Regular Projections of Spatial Graphs

Ryo NIKKUNI
ABSTRACT Study of the regular projection in spatial graph theory is more rich in content as compared with the regular projection in knot theory. Actually some interesting phenomena in the regular projection of spatial graphs which are not appeared in the one of knots and links have been discovered Since 1990s. In this article we introduce the recent topics in the regular projection of spatial graphs, in particular, knotted projection (KP), identiable projection (IP) and completely distinguishable projection (CDP). We also present some unsolved problems in this area.

1. Introduction
1.1. KP, IP and CDP Let G be a nite graph. We give a label to each of vertices and edges of G and denote the set of all vertices and the set of all edges of a graph G by V (G) and E(G), respectively. An embedding of G into R3 is called a spatial embedding of G or simply a spatial graph. Two spatial embeddings f and g of G are said to be ambient isotopic if there exists an orientation preserving homeomorphism : R3 R3 such that f = g. A graph G is said to be planar if there exists an embedding of G into R2 , and a spatial embedding of a planar graph G is said to be trivial if it is ambient isotopic to an embedding of G into R2 R3 . We note that a trivial spatial embedding of a planar graph is unique up to ambient isotopy in R3 [14]. A regular projection of G is an immersion from G to R2 whose multiple points are only nitely many transversal double points away from vertices. For a regular projection f of G with p double points, we can obtain 2p regular diagrams of the spatial embeddings of G from f by giving over/under information to each of the double points. Then we say that a spatial embedding f of G is obtained from f if f is represented by one of these 2p regular diagrams. At this time we also call f a regular projection of f. Let f be a regular projection of a graph which is homeomorphic to the disjoint union of 1-spheres. Then it is well known that a trivial spatial embedding, namely a trivial link, can be always obtained from f . This fact plays an important role in knot theory, e.g. the theory of skein polynomial invariants is based on this fact. But this fact does not always hold for a regular projection of a planar graph. For example, let G be the octahedron graph and f a regular projection of G as illustrated in Fig. 1.1.1. Taniyama pointed out that any of the spatial embeddings of G obtained from f is non-trivial [27]. Such a regular projection is said to be knotted, namely a regular projection f of a planar
1

graph G is said to be knotted [27] if any of the spatial embeddings of G obtained from f is non-trivial. We call a knotted regular projection a KP simply. The existence of a KP is one of the suggestive phenomena in spatial graph theory. In the next section, we introduce some fundamental results and the recent development in the study of KP.
1

1 4 5 2 3 6

f
6 2

G
Fig. 1.1.1. A knotted regular projection

While a KP detects a non-triviality of a regular projection, let us consider a unifor mity of a regular projection. A regular projection f of G is said to be identiable [10] if all regular diagrams of the spatial embeddings of G obtained from f represent mutually ambient isotopic spatial embeddings of G. For example, each of the regular projections as illustrated in Fig. 1.1.2 (1), (2) and (3) is identiable. Actually we can see that only trivial spatial embeddings can be obtained from each of these regular projections. We call an identiable regular projection an IP simply. As the examples above suggest, an IP detects a triviality of a regular projection. Namely, an IP cannot be a KP. We explain the details in section 3.
1 1 2 4 3 2

(1)

(2)

(3)

Fig. 1.1.2. Identiable regular projections

On the other hand, a regular projection f of G is said to be completely distinguishable [17] if all regular diagrams of the spatial embeddings of G obtained from f represent mutually dierent spatial embeddings of G up to ambient isotopy. For example, let us recall a regular projection f of the octahedron graph as illustrated in Fig. 1.1.1. As we said before, f is a KP, namely we can obtain the eight spatial embeddings of G from
2

 f as illustrated in Fig. 1.1.3 and it can be checked that they are non-trivial spatial embeddings. Moreover, by observing the constituent knots and links carefully, we can see that these eight spatial embeddings are mutually not ambient isotopic. Thus f is completely distinguishable. We call a completely distinguishable regular projection a CDP simply. As the example above suggests, a CDP of a planar graph is related to KP. Moreover, a CDP of a non-planar graph can be also considered and relates to the minimal crossing number of a graph. In section 4, we introduce the recent results in the study of CDP.
1 5 1 5 1 5 1 5

6 2

6 2

6 2

6 2

1 5

1 5

1 5

1 5

6 2

6 2

6 2

6 2

Fig. 1.1.3.

2. Knotted projection (KP)

2.1. Trivializable graphs A planar graph is said to be trivializable if the graph does not have a KP. For example, a graph homeomorphic to the disjoint union of 1-spheres is trivializable, and the octahedron graph is not trivializable. Question 2.1.1. When is a planar graph trivializable? It is known that there exist innitely many trivializable graphs. Taniyama showed that every bifocal as illustrated on the left-hand side in Fig. 2.1.1 is trivializable [27]. On the other hand, Sugiura and Suzuki showed that every 3-line web as illustrated on the right-hand side in Fig. 2.1.1 is trivializable [23]. Tamura showed that every neo-bifocal is trivializable and gave a systematic construction of trivializable graphs in terms of an edge

sum of graphs [24].1 We refer the reader to [27], [24] and [23] for the precise denitions of the bifocal, the neo-bifocal and the 3-line web, respectively. But trivializable graphs have not been characterized completely yet.

Fig. 2.1.1. A bifocal and a 3-line web

2.2. Forbidden graphs for the trivializability We investigate trivializable graphs from a stand point of graph minor theory. A graph H is called a minor of a graph G if H can be obtained from G by a nite sequence of an edge contraction or taking a subgraph. We say that a property P of a graph is inherited by minors if a graph has P then each proper minor of the graph also has P. Let (P) be the set of all graphs which do not have P and whose all proper minors have P. This set is called the obstruction set for P and each of the elements in (P) is called a forbidden graph for P. It is clear that a graph G has P if and only if G does not have a minor which belongs to (P). Then, according to Robertson-Seymours Graph Minor Theorem [20], the following fact holds. Theorem 2.2.1. (Robertson-Seymour [20]) (P) is a nite set. In particular for the trivializability T , it is known the following. Proposition 2.2.2. (Taniyama [27]) The trivializability is inherited by minors. Therefore by Theorem 2.2.1, we have that (T ) is a nite set, namely the trivializability of a graph can be determined by nitely many forbidden graphs. Thus we want to determine all elements in (T ), namely we consider the following problem. Problem 2.2.3. Find all forbidden graphs for the trivializability.
1 Her method can produce a trivializable graph which is not a minor of a bifocal or neo-bifocal. But the author does not know whether her method can produce a trivializable graph which is not a minor of a 3-line web or not.

By the works of Sugiura-Suzuki and the author-Ozawa-Taniyama-Tsutsumi, sixteen forbidden graphs for the trivializability have been found as follows. Theorem 2.2.4. (1) (Sugiura-Suzuki [23]) The seven graphs G1 , G2 , . . . , G7 as illustrated in Fig. 2.2.1 belong to (T ). (2) (Nikkuni-Ozawa-Taniyama-Tsutsumi [18]) The nine graphs G8 , G9 , . . . , G16 as illustrated in Fig. 2.2.1 belong to (T ).

G1

G2

G3

G4

G5

G6

G7

G8

G9

G 10

G 11

G 12

G 13

G 14

G 15

G 16

Fig. 2.2.1. Forbidden graphs G1 , G2, . . . , G16 for the trivializability

Indeed, each Gi has a KP fi (i = 1, 2, . . . , 16) as illustrated in Fig. 2.2.2. Actually any of the spatial embeddings of Gi obtained from fi contains a Hopf link (i = 1, 2, . . . , 16).
5

Therefore each Gi is not trivializable. Moreover it can be seen that each of the proper minors of Gi is also a minor of a 3-line web. Thus by Proposition 2.2.2 we have that G1 , G2 , . . . , G16 (T ). Now, we show a new forbidden graph for the trivializability. Let G17 be a planar graph and f17 a regular projection of G17 as illustrated in Fig. 2.2.3. We can see easily that f17 is a KP, actually we can check that each of the spatial embeddings of G17 obtained from f17 also contains a Hopf link. Thus G17 is non-trivializable. Moreover we have the following. Theorem 2.2.5. The graph G17 belongs to (T ). The proof is the same as the proof for G1 , G2 , . . . , G16 , actually we can see that each of the proper minors of G17 is a minor of a 3-line web and G17 is not homeomorphic to G1 , G2 , . . . , G16 . So we omit the details. It seems that (T ) is not determined by these seventeen forbidden graphs. Actually we have candidates for forbidden graphs for the trivializability. Let H be the graph as illustrated in Fig. 2.2.4. Sugiura asked the following question in his master thesis [22]. Question 2.2.6. (Sugiura [22]) Is the graph H trivializable? This question is still open. Since we can see that each of the proper minors of H is trivializable, if H is not trivializable then H (T ). We remark here that for any regular projection f of H there exists a spatial embedding f of H obtained from f such that f does not contain any non-trivial link [18]. Therefore if H is non-trivializable then we can nd a spatial embedding f of H which is obtained from its KP and does not contain any non-trivial link. Besides, let H1 , H2 , H3 and H4 be four graphs as illustrated in Fig. 2.2.5. Then we have that each Hi has a KP gi (i = 1, 2, 3, 4) as illustrated in Fig. 2.2.6. Question 2.2.7. (Nikkuni-Ozawa-Taniyama-Tsutsumi [18]) Is each of Hi a forbidden graph for the trivializability? Note that each of Hi has a minor which is homeomorphic to H. Thus if H is not trivializable then H1 , H2 , H3 and H4 are not forbidden graphs for the trivializability, and if H is trivializable then there is a possibility that H1 , H2 , H3 , H4 (T ). 2.3. Spatial subgraph in a spatial graph obtained from a KP For the present, any spatial graph obtained from a known KP contains a Hopf link. Then the following question can be considered in general. Question 2.3.1. What kind of spatial subgraph does appear in a spatial graph obtained from a KP?
6

f1

f2

f3

f4

f5

f6

f7

f8

f9

f 10

f 11

f 12

f 13

f 14

f 15

f 16

Fig. 2.2.2. A KP fi : Gi R2 (i = 1, 2, . . ., 17)

4 7

6 8 5 1

f17
4

8 6 1

7 2

G17

Fig. 2.2.3. New forbidden graph and its KP

Fig. 2.2.4.

H1

H2

H3

H4

Fig. 2.2.5.

g1

g2

g3

g4

Fig. 2.2.6. A KP gi : Hi R2 (i = 1, 2, 3, 4)

As a spacial case of Question 2.3.1, the following question was asked in [18]. Question 2.3.2. (Nikkuni-Ozawa-Taniyama-Tsutsumi [18]) Let f be a KP of a planar graph G and f a spatial embedding of G obtained from f . Does there exist a proper subgraph H of G such that f|H is a non-trivial spatial embedding? We remark here that Question 2.3.2 is equivalent to the following question. Question 2.3.3. (Nikkuni-Ozawa-Taniyama-Tsutsumi [18]) Is a regular projection of a minimally knotted spatial embedding of a planar graph not a KP? Here a spatial embedding f of a planar graph G is said to be minimally knotted if f is non-trivial and f|H is trivial for any proper subgraph H of G. For example, each of the spatial graphs f, g and h as illustrated in Fig. 2.3.1 is known as a minimally knotted spatial graph. In the following we give a partial answer for Question 2.3.3. A spatial embedding f of a planar graph G is said to be strongly almost trivial [7] if f is non-trivial and there exists a regular projection f of f such that any of the spatial embeddings of H obtained from f |H is trivial for any proper subgraph H of G. At this time we also call f a strongly almost trivial projection of G. It is clear that if f is strongly almost trivial then it is minimally knotted. For example, each of the spatial graphs f and g as illustrated in Fig. 2.3.1 is strongly almost trivial, actually they have f and g as strongly almost trivial projections, respectively. But a spatial graph h as illustrated in Fig. 2.3.1 is not strongly almost trivial [9]. Then we have the following. Proposition 2.3.4. A strongly almost trivial projection of a planar graph is not a KP. We prepare some denitions and lemmas that are needed to prove Proposition 2.3.4 (We also need them in the next section). A spatial graph is said to be free if the fundamental group of the spatial graph complement is a free group. Then the following is ScharlemannThompsons famous criterion. Lemma 2.3.5. (Scharlemann-Thompson [21]) A spatial embedding of a planar graph is trivial if and only if any of the spatial subgraphs is free. On the other hand, the following is known. Lemma 2.3.6. (Nikkuni-Ozawa-Taniyama-Tsutsumi [18]) Let f be a regular projection of a graph (does not need to be planar). Then there exists a free spatial embedding of the graph obtained from f. Now we give a proof of Proposition 2.3.4. The proof is essentially the same as the proof of [18, Proposition 3.4]. Let f be a strongly almost trivial spatial embedding of a planar
9

Fig. 2.3.1.

graph G and f the strongly almost trivial projection. Since f is minimally knotted, by Lemma 2.3.5 we have that f is not free. On the other hand, by Lemma 2.3.6, there exists a free spatial embedding g of G obtained from f. Then, by the denition of the strongly almost triviality, we have that g|H is trivial for any proper subgraph H of G, namely g|H is free for any subgraph H of G. Thus by Lemma 2.3.5 we have that g is trivial. So f is not a KP.

3. Identiable projection (IP)

3.1. IP and the Simon invariant When we consider an IP, the following fact is fundamental. Proposition 3.1.1. (Huh-Taniyama [10]) A non-planar graph does not have an IP. To explain Proposition 3.1.1, we recall the Wu invariant [30] of a spatial graph. Let X be a topological space and C2(X) = {(x, y) X X | x = y} the conguration space of ordered two points of X. Let be an involution on C2 (X) dened by (x, y) = (y, x). Then we call the integral cohomology group of Ker(1 + # ) the skew-symmetric integral cohomology group of the pair (C2 (X), ) and denote it by H (C2 (X), ). It is known that H 2 (C2 (R3), ) Z [30]. We denote a generator of =
10

H 2 (C2 (R3), ) by . Let f be a spatial embedding of a graph G. Then f induces a homomorphism (f 2 ) : H 2 (C2 (R3 ), ) H 2 (C2(G), ). We call (f 2 ) () the Wu invariant of f and denote it by L(f). In particular, if G is homeomorphic to K5 or K3,3 then H 2 (C2 (G), ) Z [28] and L(f) coincides with the = Simon invariant [26]. We note that L(f) coincides with twice the linking number if G is homeomorphic to the disjoint union of two 1-spheres. We refer the reader to [28] for a diagramatic calculation of L(f). The generator of H 2 (C2 (R3 ), ) depends on the orientation of R3 [30], namely it holds that L(f!) = L(f) for a spatial embedding f of G, where f! denotes the mirror image of f. Since the Simon invariant is odd integer-valued [26], we have that any spatial embedding of K5 and K3,3 is not ambient isotopic to its mirror image. This implies that any spatial embedding of a non-planar graph G is not ambient isotopic to its mirror image because G always contains a subgraph which is homeomorphic to K5 or K3,3 [13]. Therefore a non-planar graph cannot have an IP. 3.2. Characterizing spatial graphs obtained from an IP Now let us consider an IP f of a planar graph G. If G is trivialzable, then a trivial spatial embedding of G can be always obtained from f . Hence the following is clear by the denition of the identiability and the uniqueness of a trivial spatial embedding. Proposition 3.2.1. Let G be a trivializable planar graph and f a regular projection of G. Then, f is an IP if and only if any of the spatial embeddings obtained from the projection is trivial. By the existence of a KP, the proof of Proposition 3.2.1 does not work for nontrivializable graphs as it is. But by an application of Lemmas 2.3.5 and 2.3.6, we can generalize Proposition 3.2.1 to arbitrary planar graphs as follows. Theorem 3.2.2. (Nikkuni [16]) A regular projection of a planar graph is an IP if and only if any of the spatial embeddings obtained from the projection is trivial. Actually, we have the if part by the uniqueness of the trivial spatial embeddings of a planar graph. And we have the only if part as follows. Let f be an IP of a planar graph G and f the spatial embedding of G obtained from f. We note that f|H is also an IP for any subgraph H of G. Then by Lemma 2.3.6 the spatial embedding g of H obtained from f|H is free. Since g = f|H , we have that f|H is free for any subgraph H of G. Therefore by Lemma 2.3.5 we have that f is trivial. This completes the proof of Theorem 3.2.2.
11

A regular projection of a graph is said to be reduced if the image of the projection has no local parts as illustrated in Fig. 3.2.1 (a) and (b). For example, each of the regular projections as illustrated in Fig. 1.1.2 (2) and (3) is reduced. Note that there exist innitely many planar graphs whose reduced IP are only embeddings from the graph to R2 , see [10] (We also refer the reader to [3], [19], [25], [11], [31] and [8] for related works). Theorem 3.2.2 indicates that the spatial embedding of a planar graph obtained from an IP of the graph is always trivial even if the projection is reduced and not an embedding into R2 .

(a)

(b)

Fig. 3.2.1.

4. Completely distinguishable projection (CDP)

4.1. CDP of a planar graph We recall that the knottedness of a regular projection is dened for only planar graphs, and the identiability of a regular projection is considered for only planar graphs as a consequence of Proposition 3.1.1. But the complete distinguishability of a regular projection can be considered not only for planar graphs but also for non-planar graphs. First we deal with the case of planar graphs in this subsection. A CDP of a graph is said to be trivial if it has no double points. Therefore every planar graph has a trivial CDP. Moreover, a non-trivial CDP of a planar graph is closely related to KP as follows. Proposition 4.1.1. (1) Let G be a trivializable planar graph. Then a regular projection f of G is a CDP if and only if it has no double points. (2) Let G be a non-trivializable planar graph. If a regular projection f of G is a non-trivial CDP then f is a KP. Actually it is clear that any spatial embedding of a graph which can be obtained from a non-trivial completely distinguishable projection of the graph is not ambient isotopic to its mirror image. By this fact, we can prove Proposition 4.1.1. We note that the converse of Proposition 4.1.1 (2) is not true, see Examples 4.1.2 and 4.1.3. Example 4.1.2. Let G be a non-trivializable planar graph and f a CDP of G. Note that f is also a KP by Proposition 4.1.1 (2). Then by producing the local parts in f(G)
12

as illustrated in Fig. 4.1.1, we can construct another KP which is not a CDP.

Fig. 4.1.1.

Example 4.1.3. Let G be a planar graph and f a regular projection of G as illustrated in Fig. 4.1.2. Since f (G) contains a image of a KP of the octahedron graph as illustrated in Fig 1.1.1, we have that f is a KP. We can also check that f (G) does not contain any local parts as in Fig. 4.1.1. But Fig. 4.1.3 shows that f is not a CDP.
9 1 9 4 10 7 12 11 5 8 6 4 1 6 11 12 3 2 7 5 10 3 8

Fig. 4.1.2.

1 9 1 4 11 12 2 7 5 10 3 8 2 5 7 12 6 4 11 10 8 3 5 2 7 12 6 4 11 10 3 8 9 9 1 6

Fig. 4.1.3.

13

4.2. CDP of a non-planar graph In this subsection we deal with the case of non-planar graphs. Let Kn be the complete graph on n vertices and Km,n the complete bipartite graph on m + n vertices. Note that Kn is non-planar if n 5, and Km,n is non-planar if min{m, n} 3. Theorem 4.2.1. (Nikkuni [17]) Each of Kn and Km,n has a CDP. We note that each of Kn for n 4 and Km,n for min{m, n} 2 is planar and trivializable [23]. So in this case only embeddings from the graph into R2 are CDP by Proposition 4.1.1 (1). In the following we show that Kn has a CDP. Since Kn has a trivial CDP for n 4, we assume that n 5. We set V (Kn ) = {1, 2, . . . , n}. We denote the edge of Kn connecting two vertices i and j by ei,j (1 i < j n). Let Wn = Cn1 st(n) be a subgraph of Kn embedded in R2 as illustrated in Fig. 4.2.1, where Cn1 = e1,2 e2,3 e1,n1 and st(n) = n1 ei,n . Then Cn1 divides R2 into two domains and we can construct a regular i=1 projection f : Kn R2 by embedding each of the edges in E(Kn ) E(Wn ) into the noncompact domain such that any pair of the edges ei,j and ek,l (1 i < k < j < l n 1) has exactly one double point and the other pair of the edges has no double points, see Fig. 4.2.1. Then, for the double point P between ei,j and ek,l (1 i < k < j < l n 1), we can nd a subgraph H = Cn1 ei,n ek,n ej,n el,n ei,j ek,l of Kn which is homeomorphic to K5 such that f |H is a regular projection of H with exactly one double point P . Let g and g! be exactly two spatial embeddings of H which can be obtained from f |H . Then we have that |L(g)| = |L(g!)| = 1. Since L(g!) = L(g), we have that L(g) = L(g!). This implies that f is a CDP.

1 1 5 6 2 3 3 4 4 2 5 2 3 4 1 7 5 6

Fig. 4.2.1. A CDP f : Kn R2 (n = 5, 6, 7)

We remark here that the construction above is based on a complete calculation of the Wu invariants of spatial embeddings of Kn by the author, see [15] for the details. Conjecture 4.2.2. Every graph has a CDP.
14

Since a planar graph always has a trivial CDP, our aim is to prove whether a nonplanar graph always has a CDP. For a planar graph, it is also natural to ask the following question. Question 4.2.3. Does a non-trivializable planar graph have a non-trivial CDP? 4.3. CDP and the minimal crossing projection Let cr(f ) be the number of double points of a regular projection f : G R2 . Then we call cr(G) = min{cr(f) | f : G R2 is a regular projection} a minimal crossing number of G. A regular projection f of G is called a minimal crossing projection if cr(f) = cr(G). For example, cr(K5 ) = cr(K3,3 ) = 1, cr(K3,4 ) = 2 and cr(K6 ) = 3 (cf. [6]). Fig. 4.3.1 shows an example of the minimal crossing projections of K5 , K3,3 , K3,4 and K6 .

Fig. 4.3.1. Minimal crossing projections

Let f be a CDP of Kn (resp. Km,n ) constructed by the method in the proof of Theorem 4.2.1. Then it can be easily seen that f is not a minimal crossing projection if n 6 (resp. min{m, n} 3 and max{m, n} 4). On the other hand, we can check that each of the minimal crossing projections as illustrated in Fig. 4.3.1 is a CDP in a similar way as the proof of Theorem 4.2.1. From this viewpoint, Ozawa suggested the following question. Question 4.3.1. ([17]) Is a minimal crossing projection a CDP? Note that a minimal crossing projection of a planar graph is an embedding of the graph into R2 , namely a trivial CDP. Besides we can give an armative answer to Question 4.3.1 for the case of minimal crossing number one [17]. We present a piece of circumstantial evidence as follows. Example 4.3.2. Let f be the regular projection of K2n , which is Blaek-Komans z construction of a candidate for the minimal crossing projection of K2n [1]. Fig. 4.3.2 illustrates the case of n = 4 and in this case f is a minimal crossing projection of K8 .2
2

R. K. Guy conjectured in [4] that cr(Kn ) = 1 n 4 2 n1 2 n2 2 n3 2

15

 Then we can see that f is a CDP in a similar way as the proof of Theorem 4.2.1.

1 5

4 8

1 5 8 7

6 7 2 3 2

Fig. 4.3.2. f : K2n R2 (n = 4)

Example 4.3.3. Let f be the regular projection of Km,n , which is Zarankiewiczs construction of a candidate for the minimal crossing projection of Km,n [32]. Fig. 4.3.3 illustrates the case of m = n = 6 and in this case f is a minimal crossing projection of K6,6 .3 Then we can also see that f is a CDP in a similar way as the proof of Theorem 4.2.1.
1 3 5 2 4 6 7 9 8 10 12 2 4 6 1 3 5 7 9 8 10 12

11

11

Fig. 4.3.3. f : Km,n R2 (m = n = 6)

Conjecture 4.3.4. A minimal crossing projection is a CDP. We remark here that if Conjecture 4.3.4 is solved armatively then Conjecture 4.2.2 is also solved armatively.
and showed that the conjecture above is true for n 10 [5]. 3 From the beginning, the crossing number problem for K a m,n is called Turns brick factory problem (cf. [29]). In [32], K. Zarankiewicz claimed that m m1 n n1 cr(Km,n ) = . 2 2 2 2 However Guy pointed out a gap for Zarankiewiczs proof [4], and now this is called Zarankiewiczs conjecture. At present, Zarankiewiczs conjecture is true for m 6 [12].

16

References
[1] J. Blaek and M. Koman, A minimal problem concerning complete plane graphs, Theory of Graphs z and its Applications (Proc. Sympos. Smolenice, 1963) pp. 113117 Publ. House Czechoslovak Acad. Sci., Prague. [2] J. A. Bondy and U. S. R. Murty, Graph Theory with Applications, American Elsevier Publishing Co., Inc., New York, 1976. [3] T. D. Cochran and R. E. Gompf, Applications of Donaldsons theorems to classical knot concordance, homology 3-spheres and property P , Topology 27 (1988), 495512. [4] R. K. Guy, The decline and fall of Zarankiewiczs theorem, Proof Techniques in Graph Theory (Proc. Second Ann Arbor Graph Theory Conf., Ann Arbor, Mich., 1968) pp. 6369 Academic Press, New York. [5] R. K. Guy, Crossing numbers of graphs, Graph theory and applications (Proc. Conf., Western Michigan Univ., Kalamazoo, Mich., 1972; dedicated to the memory of J. W. T. Youngs), pp. 111124. Lecture Notes in Math., Vol. 303, Springer, Berlin, 1972. [6] N. Hartseld and G. Ringel, Pearls in graph theory, A comprehensive introduction. Revised reprint of the 1990 original. Academic Press, Inc., Boston, MA, 1994. x+249 pp. [7] Y. Huh, G. T. Jin and S. Oh, Strongly almost trivial -curves, J. Knot Theory Ramications 11 (2002), 153164. [8] Y. Huh, G. T. Jin and S. Oh, An elementary set for n -curve projections, J. Knot Theory Ramications 11 (2002), 12431250. [9] Y. Huh and S. Oh, Planar graphs producing no strongly almost trivial embedding, J. Graph Theory 43 (2003), 319326. [10] Y. Huh and K. Taniyama, Identiable projections of spatial graphs, J. Knot Theory Ramications 13 (2004), 991-998 [11] S. Kinoshita and J. Mikasa, On projections of spatial theta-curves, Kwansei Gakuin University (in Japanese), 1993. [12] D. J. Kleitman, The crossing number of K5,n , J. Combinatorial Theory 9 (1970), 315323. [13] C. Kuratowski, Sur le problme des courbes gauches en topologie, (French) Fund. Math. 15 (1930), e 271283. [14] W. K. Mason, Homeomorphic continuous curves in 2-space are isotopic in 3-space, Trans. Amer. Math. Soc. 142 (1969), 269290. [15] R. Nikkuni, The second skew-symmetric cohomology group and spatial embeddings of graphs, J. Knot Theory Ramications 9 (2000), 387411. [16] R. Nikkuni, A remark on the identiable projections of planar graphs, Kobe J. Math. 22 (2005), 65-70.

17

[17] R. Nikkuni, Completely distinguishable projections of spatial graphs, J. Knot Theory Ramications 15 (2006), 11-19. [18] R. Nikkuni, M. Ozawa, K. Taniyama and Y. Tsutsumi, Newly found forbidden graphs for trivializability, J. Knot Theory Ramications 14 (2005), 523-538. [19] J. H. Przytycki, Positive knots have negative signature, Bull. Polish Acad. Sci. Math. 37 (1989), 559562. [20] N. Robertson and P. Seymour, Graph minors XVI. Wagners conjecture, preprint. [21] M. Scharlemann and A. Thompson, Detecting unknotted graphs in 3-space, J. Di. Geom. 34 (1991), 539560. [22] I. Sugiura, On trivializability of spatial graph projections, Master thesis, Waseda University (1997). [23] I. Sugiura and S. Suzuki, On a class of trivializable graphs, Sci. Math. 3 (2000), 193200. [24] N. Tamura, On an extension of trivializable graphs, J. Knot Theory Ramications 13 (2004), 211218. [25] K. Taniyama, A partial order of knots, Tokyo J. Math. 12 (1989), 205229. [26] K. Taniyama, Cobordism, homotopy and homology of graphs in R3 . Topology 33 (1994), 509523. [27] K. Taniyama, Knotted projections of planar graphs, Proc. Amer. Math. Soc. 123 (1995), 3575 3579. [28] K. Taniyama, Homology classication of spatial embeddings of a graph, Topology Appl. 65 (1995), 205228. [29] P. Turn, A note of welcome, J. Graph Theory 1 (1977), 79. a [30] W. T. Wu, On the isotopy of a complex in a Euclidean space. I, Sci. Sinica 9 (1960), 2146. [31] C. Yoshioka: Regular projections of knotted handcu graphs and knotted bouquets, Master thesis, Tokyo Womans Christian University, 1996. [32] K. Zarankiewicz, On a problem of P. Turn concerning graphs, Fund. Math. 41, (1954). 137145. a

Department of Mathematics, Faculty of Education, Kanazawa University Kakuma-machi, Kanazawa, 920-1192, Japan nick@ed.kanazawa-u.ac.jp

18