that c1c_{1} and c2c_{2} are parallel, namely Γ\Gamma is illustrated as the left of Fig. 34. Let viv_{i} (resp. vi)v_{i}^{\prime}) be the over-vertex (resp. under-vertex) at cic_{i} (i=1, 2)(i=1,\ 2). By subdividing if necessary, we can choose a spanning tree TT of GG^{\prime} and a vertex vv of TT such that dT(v,vi)=dT(v,vi)(i=1,2)d_{T}(v,v_{i})=d_{T}(v,v_{i}^{\prime})\ (i=1,2) (see the right of Fig. 34). By Lemma 5.1 we have u(D)c(D)22u(D)\leq\dfrac{c(D)-2}{2}.

Refer to caption Refer to caption
Fig. 34.

Finally, we show that if u(D)=c(D)12u(D)=\dfrac{c(D)-1}{2}, then u(γ)=u(D)u(\gamma)=u(D) and γ\gamma is a reduced alternating diagram of a (2,p)(2,\ p)- torus knot.

Let Γ\Gamma be a cycle of GG such that γ\gamma is obtained from f~(Γ)\tilde{f}(\Gamma). From the above f~(Γ)\tilde{f}(\Gamma) is a standard projection of a (2,p)(2,\ p)-torus knot as the case p=5p=5 is illustrated in the left of Fig. 33 and c(γ)=c(D)=pc(\gamma)=c(D)=p. If we can join two components of f~(Γ)\C(f~(G))\tilde{f}(\Gamma)\backslash C(\tilde{f}(G)) by a path PP of f~(G)\tilde{f}(G) as illustrated in the left of Fig. 35, then there exists a cycle γ\gamma^{\prime} of DD that has a crossing between γ\gamma and DγD-\gamma as illustrated in the right of Fig. 35. This is contradict to Lemma 5.2 (1)(1).

Refer to caption Refer to caption
Fig. 35.

Therefore we may assume that f~(G)\tilde{f}(G) has no paths as illustrated in the left of Fig. 35, namely f~(G)\tilde{f}(G) is a projection as illustrated in Fig. 36. Then by changing over/under informations at u(γ)u(\gamma) crossings on DD we can obtain a diagram of a trivial handlebody-knot. Since c(γ)=c(D),u(D)=c(D)12c(\gamma)=c(D),\ u(D)=\dfrac{c(D)-1}{2} and u(D)u(γ)u(D)\leq u(\gamma), we have u(γ)=c(γ)12u(\gamma)=\dfrac{c(\gamma)-1}{2}. By Theorem 1.5, γ\gamma is a reduced alternating diagram of a (2,p)(2,\ p)-torus knot for some odd number p±1p\neq\pm 1.

Refer to caption
Fig. 36.

Proof of Thm 1.7
Let HH be a nontrivial handlebody-knot that satisfies the equality u(H)=c(H)12u(H)=\dfrac{c(H)-1}{2}. Let DD be a minimal crossing diagram of HH. Since u(H)u(D)u(H)\leq u(D) and c(D)12=c(H)12\dfrac{c(D)-1}{2}=\dfrac{c(H)-1}{2} we have u(D)c(D)12u(D)\geq\dfrac{c(D)-1}{2}. Thus by the proof of Theorem 1.6 we have u(D)=c(D)12u(D)=\dfrac{c(D)-1}{2}. Then by Lemma 5.2 we see that HH is a handlebody-knot represented by one of diagrams illustrated in Fig. 10. We note that the unknotting number of handlebody-knot represented by D2n1(n0,1)D_{2n-1}(n\neq 0,1) are one by Proposition 3.3. Therefore HH is a handlebody-knot represented by D3D_{3} or D3D_{-3} as desired. ∎

Acknowledgements

The author would like to thank Professor Tomo Murao for his helpful comments. He is particularly grateful to Professor Kouki Taniyama for invaluable advice and his suggestions.

References

  • [1] Y. Akimoto and K. Taniyama, Unknotting numbers and crossing numbers of spatial embeddings of a planar graph, J. Knot Theory Ramifications, 29(14):2050095, 13, 2020.
  • [2] B. Dorothy and O. Danielle, Unknotting numbers for prime θ\theta-curves up to seven crossings, preprint. (arXiv:1710.05237).
  • [3] C. H. Dowker and M. B. Thistlethwaite, Classification of knot projections, Topology Appl., 16(1):19–31, 1983.
  • [4] R. Hanaki, Pseudo diagrams of knots, links and spatial graphs, Osaka J. Math., 47(3):863–883, 2010.
  • [5] A. Ishii, Moves and invariants for knotted handlebodies, Algebr. Geom. Topol., 8(3):1403–1418, 2008.
  • [6] M. Iwakiri, Unknotting numbers for handlebody-knots and Alexander quandle colorings, J. Knot Theory Ramifications, 24(14):1550059, 13, 2015.
  • [7] T. Kobayashi, A criterion for detecting inequivalent tunnels for a knot, Math. Proc. Cambridge Philos. Soc., 107(3):483–491, 1990.
  • [8] R. Nikkuni, M. Ozawa, K. Taniyama, and Y. Tsutsumi, Newly found forbidden graphs for trivializability, J. Knot Theory Ramifications, 14(4):523–538, 2005.
  • [9] I. Sugiura and S. Suzuki, On a class of trivializable graphs, Sci. Math., 3(2):193–200, 2000.
  • [10] N. Tamura, On an extension of trivializable graphs, J. Knot Theory Ramifications, 13(2):211–218, 2004.
  • [11] K. Taniyama, Knotted projections of planar graphs, Proc. Amer. Math. Soc., 123(11):3575–3579, 1995.
  • [12] K. Taniyama, Unknotting numbers of diagrams of a given nontrivial knot are unbounded, J. Knot Theory Ramifications, 18(8):1049–1063, 2009.