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

Strong and weak (1, 2) homotopies on knot projections and new invariants

Noboru Ito  and  Yusuke Takimura Waseda Institute for Advanced Study, 1-6-1 Nishi-Waseda Shinjuku-ku Tokyo 169-8050 Japan [email protected] (Current address) Graduate School of Mathematical Sciences, The University of Tokyo, 3-8-1 Komaba Meguro-ku Tokyo 153-8914 Japan [email protected] Gakushuin Boy’s Junior High School, 1-5-1 Mejiro Toshima-ku Tokyo 171-0031 Japan [email protected]
Abstract.

Every second flat Reidemeister move of knot projections can be decomposed into two types thorough an inverse or direct self-tangency modification, respectively called strong or weak, when orientations of the knot projections are arbitrarily provided. Further, we introduce the notions of strong and weak (1, 2) homotopies; we define that two knot projections are strongly (resp. weakly) (1, 2) homotopic if and only if two knot projections are related by a finite sequence of first and strong (resp. weak) second flat Reidemeister moves. This paper gives a new necessary and sufficient condition that two knot projections are not strongly (1, 2) homotopic. Similarly, we obtain a new necessary and sufficient condition in the weak (1, 2) homotopy case. We also define a new integer-valued strong (1, 2) homotopy invariant. Using it, we show that the set of the non-trivial prime knot projections without 11-gons that can be trivialized under strong (1, 2) homotopy is disjoint from that of weak (1, 2) homotopy. We also investigate topological properties of the new invariant and give its generalization, a comparison of our invariants and Arnold invariants, and a table of invariants.

Key words and phrases:
knot projection; spherical curve; strong (1, 2) homotopy; weak (1, 2) homotopy; non-Seifert resolution
MSC2010: 57M25; 57Q35.

1. Introduction

A knot projection is defined as the image of a generic immersion of a circle into a 22-dimensional sphere. Historically, the equivalence classes of knot projections generated by the first and second flat Reidemeister moves shown in Fig. 1 are determined by Khovanov [7, Theorem 2.2] via the notion of doodles, introduced by Fenn and Taylor [3, 2] (cf. [5, Theorem 2.2]). Essentially, by using (1, 2)-reduced knot projections, which are knot projections without any 11- or 22-gons, we can detect whether two given knot projections are equivalent under the equivalence relations obtained by the first and the second flat Reidemeister moves. Here, a 11-gon (resp. 22-gon) is the boundary of a disk with exactly one (resp. two) vertex and exactly one (resp. two) edges, where a knot projection consists of vertexes and edges (Fig. 1).

Refer to caption
Figure 1. The first and second flat Reidemeister moves for knot projections.

Nonetheless, to the best of our knowledge, there have still been two open problems: find a nice necessary and sufficient condition for two knot projections being equivalent under a given equivalence relation, called strong (1,2)(1,2) homotopy (resp. weak (1,2)(1,2) homotopy), consisting of the first flat Reidemeister moves denoted by RIs and the strong RI​Is (resp.  weak RI​Is ) as defined by Fig. 2  (center) (resp. Fig. 2  (right)). The equivalence induced by RI is denoted by 1\stackrel{{\scriptstyle 1}}{{\sim}}. The equivalence induced by strong RI​I (resp. weak RI​I) is denoted by s2\stackrel{{\scriptstyle s2}}{{\sim}} (resp. w2\stackrel{{\scriptstyle w2}}{{\sim}}). The 22-gon appearing in a strong (resp. weak) RI​I is called a coherent ( resp. incoherent ) 22-gon if the 22-gon is (resp. is not) be oriented by orienting a knot projection.

Refer to caption
Figure 2. Strong RI​I and weak RI​I. Dotted arcs stand for the connection of the branches of two double points, which we focused on.

In this paper, let s\sim_{s} (resp. w\sim_{w}) be strong (resp. weak) (1, 2) homotopy and let \simeq be sphere isotopy (often simply called isotopy). The definition of the strong or weak RI​I is a natural notion as a second flat Reidemeister move (Fig. 1, right) is decomposed into just the two kinds of local moves shown in Fig. 2; one is strong RI​I, and the other is a weak RI​I. The strong (resp. weak) RI​I is also treated as the inverse (resp. direct) self-tangency perestroika as in Arnold [1] if the knot projections are oriented. In the rest of this paper, a knot projection having no double points is simply called a simple closed curve or a trivial knot projection.

The answers to the two aforementioned problems appear in this paper (Theorem 1), serving as the starting point.

Theorem 1.

Let a knot projection with no 11-gons and no incoherent (resp. coherent) 22-gons be called a weak (resp. strong) reduced knot projection PiwrP_{i}^{wr} (resp. PisrP_{i}^{sr}), only decreasing double points by any RIs or weak (resp. strong) RI​Is from a knot projection PiP_{i}.

(1)(1) P1P_{1} and P2P_{2} are weakly (1, 2) homotopic if and only if P1wrP_{1}^{wr} and P2wrP_{2}^{wr} are isotopic.

(2)(2) P1P_{1} and P2P_{2} are strongly (1, 2) homotopic if and only if P1srP_{1}^{sr} and P2srP_{2}^{sr} are isotopic.

Here, the next problem arises: how do the two equivalence relations, weak and strong (1, 2) homotopies, classify knot projections? This paper obtains a partial answer to that problem.

Theorem 2.

Let OO be a knot projection having no double points, namely a simple circle on the 22-sphere. Then a knot projection PP is strongly (1, 2) homotopic and weakly (1, 2) homotopic to OO if and only if PP and OO are transformed into each other by RIs and isotopies.

To prove Theorem 2, this paper introduces a topological invariant of knot projections obtained by applying the replacement shown in Fig. 3 at every double point. The replacements do not depend on an orientation of a knot projection and are denoted as “A1A^{-1}” in [4]. We call replacement A1A^{-1}, or non-Seifert resolution.

Refer to caption
Figure 3. The local replacement around a double point (A1A^{-1}).

One may feel that this local replacement is similar to Seifert resolution (Fig. 4).

Refer to caption
Figure 4. Seifert resolution.

Seifert resolution has many crucial roles in the basics of today’s Knot Theory; for instance, genus, Alexander polynomial, or Khovanov-Lee homology. It also has an important role in the theory of generic immersed plane curves: e.g., it provides the Alexander numbering and Arnold invariants of plane curves [10, 8]. In these roles, one of the advantages of Seifert resolution is that it preserves the orientation of an oriented curve. In comparison, A1A^{-1} does not preserve the orientation if the curve is oriented, and thus one may think that A1A^{-1} would not have such an important role.

However, this paper shows that A1A^{-1} has nice properties and applications. One of the advantages of A1A^{-1} is that for a non-oriented curve, A1A^{-1} does not change under RI, while each Seifert resolution changes the number of components under RI. In fact, A1A^{-1} gives invariants of knot projections under both RI and strong RI​I as shown in Fig. 2.

For example, let us consider our familiar object, the chord diagram (often called Gauss diagram), which is one circle with finitely many chords where each chord connects the preimages of each double point of a knot projection. The number X(P)X(P), introduced by [6], is defined as the number of sub-chords as “\otimes” embedded in the whole chord diagram of a given knot projection PP; X(P)X(P) modulo 22 is then an invariant under strong (1, 2) homotopy, though it is a /2\mathbb{Z}/2\mathbb{Z}-valued invariant.

An easily calculated \mathbb{Z}-valued strong (1, 2) homotopy invariant, called circle number, is introduced in this paper using non-Seifert resolution A1A^{-1}. We also give a generalization of the \mathbb{Z}-valued invariant, called circle arrangement, under strong (1, 2) homotopy and obtain its table of prime knot projections with small numbers of double points. We call these new invariants as a whole circle invariants.

This paper is constructed as follows. Sec. 2 proves Theorem 1. Sec. 3 defines a new \mathbb{Z}-valued invariant, circle numbers, under strong (1, 2) homotopy and also gives a new, strictly stronger invariant, circle arrangements. Sec. 4 explores the characteristics of circle numbers. Sec. 5 proves Theorem 2. Sec. 6 explores the characteristics of circle arrangements. Finally, Sec. 7 obtains a table of prime knot projections with small numbers of double points including information on circle numbers and circle arrangements.

2. Proof of Theorem 1

We need two lemmas which are needed to prove Theorem 1.

Lemma 1.

A move RI increasing (resp. decreasing) double points is denoted by 1a1a (resp. 1b1b). Weak RI​I increasing (resp. decreasing) double points is denoted by w2aw2a (resp. w2bw2b). Any finite sequence generated by RIs and weak RI​Is from a weak reduced knot projection to a knot projection can be replaced with a sequence of moves only of types 1a1a and w2aw2a.

Lemma 2.

A move RI increasing (resp. decreasing) double points is denoted by 1a1a (resp. 1b1b). Strong RI​I increasing (resp. decreasing) double points is denoted by s2as2a (resp. s2bs2b). Any finite sequence generated by RIs and strong RI​Is from a strong reduced knot projection to a knot projection can be replaced with a sequence of moves only of types 1a1a and s2as2a.

Proofs of Lemma 1 and Lemma 2. We can prove Lemmas 1 and 2 by restricting the general second flat Reidemeister move into either a weak RI​I (for claim (1)) or a strong RI​I (for claim (2)) in [5, Proof of Theorem 2.2 (c)]. The strategy of the two proofs of (1) and (2) shows up in Table 1. In Table 1, “same” means the same argument as that of [5, Theorem 2.2 (c)], \emptyset means that it cannot occur, and the term “restricted into strong RI​I (resp. weak RI​I)” means that the same argument is used, though considering the second flat Reidemeister move restricted to only a strong RI​I (resp.  weak RI​I). The proof [5, Theorem 2.2 (c)] consists of Case 1–Case 4 by the last two moves when, in the sequence of the first and second flat Reidemeister moves, the first appearance of moves decreasing the number of double points occurs. Each of Case 1–Case 4 has that (i): the last two moves are close and (ii): the last two moves are apart from one another. Thus, using the proof of [5, Theorem 2.2 (c)], we have proofs of Lemmas 1 and 2. \Box

Weak (1, 2) Strong (1, 2)
Case 1-(i) same same
Case 1-(ii) same same
Case 2-(i) \emptyset same
Case 2-(ii) restricted into weak RI​I restricted into strong RI​I
Case 3-(i) \emptyset same
Case 3-(ii) restricted into weak RI​I restricted into strong RI​I
Case 4-(i) restricted into weak RI​I restricted into strong RI​I
Case 4-(ii) restricted into weak RI​I restricted into strong RI​I
Table 1. Strategy to prove Lemmas 1 and 2.

Proof of Theorem 1. Assume that P1wr≄P2wrP_{1}^{wr}\not\simeq P_{2}^{wr}. If P1wP2P_{1}\sim_{w}P_{2}, then P1wrwP2wrP_{1}^{wr}\sim_{w}P_{2}^{wr}. By Lemma 1, there exists a finite sequence of moves only of types 1a1a and w2aw2a from P1wrP_{1}^{wr} to P2wrP_{2}^{wr}. Thus P2wrP_{2}^{wr} has at least one 11- or incoherent 22-gon, which contradicts the definition of P2wrP_{2}^{wr}. Therefore, if P1wP2P_{1}\sim_{w}P_{2}, P1wrP2wrP_{1}^{wr}\simeq P_{2}^{wr}. If P1wrP2wrP_{1}^{wr}\simeq P_{2}^{wr}, by the definitions of P1wrP_{1}^{wr} and P2wrP_{2}^{wr}, P1wP2P_{1}\sim_{w}P_{2}. This completes the proof of claim (1).

For claim (2) of Theorem 1, a very similar proof to the above is established thorough appropriate replacements (e.g., wsw\to s, “incoherent” \to “coherent”). That complete the proof of Theorem 1. \Box

Corollary 1.

For an arbitrary knot projection PP, there exists a unique knot projection PwrP^{wr} realizing the minimal number of double points up to weak (1, 2) homotopy.

Corollary 2.

For an arbitrary knot projection PP, there exists a unique knot projection PsrP^{sr} realizing the minimum number of double points up to strong (1, 2) homotopy.

3. Definition of circle invariants

When an oriented knot projection is given, let us consider the local replacement of Fig. 3 from the left figure to the right figure for every double point. By this definition, this local replacement does not depend on an orientation of a knot projection. This local replacement is introduced by [4] for knot projections and denoted by A1A^{-1} following [4]. In this paper, we also call the local replacement A1A^{-1} non-Seifert resolution.

Definition 1 (circle arrangements and circle numbers).

For a knot projection PP, we apply A1A^{-1} at every double point, and then we have a collection of circles having no double point on the sphere. The collection of circles resulting from a knot projection PP is denoted by τ(P)\tau(P). Let us call this τ(P)\tau(P) circle arrangements (e.g. Fig. 5). The number of circles in τ(P)\tau(P) is called a circle number and denoted by |τ(P)||\tau(P)| (e.g.  |τ(P)||\tau(P)| == 33 in Fig. 5). All together, the notions of circle numbers and circle arrangements are called circle invariants. Note that circle invariants can be defined on both oriented and unoriented knot projections.

Refer to caption
Figure 5. (a) a knot projection PP, (b) τ(P)\tau(P), (c) the same τ(P)\tau(P) as (b) under sphere isotopy.
Theorem 3.

Let PP be an arbitrary knot projection. τ(P)\tau(P) is invariant under strong (1, 2) homotopy.

Proof.

Fig. 6 proves the claim.

Refer to caption
Figure 6. Non-Seifert resolution A1A^{-1} for RI and strong RI​I.

Corollary 3.

Let PP be an arbitrary knot projection. |τ(P)||\tau(P)| is invariant under strong (1, 2) homotopy.

Proposition 1.

Let PP be an arbitrary knot projection. τ(P)\tau(P) is strictly stronger than |τ(P)||\tau(P)|.

Proof.

It is easy to see the canonical map τ(P)|τ(P)|\tau(P)\mapsto|\tau(P)| by counting the number of circles in τ(P)\tau(P). In Fig. 20, |τ(31)||\tau(3_{1})| == |τ(63)||\tau(6_{3})| == 33 but τ(31)\tau(3_{1}) \neq τ(63)\tau(6_{3}). ∎

Remark 1.

There exists two knot projections with the same circle arrangement which are not strongly (1, 2) homotopic. See 7C7_{C} and 757_{5} (or 737_{3}).

4. Properties of circle numbers

The connected sum of two knot projections PP and PP^{\prime} are defined by Fig. 7.

Refer to caption
Figure 7. Connected sum of two knot projections PP and PP^{\prime}.
Theorem 4.

Let PP be an arbitrary knot projection. |τ(P)||\tau(P)| has the following properties.

  1. (1)

    |τ(P)||\tau(P)| is an odd integer.

  2. (2)

    |τ(PP)||\tau(P\sharp P^{\prime})| == |τ(P)||\tau(P)| ++ |τ(P)||\tau(P^{\prime})| - 11 where PPP\sharp P^{\prime} is a connected sum of PP and PP^{\prime}.

  3. (3)

    For a knot projection PP, we give PP an arbitrary orientation. If PP contains an element of the following list (Fig. 8) as a sub-diagram, |τ(P)|3|\tau(P)|\geq 3.

    Refer to caption
    Figure 8. List of 2n2n-gons.
Proof.
  1. (1)

    By the definition, |τ(O)||\tau(O)| is 11. It is well known that an arbitrary knot projection and the simple circle OO can be related by a finite sequence of the first, second, and third flat Reidemeister moves, where the first and second flat Reidemeister moves are defined by Fig. 1 and the third flat Reidemeister move is defined by Fig. 9.

    Refer to caption
    Figure 9. The third flat Reidemeister move.

    We have already shown that |τ(P)||\tau(P)| is invariant under RI and strong RI​I. Thus, it is sufficient to show that |τ(P)||\tau(P)| \equiv 0 (mod 22) under weak RI​I and the third flat Reidemeister move.

    First, let us consider weak RI​I by looking at Fig. 10. For a knot projection PP with an orientation (the upper-left of Fig. 10), we have τ(P)\tau(P) appearing as in the left figure of the second row. We consider the possibilities of the connections of arcs of finitely many simple circles which we focus on. We can draw them as locally two types of τ(P)\tau(P), as in the upper right of the first and second rows in Fig. 10. These two figures imply that |τ(P)||τ(P)||\tau(P)|-|\tau(P^{\prime})| is either ±2\pm 2 or 0 if PP and PP^{\prime} is related by a single weak RI​I.

    Refer to caption
    Figure 10. The symbol \Leftrightarrow in the figure (upper part) stands for weak RI​I. In the lower part of the figure, the symbol w\stackrel{{\scriptstyle w}}{{\sim}} (resp. s\stackrel{{\scriptstyle s}}{{\sim}}) represents the weak (resp. strong) RI​I​I.

    Next, we consider the third flat Reidemeister move. The third flat Reidemeister move can be split into the two kinds of moves by the way of connection of three branches as in Fig. 9 (see Fig. 11).

    Refer to caption
    Figure 11. The strong (upper) and weak (lower) RI​I​I.

    The upper (resp. lower) local move of Fig. 11 is called the strong (resp. weak) RI​I​I.

    These two local moves can be detected easily by giving an orientation to a knot projection. In fact, one appears in the left of the third line and the other appears in the fourth line of Fig. 10 if any orientations of knot projections are given.

    We apply non-Seifert resolutions to the two types of the third flat Reidemeister move (the left column in the third and fourth lines of Fig. 10). Therefore, it is sufficient to consider to the two cases of connections of three arcs that are parts of simple circles (the bottom line of Fig. 10). The two figures in the bottom line of Fig. 10 shows that |τ(P)||τ(P)||\tau(P)|-|\tau(P^{\prime})| is either 0 or ±2\pm 2 if a knot projection PP is related to the other knot projection PP^{\prime} by a single third flat Reidemeister move. This completes the proof of (1).

  2. (2)

    For two knot projections P1P_{1} and P2P_{2}, we give them any orientations, and denote by P1oP_{1}^{o} and P2oP_{2}^{o} these oriented knot projections. The knot projection P2P_{2} with the opposite orientation is denoted by P2o¯P_{2}^{\bar{o}}. Now we consider a connected sum P1P2P_{1}\sharp P_{2} of P1P_{1} and P2P_{2}. We can preserve the orientation of P1oP_{1}^{o} under this connected sum by choosing either P2oP_{2}^{o} or P2o¯P_{2}^{\bar{o}}. Note that the τ(P2o)\tau(P_{2}^{o}) == τ(P2o¯)\tau(P_{2}^{\bar{o}}). Apply non-Seifert resolution at all double points P1P2P_{1}\sharp P_{2}. The way of smoothing them is the same as those for P1P_{1} and P2P_{2} since we consider either P1oP2oP_{1}^{o}\sharp P_{2}^{o} or P1oP2o¯P_{1}^{o}\sharp P_{2}^{\bar{o}}.

  3. (3)

    If there exists at least one part as one of the list, we have at least two circles in τ(P)\tau(P). By (1), we have |τ(P)|3|\tau(P)|\geq 3.

Remark 2.

Let PP be a knot projection and set a(P)a(P) == 12(J++2St)-\frac{1}{2}(J^{+}+2St) == J+/2St-J^{+}/2-St, where J+J^{+} and StSt are Arnold invariants of immersed plane curves when we assign \infty to one arbitrary chosen region. It is easy to see that a(P)a(P) does not depend on the choice of \infty and is an invariant of knot projections [9, Page 997, Corollary 2]. Moreover, it is easy to note that a(P)a(P) is invariant under strong (1, 2) homotopy.

We can also demonstrate that |τ(P)||\tau(P)| is independent of a(P)a(P). For instance, let OO (resp. 313_{1}) be the simple circle (resp. the trefoil projection) and let 515_{1} (resp. 636_{3}) be the 515_{1} (resp. 636_{3}) projection as shown in Fig. 20. On one hand, a(O)=0a(O)=0, a(31)=1a(3_{1})=-1, and a(51)=a(63)=2a(5_{1})=a(6_{3})=-2. On the other hand, |τ(O)|=1|\tau(O)|=1, |τ(31)|=|τ(63)|=3|\tau(3_{1})|=|\tau(6_{3})|=3, and |τ(51)|=5|\tau(5_{1})|=5. As previously demonstrated, circle arrangement is strictly stronger than circle number, and as shown in Fig. 20, τ(O)\tau(O), τ(31)\tau(3_{1}), τ(51)\tau(5_{1}), and τ(63)\tau(6_{3}) are all mutually different.

5. Proof of Theorem 2

In this section, we prove Theorem 2. To avoid confusion, recall that the symbol s\sim_{s} (resp. w\sim_{w}) represents strong (1, 2) homotopy (resp. weak (1, 2) homotopy) (see the definition of s\sim_{s} and w\sim_{w} in the statement of Theorem 1).

Proof.

The assumption is that there exists a knot projection PP such that PP is equivalent to the simple circle OO up to weak (1, 2) homotopy (i.e. PP w\sim_{w} OO) and PP is also equivalent to OO up to strong (1, 2) homotopy (i.e. PP s\sim_{s} OO). The condition that PP s\sim_{s} OO implies that |τ(P)||\tau(P)| == 11. On the other hand, PP w\sim_{w} OO implies that there exists a finite sequence (\ast) of 1a1a and w2aw2a from OO to PP by Lemma 1.

Now, we assume that the sequence (\ast) contains at least one w2aw2a (\star). Therefore, there exists a knot projection PP^{\prime} with at least one incoherent 22-gon such that P1PP^{\prime}\stackrel{{\scriptstyle 1}}{{\sim}}P. Then, |τ(P)||\tau(P^{\prime})| == |τ(P)||\tau(P)| == 11. By Theorem 4 (3), |τ(P)||\tau(P^{\prime})| \neq 11 is a contradiction. Thus, the assumption (\star) is false and then the sequence (\ast) contains only 1a1a. Then PP 1\stackrel{{\scriptstyle 1}}{{\sim}} OO.

Conversely, assume that P1OP\stackrel{{\scriptstyle 1}}{{\sim}}O. Then, in particular, PP satisfies both PwOP\sim_{w}O and PsOP\sim_{s}O. This completes the proof of Theorem 2. ∎

6. Properties of circle arrangements

We call PP a prime knot projection if PP is non-trivial and cannot be a connected sum of non-trivial knot projections.

Theorem 5.

For any circle arrangement ss, there exists a knot projection PP such that τ(P)=s\tau(P)=s. In particular, for any positive integer mm, there exists a knot projection PP such that |τ(P)|=m|\tau(P)|=m.

Moreover, we can choose a prime knot projection as PP.

After we demonstrate Lemmas 3, 4, and 5, we will prove Theorem 5.

Lemma 3.

A circle arrangement τ(P)\tau(P) can be locally changed as in Fig. 12 from τ(P)\tau(P) to τ(PP)\tau(P\sharp P^{\prime}) by considering a connected sum PPP\sharp P^{\prime} if PP^{\prime} is the trefoil projection or 636_{3} projection where the trefoil projection and 636_{3} projection look as in Fig. 12.

Refer to caption
Figure 12. The connected sums of a knot projection and either the trefoil projection (upper) or 636_{3} projection (lower). In the second and fourth lines, we omit other circles which remain.
Proof.

It is easy to see this proof by the definition of the circle arrangement. ∎

Lemma 4.

Let PP be a knot projection and assume |τ(P)|3|\tau(P)|\geq 3. There exists a 22-disk DD on the 22-sphere such that DD contains exactly two circles that are separated or nested in DD as illustrated in Fig. 13.

Refer to caption
Figure 13. Separated circles (left) and nested circles (right).
Proof.

We pick up the innermost circle in a nested collection of circles and choose one of them arbitrarily, denoting it CC. If the circle CC is not encircled as in the right figure of Fig. 13, then there should exist at least one circle as in the left figure of Fig. 13. Thus, the proof is completed. ∎

Lemma 5.

There exists a finite sequence generated by 1a1a and s2as2a to obtain a prime knot projection from any non-prime knot projection.

Proof.
Refer to caption
Figure 14. A prime decomposition.

If a knot projection PP is a simple curve, it is easy to see that 414_{1} projection (cf. Fig. 20) is obtained by 1a1a and s2as2a. In the following, we assume that a knot projection is a non-trivial knot projection. We consider a prime decomposition as Fig. 14. If a knot projection PP has a non-trivial sub-curve such that the sub-curve has exactly two end points and if we can find a simple arc closing the two endpoints of the curve that produces a prime knot projection and where the closing does not create double points, then the non-trivial sub-curve is called a prime part of PP. Let us encircle the prime part at one of the most inner regions by one circle, called a red circle, where we arbitrarily select an innermost prime part. If we replace the prime part with a simple arc, we precede to the next stage and repeat; we encircle the prime part of one of the innermost regions by a red circle that must not intersect the other red circles. We continue to draw red circles not intersecting the other red circles. Finally, we draw a circle whose part is contained by the outermost region and select a next-innermost prime part and repeat the above process. For each step, if the outermost red circle is already given by the former process, we omit drawing it (e.g., Fig. 14). Since the number of double points in a knot projection is finite, every encircling must eventually stop. As a result, we can choose a decomposition where each red circle rr contains a prime part where we replace prime parts t1,t2,,tmt_{1},t_{2},\dots,t_{m} with simple arcs a1,a2,,ama_{1},a_{2},\dots,a_{m} if the other red circles r1,r2,,rmr_{1},r_{2},\dots,r_{m} are contained in rr.

Refer to caption
Figure 15. Marked arcs and a circle encircling a prime part.

Let us look at Fig. 15. Every red circle can be drawn as in Fig. 15. Call the two arcs of Fig. 15 marked arcs. Every arc of the sub-diagram within the red circle of Fig. 15 is called a prime part’s arc. Choose two non-marked arcs α1\alpha_{1} and α2\alpha_{2} in the same region where one is picked from a prime part’s arc α1\alpha_{1} and the other is picked from a distinct neighboring prime part’s arc α2\alpha_{2}. Apply the operation shown in Fig. 16 to arcs α1\alpha_{1} and α2\alpha_{2}. This operation is one s2as2a or a pair of s2as2a and 1a1a; if two arcs α1\alpha_{1} and α2\alpha_{2} cannot make a coherent 22-gon directly, we make the allowable situation by applying one 1a1a (see Fig. 16).

Refer to caption
Figure 16. Operation s2as2a or a pair of s2as2a and 1a1a to make a prime knot projection from a connected sum of knot projections. Dotted arcs represent the possible need for 1a1a before applying s2as2a.

As a result, the relation of the two neighboring prime parts is not a connected sum and so we should delete the red circle between the two prime parts. We apply this operation to every red circle, and so we can eventually resolve all the connected summations to obtain a prime knot projection. Thus, the proof is completed. ∎

Next, we prove Theorem 5.

Proof.

If |τ(P)|2|\tau(P)|\leq 2, |τ(P)|=1|\tau(P)|=1 by Theorem (1). We have |τ(41)|=1|\tau(4_{1})|=1 (the symbol 414_{1} presents a knot projection, see Fig. 20). Thus, we can assume that |τ(P)|3|\tau(P)|\geq 3 in the following. Let Q1Q_{1} (resp. Q2Q_{2}) be the left (resp. right) figure of Fig. 13. By Lemma 4, we can find at least one copy of either Q1Q_{1} or Q2Q_{2} within any arbitrary circle arrangement. For such an arbitrary circle arrangement s=s1s=s_{1}, obtain s2s_{2} by deleting this part, denoted by R1R_{1} == Q1Q_{1} or Q2Q_{2}. In s2s_{2}, we can find at least one copy of Q1Q_{1} or Q2Q_{2} which we denoted by R2R_{2}. Repeating the deletions, we eventually have exactly one circle sm+1s_{m+1} and a sequence R1R2R3Rm1RmR_{1}R_{2}R_{3}\dots R_{m-1}R_{m} for some positive integer mm.

Next, we will construct any circle arrangement from one circle OO. For one circle OO, we put RmR_{m} in the two regions given by OO on S2S^{2}. Using Lemma 3, if RmR_{m} == Q1Q_{1} (resp. Q2Q_{2}), we consider the connected sum of the trefoil (resp. 636_{3}) projection and OO as in Fig. 12. Next, for Rm1R_{m-1}, we consider an appropriate connected sum by specifying Rm1R_{m-1} and sm1s_{m-1}. By applying connected sums step by step as above, we finally obtain a knot projection PP such that τ(P)\tau(P) == s1s_{1} == ss and PP consists of finitely many trefoil and 636_{3} projections.

Next we consider the goal of obtaining a prime knot projection such that τ(P)\tau(P) == ss. To obtain such a knot projection, every time we consider a connected sum in the above step RkRk1R_{k}\to R_{k-1}, we apply an operation s2as2a or a pair (1a)(s2a)(1a)(s2a) as Fig. 16 (e.g., Fig. 17).

Refer to caption
Figure 17. Example of a prime knot projection from the connected sum of two 636_{3} projections.

7. Table of circle numbers and circle arrangements

Finally, we include a table of circle invariants on prime knot projections without 11-gons up to seven double points (Fig. 20). The symbols for circle arrangements appearing in Fig. 20 are defined by Fig. 19. Note that the following two points. The first point explains the meaning of the symbols 7A7_{A}, 7B7_{B}, and 7C7_{C}. Traditionally, cmc_{m} denotes the projection image of the knot cmc_{m} in Rolfsen table. To tabulate prime knot projections without 11-gons up to seven double points, all knot projections are obtained from cmc_{m} by flypes of knot projections defined by Fig. 18 (cf. Tait Conjecture).

Refer to caption
Figure 18. A flype.

Thus, exactly three knot projections should be added, i.e., 7A7_{A} (from 767_{6}), 7B7_{B} (from 777_{7}), and 7C7_{C} (from 757_{5}). The second point is that we obtain the definitions of lines with ww and ss in Fig. 20. Two knot projections are connected by a line with ss (resp. ww) if a finite sequence generated by moves of type 1a1a and a single s2as2a (resp. w2aw2a) that connects the two knot projections is found.

Refer to caption
Figure 19. Definition of symbols assigned to circle arrangements for up to five circles on a sphere.
Refer to caption
Figure 20. Prime knot projections without 11-gons for small numbers of double points with circle arrangements. Numbers in this table represent circle numbers (e.g., 33 of the symbol 3a3a).

Acknowledgements

The authors would like to thank Professor Kouki Taniyama for his fruitful comments. The authors would also like to thank the referee for his/her comments on an earlier version of this paper. The part of this work of N. Ito was supported by a Waseda University Grant for Special Research Projects (Project number: 2014K-6292) and the JSPS Japanese-German Graduate Externship. N. Ito is a project researcher of Grant-in-Aid for Scientific Research (S) 24224002 (April 2016–).

References

  • [1] V. I. Arnold, Topological invariants of plane curves and caustics, University Lecture Series, 5. American mathematical Society, Providence, RI, 1994.
  • [2] R. Fenn, Techniques of geometric topology, London Mathematical Society Lecture Note Series, 57. Cambridge University Press, Cambridge, 1983.
  • [3] R. Fenn and P. Taylor, Introducing doodles, Topology of low-dimensional manifolds (Proc. Second Sussex Conf., Chelwood Gate, 1977), pp.37–43.
  • [4] N. Ito and A. Shimizu, The half-twisted splice on reduced knot projections, J. knot Theory Ramifications 21 (2012), 1250112, 10pp.
  • [5] N. Ito and Y. Takimura, (1, 2) and weak (1, 3) homotopies on knot projections, J. Knot Theory Ramifications 22 (2013), 1350085, 14pp.
  • [6] N. Ito, Y. Takimura, and K. Taniyama, Strong and weak (1, 3) homotopies on knot projections, Osaka J. Math. 52 (2015), 617–646.
  • [7] M. Khovanov, Doodle groups, Trans. Amer. Math. Soc. 349 (1997), 2297–2315.
  • [8] A. Shumakovitch, Explicit formulas for the strangeness of plane curves, (Russian) Algebra i Analiz 7 (1995), 165–199; translation in St. Petersburg Math. J. 7 (1996), 445–472.
  • [9] M. Polyak, Invariants of curves and fronts via Gauss diagrams, Topology 37 (1998), 989–1009.
  • [10] O. Viro, Generic immersions of the circle to surfaces and the complex topology of real algebraic curves, Topology of real algebraic varieties and related topics, 231–252, Amer. Math. Soc. Transl. Ser. 2, 173, Amer. Math. Soc., Providence, RI, 1996.