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

Cubic planar bipartite graphs are dispersable thanks: As submitted for publication on 7/20/2020; see Postscript for additional information.

Paul C. Kainen
[email protected]
Shannon Overbay
[email protected]
Abstract

A graph is called dispersable if it has a book embedding in which each page has maximum degree 1 and the number of pages is the maximum degree. Bernhart and Kainen conjectured every kk-regular bipartite graph is dispersable. Forty years later, Alam, Bekos, Gronemann, Kaufmann, and Pupyrev have disproved this conjecture, identifying nonplanar 3- and 4-regular bipartite graphs that are not dispersable. They also proved all cubic planar bipartite 3-connected graphs are dispersable and conjectured that the connectivity condition could be relaxed. We prove that every cubic planar bipartite multigraph is dispersable.

Key Phrases: book thickness; dispersable book embeddings; matching book thickness; subhamiltonian vertex order; cubic bipartite planar multigraphs.

1 Introduction

The book thickness of a graph, [11, 15] is now well-studied and has a variety of applications such as manipulation of stacks and queues (Even & Itai [8] and Tarjan [17]), fault-tolerant computing [4], traffic engineering [12], RNA folding [10], and graph drawing and visualization, e.g., [6, 7, 20]. There is also an extensive literature on the purely mathematical aspects of this measure of graph complexity; [19] lists over a couple of dozen and is by no means complete.

The book thickness of a graph is the least number of colors needed for the edges such that, for some outerplane drawing of the graph, no two edges of the same color cross. In [3], we introduced the additional constraint that same-color edges do not share an endpoint, so pages must be matchings.

A book embedding of a graph is an outerplane drawing (i.e., a cyclic order on the vertices) and an edge-coloring such that edges of each color (called a page of the embedding) induce an outerplane subgraph. A graph or vertex-order with a 2-page book embedding is called subhamiltonian [3], [16].

The matching book thickness [13] of a graph GG is the least number of pages, mbt(G)mbt(G), needed if the pages are matchings. A graph is said to be dispersable [3] if it has a matching book embedding in Δ(G)\Delta(G) pages, where Δ(G)\Delta(G) denotes maximum degree; we refer to a dispersable book embedding.

Bernhart and Kainen [3] conjectured (in 1979):
If G is regular bipartite, then G has a dispersable book embedding.\mbox{\it If $G$ is regular bipartite, then $G$ has a dispersable book embedding}.
This BK Conjecture was recently disproved by Alam et al. [1] who showed:
There exist bipartite 3- and 4-regular graphs GG with mbt(G)=4mbt(G)=4 or 55, resp.
These graphs, due to Gray and Folkman, respectively, are edge-transitive but not vertex-transitive. In fact, Tutte [18, p. 67] proved that any such graph must be bipartite. Further, both the Gray and Folkman graphs are nonplanar.

In contrast, we know of many graphs, all vertex transitive, for which Conjecture BK does hold, including cycles, hypercubes, complete bipartite graphs, degree-3 circulants, and the Heawood graph [3, 16, 13]. Also, the Cartesian product of dispersable graphs is dispersable, and if GG is bipartite, then mbt(G)mbt(G) is bounded above by the minimum sum of the page-maximum degrees over all book embeddings and trivially by bt(G)Δ(G)bt(G)\Delta(G); see [14].

In [1] Alam-et-al proved Conjecture BK holds if GG is cubic planar bipartite 3-connected and conjectured the condition, that the graph be 3-connected, could be relaxed. We confirm their conjecture by proving that all cubic planar bipartite multigraphs are dispersable, so Conjecture BK holds in this case.

The paper is organized as follows. Section 2 has some combinatorial preliminaries, including two “folklore” results for regular bipartite graphs—non-existence of a cutpoint and an “entanglement” lemma: any two cut-edges have vertices of attachment which are in different parts of the bipartition. Section 3 contains a proof of the main result.

2 Combinatorial preliminaries

Undefined graph terminology is standard [9], [5]. Let |||\cdot| denote cardinality.

By a graph G=(V,E)G=(V,E) we mean a 1-dimensional simplicial complex; i.e., no loops and at most one edge between any pair of vertices; GG is a multigraph if we allow parallel edges, which are sets of more than one edge between a pair of vertices. Thus, every graph is a multigraph. A digraph is a multigraph whose edges have been oriented (that is, assigned a direction).

A set of edges in a multigraph GG is a matching if the edges are pairwise disjoint; a matching is perfect if each vertex is in an edge of the matching. Each page of a dispersable embedding of a regular graph is a perfect matching.

The following must be known but we haven’t found it in the literature.

Lemma 1.

Let GG be a bipartite, connected, and kk-regular multigraph, k3k\geq 3. Let e,e′′e^{\prime},e^{\prime\prime} be two disjoint edges in GG such that G{e,e′′}G\setminus\{e^{\prime},e^{\prime\prime}\} contains a connected component HH, and let u,u′′u^{\prime},u^{\prime\prime} be the endpoints of e,e′′e^{\prime},e^{\prime\prime} which are in HH. Then u,u′′u^{\prime},u^{\prime\prime} are in distinct parts of the unique bipartition of V(G)V(G).

Proof.

The oriented degree deg(v)\deg(v) of a vertex in a digraph is the number of edges in minus the number of edges out. The sum of all oriented vertex degrees for any digraph is zero.

Let (H(H, uu^{\prime}, u′′)u^{\prime\prime}) be as in the statement; give GG and HH the edge-orientation induced from the bipartition by orienting the edges from white to black, so each vertex has oriented degree ±k\pm k according to whether it’s black or white. Put

W:=V(H{u,u′′})=V(H){u,u′′}.W:=V(H\setminus\{u^{\prime},u^{\prime\prime}\})=V(H)\setminus\{u^{\prime},u^{\prime\prime}\}.

We have

0=vV(H)degH(v)=degH(u)+degH(u′′)+vWdegH(v).0\;\;=\sum_{v\in V(H)}\deg_{H}(v)\,=\,\deg_{H}(u^{\prime})+\deg_{H}(u^{\prime\prime})+\sum_{v\in W}\deg_{H}(v). (1)

For each vWv\in W, degH(v)=degG(v)\deg_{H}(v)=\deg_{G}(v), so vWdegH(v)0\sum_{v\in W}\deg_{H}(v)\cong 0 (mod kk), hence, by (1), degH(u)+degH(u′′)0\deg_{H}(u^{\prime})+\deg_{H}(u^{\prime\prime})\cong 0. As k3k\geq 3, 2k202k-2\ncong 0 (mod kk), so the degrees of uu^{\prime} and u′′u^{\prime\prime} must be of opposite sign, i.e., uu^{\prime} and u′′u^{\prime\prime} are in different parts. ∎

By a similar argument, one can prove the following “folklore” result.

Proposition 1.

Connected regular bipartite graphs have no cutpoints.

Proof.

Let GG be a connected kk-regular bipartite graph. If k=0k=0 or 11, then GG is K1K_{1} or K2K_{2}, resp., and there are no cutpoints. For k=2k=2, bipartite or not, GG is a cycle and so has no cutpoint or bridge. Suppose now k3k\geq 3. Let uu be a cutpoint of GG and let Γ\Gamma be a connected component of GuG-u.

We define J:=Γ+u:=G(V(Γ){u})J:=\Gamma+u:=G(V(\Gamma)\cup\{u\}) to be the smallest subgraph of GG containing Γ\Gamma and uu. Then again using oriented degrees,

0=vV(J)degJ(v)=degJ(u)+vV(Γ)degG(v);0\;\;=\sum_{v\in V(J)}\deg_{J}(v)\,=\,\deg_{J}(u)+\sum_{v\in V(\Gamma)}\deg_{G}(v); (2)

the second summand is congruent to zero (mod kk) but degJ(u)0\deg_{J}(u)\ncong 0 (mod kk) so no cutpoint uu can exist. ∎

Observe that for k2k\geq 2, there is also no bridge.

3 Cubic planar bipartite graphs

Alam-et-al. proved the following result [1, pp. 18–21].

Theorem 1 ([1]).

Let GG be a bipartite cubic planar 33-connected graph. Then GG has a subhamiltonian vertex ordering ω\omega such that mbt(G)=mbt(G,ω)=3mbt(G)=mbt(G,\omega)=3.

We say that such a GG is subhamiltonian dispersable. It is conjectured in [1] that 33-connected is not needed, as we now show.

Theorem 2.

Bipartite cubic planar multigraphs are subhamiltonian dispersable.

Our proof of Theorem 2 uses multigraphs. The trivial cubic multigraph, which we denote by Θ\Theta, has 2 vertices joined by 3 parallel edges.

Lemma 2.

A cubic multigraph Θ\neq\Theta with parallel edges cannot be 33-connected.

Proof.

If GG is cubic and f1,f2f_{1},f_{2} are parallel edges joining vertices aa and bb, then either (i) there is a third parallel edge f3f_{3} and GG contains the trivial cubic graph as a connected component or (ii) the remaining two edges f3f_{3} and f4f_{4} at aa and bb, resp., are a cutset for GG, as they separate {a,b}\{a,b\} from the other vertices. ∎

[Uncaptioned image]

Figure 1   Parallel edges in a cubic graph


Hence, a nontrivial 3-connected cubic multigraph has no parallel edges and is a cubic graph. We are now ready for the proof of Theorem 2.

Proof.

Let GG be a bipartite cubic planar multigraph. Assume without loss of generality ([3], [16]) that GG is connected. We need to show that GG has a subhamiltonian vertex order ω\omega and an edge-3-coloring cc such that in the outerplane drawing (G,ω)(G,\omega), each color class of cc is crossing-free; that is,

bt(G)=bt(G,ω)2andmbt(G)=mbt(G,ω)=3.bt(G)=bt(G,\omega)\leq 2\;\;\mbox{{\rm and}}\;\;mbt(G)=mbt(G,\omega)=3. (3)

Suppose that GG is a counterexample with the minimum number of vertices. Note that GG must be a nontrivial cubic multigraph as the required ω\omega and cc do exist for Θ\Theta, and GG cannot be 33-connected else it is dispersable by Theorem 1 which applies as there can be no parallel edges.

Since GG is bipartite and cubic, by Proposition 1 it has no cutpoint and hence must be 2-connected. As line-connectivity and point-connectivity coincide for cubic graphs [9, p. 55, ex. 5.6 ], there must be a cutset of two edges. But the two edges of such a cutset cannot share a vertex – the third edge at a common vertex would be a bridge. Hence, GG contains vertex-disjoint edges e=uw,e′′=u′′w′′e^{\prime}=u^{\prime}w^{\prime},\;e^{\prime\prime}=u^{\prime\prime}w^{\prime\prime} (neither of which is a parallel edge) such that

G{e,e′′}=G1G2,u,u′′V(G1),w,w′′V(G2),G\setminus\{e^{\prime},e^{\prime\prime}\}=G_{1}\cup G_{2},\;\;u^{\prime},u^{\prime\prime}\in V(G_{1}),\;\;w^{\prime},w^{\prime\prime}\in V(G_{2}), (4)

where G1G_{1} and G2G_{2} are vertex-disjoint. Further, by Lemma 1, uu^{\prime} and u′′u^{\prime\prime} are in distinct parts of the bipartition of G1G_{1} induced by the bipartition of GG, and similarly for w,w′′w^{\prime},w^{\prime\prime} in the corresponding bipartition of G2G_{2}.

Let H1=G1+e1H_{1}=G_{1}+e_{1}, where e1:=uu′′e_{1}:=u^{\prime}u^{\prime\prime}, H2=G2+e2H_{2}=G_{2}+e_{2}, where e2:=ww′′e_{2}:=w^{\prime}w^{\prime\prime}. Then H1H_{1} and H2H_{2} are bipartite cubic planar multigraphs, either of which can contain multiple edges even if GG has no parallel edges. As |H1|,|H2|<|G||H_{1}|,|H_{2}|<|G|, there exist cyclic orderings ω1,ω2\omega_{1},\omega_{2} on V(H1)=V(G1)V(H_{1})=V(G_{1}) and V(H2)=V(G2)V(H_{2})=V(G_{2}), resp., and edge colorings c1,c2c_{1},c_{2} which constitute dispersable subhamiltonian book embeddings of H1H_{1} and H2H_{2}. Let λ1\lambda_{1} be one of the two possible linear orderings obtained from ω1\omega_{1} which starts at u′′u^{\prime\prime}. Similarly, let λ2\lambda_{2} be one of the two possible linear orderings obtained from ω2\omega_{2} which starts at w′′w^{\prime\prime}.

Define λ\lambda a linear order on V(G)=V(H1)V(H2)V(G)=V(H_{1})\cup V(H_{2}) by the equation λ:=λ1opλ2\lambda:=\lambda_{1}^{op}\,*\,\lambda_{2}, where λ1op\lambda_{1}^{op} is the opposite (i.e., reverse) order to λ1\lambda_{1} and * denotes concatenation. Let ω\omega be the cyclic order determined by λ\lambda. Observe that u′′u^{\prime\prime} and w′′w^{\prime\prime} are consecutive in the subhamiltonian ordering of V(G)V(G) given by ω\omega.

By renaming the colors, we can assume that the colorings c1c_{1} and c2c_{2} both use colors {α,β,γ}\{\alpha,\beta,\gamma\} and that e1e_{1} and e2e_{2} are both assigned γ\gamma. Define cc an edge 3-coloring of E(G)E(G) to be c1c_{1} on E(G1)E(G_{1}), c2c_{2} on E(G2)E(G_{2}), and put c(e)=c(e′′)=γc(e^{\prime})=c(e^{\prime\prime})=\gamma. If eE(G1)e\in E(G_{1}) is adjacent to either ee^{\prime} or e′′e^{\prime\prime}, then ee was adjacent to e1e_{1}, so c(e)c(e)c(e)\neq c(e^{\prime}) and similarly for G2G_{2}. Observe that e′′e^{\prime\prime} has no crossings in (G,ω)(G,\omega).

[Uncaptioned image]

Figure 2   Subhamiltonian dispersable embedding of GG


We claim ee^{\prime} crosses no edge colored γ\gamma. For if e=xye=xy crosses ee^{\prime} in the drawing (G,ω)(G,\omega), then xx and yy both belong to G1G_{1} or both belong to G2G_{2}; else {e,e′′}\{e^{\prime},e^{\prime\prime}\} wouldn’t be a cutset. Thus xyxy must cross e1e_{1} or e2e_{2}, hence c(xy)γc(xy)\neq\gamma. So (ω,c)(\omega,c) is a dispersable subhamiltonian book embedding of GG which is impossible. ∎

4 Discussion

Since the examples of non-dispersable bipartite graphs given in Alam-et-al. are nonplanar and not vertex transitive, natural questions arise. They ask [1, question 6]: Does the BK Conjecture in [3] hold for planar bipartite graphs?

All examples we know support the following conjecture which would include a positive answer to [1, questions 2 and 4]: Every vertex-transitive graph GG with Δ(G)=r\Delta(G)=r has mbt(G)=rmbt(G)=r or r+1\,r{+}1 according to whether or not GG is bipartite. However, some condition on GG is needed as Barat et al.[2] show that if Δ(G)9\Delta(G)\geq 9, bounded degree graphs cannot have bounded book thickness.

References

  • [1] J. Md. Alam, M. A. Bekos, M. Gronemann, M. Kaufmann & S. Pupyrev, On dispersable book embeddings, Graph-theoretic Concepts in Computer Science (44th International Workshop, WG 2018, Cottbus, Germany), A. Brandestädt, E. Köhler, & K. Meer, Eds., LNCS 11159 (2018) 1–14, Springer, Cham, Switzerland.
  • [2] J. Barat, J. Matousek & D. R. Wood, Bounded-degree graphs have arbitrarily large geometric thickness, The Elec. J. of Comb. 13 (2006) #R13.
  • [3] F. R. Bernhart & P. C. Kainen, The book thickness of a graph, J. Comb. Theory, Series B, 27 (1979) 320–331.
  • [4] F. R. K. Chung, F. T. Leighton, & A. L. Rosenberg, Embedding graphs in books: A layout problem with applications to VLSI design, SIAM J. Alg. Discr. Meth. 8 (1) (1987) 33–58.
  • [5] R. Diestel, GraphTheory, Springer, Berlin, 2005.
  • [6] D. Eppstein, Separating geometric thickness from book thickness, 2001.
    http://arXiv.org/math/0109195.
  • [7] D. Eppstein, Separating thickness from geometric thickness. In Graph Drawing, (10th International Symp. GD 2002), M. T. Goodrich & S. G. Kobourov, eds., 150–161, Springer, Berlin, 2002. Also in Towards a Theory of Geometric Graphs, Janos Pach, ed., Contemporary Math., 342 (2004) 75–86. Amer. Math. Soc., Providence, RI.
  • [8] S. Even & A. Itai, Queues, stacks, and graphs, in Theory of Machines and Computations, A. Press, Ed., pp. 71–86, 1971.
  • [9] F. Harary, Graph Theory, Addison-Wesley, Reading, MA 1969.
  • [10] C. Haslinger & P. F. Stadler, RNA structures with pseudo-knots: Graph-theoretical, combinatorial, and statistical properties, Bull. of Mathematical Biology 61 (3) (1999) 437–467.
  • [11] P. C. Kainen, Some recent results in topological graph theory, in Graphs and Combinatorics, R. A. Bari & F. Harary, Eds., Springer Lect. Notes in Math. 406, Berlin, 1974, pp. 76–108.
  • [12] P. C. Kainen, The book thickness of a graph, II, Congr. Numeratium 71 (1990), 127–132.
  • [13] P. C. Kainen, Circular layouts for crossing-free matchings, preprint 2009. http://faculty.georgetown.edu/kainen/circLayouts.pdf
  • [14] P. C. Kainen, S. Overbay, On matching book thickness, in preparation.
  • [15] L. T. Ollmann, On the book thicknesses of various graphs, Congressus Numerantium VIII (1973) 459 (Proc. 4th Southeastern Conference on Combinatorics, Graph Theory and Computing, F. Hoffman, R. B. Levow, R. S. D. Thomas, Eds.
  • [16] S. Overbay, Generalized Book Embeddings, Ph. D. Dissertation, Colorado State University, Fort Collins, CO, 1998.
  • [17] R. E. Tarjan, Sorting using networks of queues and stacks, J. Assoc. of Computing Machinery 19 (1972) 341–346.
  • [18] W. T. Tutte, Connectivity in Graphs, U. of Toronto Press, London, 1966.
  • [19] Wikipedia, Book Embedding, accessed 18 July 2020
    https://en.wikipedia.org/wiki/Book_embeddinghttps://en.wikipedia.org/wiki/Book\_embedding
  • [20] D. R. Wood, Bounded degree book embeddings and three-dimensional orthogonal graph drawing, Graph Drawing: 9th International Symposium, P. Mutzel, M. Jünger, & S. Leipert, Eds., LNCS, 2265 (2002) 312–327 Springer, Berlin.

Paul C. Kainen, Department of Mathematics and Statistics,
Georgetown University, 37th and O Streets, N.W., Washington DC 20057

Shannon Overbay, Department of Mathematics, Gonzaga University,
502 E. Boone Ave., Spokane WA 99258


Postscript

The previous paper [2] proves the conjecture of Alam et al. [1] made in their 2018 conference paper which we discovered in 2019. We worked during the Spring of 2020 and on July 20, 2020, our paper was submitted for publication in the form above. After 8 months, on March 26, 2021, we contacted the assistant to the managing editor of the journal and were told that no referee had been found. We then supplied names, affiliations, and emails for several (we thought obvious) choices for possible referees, which were accepted with thanks. On May 30, 2021, the journal rejected the paper. The referee’s report said that the problem had been solved and included a link to the now-published expanded version [3] of Alam et al.’s results.

Indeed, on March 3, 2020, Alam, et al. first submitted a version of the conference paper [1] to the Elsevier journal, Theoretical Computer Science and submitted the second version [3] on Jan. 23, 2021, and it appeared electronically on Feb. 4, 2021. This new journal version [3], with a sixth coauthor, Dujmović, includes a proof of the conjecture in [1].

Another proof of the Alam et al. conjecture was recently given by Shao, Liu, and Li [4]. The paper [4] uses a different method, while [2] and the portion of [3], which proves the conjecture, use very similar techniques.

We hope the comparison of these approaches will be helpful.

References

[1] J. Md. Alam, M. A. Bekos, M. Gronemann, M. Kaufmann & S. Pupyrev, On dispersable book embeddings, Graph-theoretic Concepts in Computer Science (44th International Workshop, WG 2018, Cottbus, Germany), A. Brandestädt, E. Köhler, & K. Meer, Eds., LNCS 11159 (2018) 1–14, Springer, Cham, Switzerland.

[2] P. C. Kainen & S. Overbay, Cubic planar bipartite graphs are dispersable, July 20, 2020 (7 pages).

[3] J. Md. Alam, M. A. Bekos, V. Dujmović, M. Gronemann, M. Kaufmann & S. Pupyrev, On dispersable book embeddings, Theoretical Computer Science, 861 (2021), 1–22.

[4] Z. Shao, Y. Liu & Z. Li, Bipartite cubic planar graphs are dispersable, arXiv:2107.00907v1, July 2, 2021, (10 pages).