The Geometry Junkyard

Layered Graph Drawing

A planar graph is one that can be drawn in the plane with no crossing edges. Planar graphs have been investigated for a long time, and it is known (Fáry's theorem) that any planar graph can be drawn in such a way that all edges are straight line segments.

But what can you do when a graph is not planar? I and my co-authors have been studying drawings in which the edges are grouped into layers (shown here as colors) such that no two edges in the same layer cross. It turns out (unlike for planar graphs) that it makes a difference whether you require the edges to be straight line segments. The minimum number of layers in a drawing, if you allow curved edges, is a well studied concept known as the thickness of the graph. The minimum number of layers in a drawing with straight edges is a less well studied concept which we call the geometric thickness.

Complete Graphs

The following drawing shows that the geometric thickness of the complete graph on twelve vertices (K12) is at most three:

Three-layer drawing of K_{12}

The central hexagon of this drawing is a little hard to see, so here is a closer view:

Close-up drawing of K_{12}

The pattern used for this drawing can be generalized to show that, for any n, the graph Kn can be drawn in at most ceiling(n/4) layers [Eppstein et al, JGAA 2000]. We also have a lower bound that at least n/5.646 layers are needed. Since Kn is known to have n/6-layer drawings with curved edges, this lower bound shows that geometric thickness is not the same concept as thickness.

The following table shows what we know about lower bounds and upper bounds for the number of layers required for complete graphs Kn.

n 1 - 4 5 - 8 9 - 12 13 - 14 15 - 16 17 - 20 21 - 24 25 - 26 27 - 28 29 - 31 32
lower bound 1 2 3 3 4 4 5 6 7
upper bound 4 5 6 7 8

Subdivisions of Complete Graphs

If one splits each edge of a complete graph (or any other graph) into a two-edge path, the resulting subdivided graph has geometric thickness two. The following picture shows a drawing for K8:

Two-layer drawing of subdivision of K_8

Interestingly, a Ramsey-theoretic argument shows that the book thickness of subdivisions of Kn is not bounded by any constant. So, there are graphs with geometric thickness two and arbitrarily large book thickness [Eppstein, 2001]. A very similar Ramsey-theoretic argument, applied to graphs formed by starting with n points and adding a new point adjacent to each triple of the n points, shows that there are also graphs with thickness three and arbitrarily large geometric thickness [Eppstein, Contemp. Math. 2004].

Complete Bipartite Graphs

In our JGAA 2000 paper we also investigated the geometric thickness of complete bipartite graphs. When a is much smaller than b, the geometric thickness of Ka,b is just a/2, but complete bipartite graphs in which both sides are of nearly equal size are more interesting. For instance, here is a two-layer drawing of the complete bipartite graph K6,6:

Two-layer drawing of K_{6,6}

K1,n and K2,n are planar (one layer), and K3,n and K4,n always require exactly two layers, so the first nontrivial cases are K5,n and K6,n. K5,n requires two layers for n < 8 and three layers for n > 11. For 9 < n < 10 we don't know whether two or three layers are needed. A drawing of K5,8 is below. K6,n requires three layers for n > 8. For K6,7 we don't know whether two or three layers are needed.

Two-layer drawing of K_{5,8}


Below we show a two-layer drawing of a six-dimensional hypercube. More generally the d-cube has geometric thickness at most ceiling(d/3) [Eppstein, GD 2004].

Two-layer drawing of six-dimensional hypercube

Minor-Closed Graph Families

Blankenship and Oporowski [2001,2003] showed that all proper minor-closed graph families have bounded book thickness. Therefore, they also have bounded geometric thickness.

For graphs of treewidth k, the largest possible thickness and the largest possible geometric thickness are both ceiling(k/2), roughly half the largest possible book thickness [Dujmović and Wood, DCG 2007].

Low Degree Graphs

Graphs with maximum degree up to four require geometric thickness at most two [Duncan et al., SCG 2004]. How does the geometric thickness behave for bounded degree graphs with higher degree bounds? Some progress has been made by [Dujmović et al., GD 2004] who showed that the geometric thickness is O(1) for bounded-degree interval graphs, co-comparability graphs, and AT-free graphs, and that it is O(log n) for bounded-degree bounded-treewidth graphs. Duncan [CCCG 2009] showed that the geometric treewidth is also O(log n) for graphs that can be decomposed into the union of two bounded-treewidth graphs, without degree restrictions.

However, Barát, Matoušek, and Wood [EJC 2006] have shown that larger constant bounds on degree need not imply bounded thickness for arbitrary graphs. In particular, they use counting arguments to prove the existence of Δ-regular graphs for any Δ > 8 with geometric thickness Ω(nc) for some c > 0, where c depends on Δ and approachs 1/2 in the limit of large Δ.

Computational complexity

For arbitrary graphs, testing whether the graph has thickness k is NP-complete, even for k=2 [Mansfield, MPCPS 1983]. Testing the geometric thickness of a particular drawing is polynomial time for thickness two but NP-complete for higher thickness [Eppstein, SODA 2004]. Recognizing whether a graph has geometric thickness two is also NP-complete, as is coloring graphs of geometric thickness t with 4t − 1 colors [Durocher, WG 2013].

Open Problems

  1. The complete graphs Kn should require a number of layers of the form c*n, for some constant c. What is c?
  2. Similarly, what is the asymptotic number of layers needed for Kn,n?
  3. If a multi-layer drawing of a graph is made in which all the vertices have integer coordinates, how big do those integers need to be? (For planar graphs, it is known that all coordinates can be O(n).)
  4. How large can the ratio between thickness and book thickness of a graph be, as a function of the number of vertices in the graph? Similarly how large can the ratio between geometric thickness and thickness be? The examples of Barát, Matoušek, and Wood [EJC 2006] show that this ratio can at least approach sqrt n.
  5. What happens if we allow restricted types of curves other than straight line segments? For instance, Wood [ENDM 2001] has looked at thickness for curves with one bend that are allowed to change layers at the bend. What if the curves can bend in a restricted way but can't change layers?
  6. Hutchinson et al. proved that the number of edges of a n-vertex graph with geometric thickness two is at most 6n − 18. However, the largest number of edges that is known for an example of such a graph is 6n − 19 (for all n at least 9, by Durocher et al.) Which of these two is the correct bound?
  7. Determining the chromatic number of thickness-two graphs is Ringel's famous Earth-Moon problem. What about the chromatic number of graphs with low geometric thickness? The best lower bound for Ringel's problem comes from the graph K6+C5, which has thickness two and requires nine colors; however this graph has 50 edges, violating a 6n-18 upper bound on the number of edges of a geometric thickness two graph due to Hutchinson et al. Is there a different nine-chromatic graph with geometric thickness two?


J. Barát, J. Matoušek, and D. R. Wood.
Bounded-degree graphs have arbitrarily large geometric thickness.
Electron. J. Combin. 13(1):R3:1–R3:14, 2006.

L. W. Beineke.
Biplanar graphs: a survey.
Graph theory in computer science, chemistry, and other fields, Las Cruces, New Mexico, USA, 1991.
Comput. Math. Appl. 34(11):1-8, 1997.

R. Blankenship.
Book Embeddings of Graphs.
Ph. D. thesis, Louisiana State Univ., Dept. of Mathematics, 2003.

R. Blankenship and B. Oporowski.
Book embeddings of graphs and minor-closed classes.
Proc. 32nd Southeastern International Conf. on Combinatorics, Graph Theory and Computing, 2001.

M. B. Dillencourt, D. Eppstein, and D. S. Hirschberg.
Geometric thickness of complete graphs.
6th Int. Symp. Graph Drawing, Montréal, 1998.
Lecture Notes in Comp. Sci. 1547, pp. 102-110, 1998.
J. Graph Algorithms and Applications 4(3):5-17, 2000.

V. Dujmović, M. Suderman, and D. R. Wood.
Really straight graph drawings.
Proc. 12th Int. Symp. Graph Drawing, New York, 2004.

V. Dujmović and D. R. Wood.
Graph treewidth and geometric thickness parameters.
Discrete Comput. Geom. 37(4):641–670, 2007.

C. Duncan.
On graph thickness, geometrick thickness, and separator theorems.
Proc. 21st Canad. Conf. Comput. Geom., Vancouver, 2009, pp. 13-16.

C. Duncan, D. Eppstein, and S. Kobourov.
The geometric thickness of low degree graphs.
20th ACM Symp. Comp. Geom., Brooklyn, 2004, pp. 340-346.

S. Durocher, E. Gethner, and D. Mondal.
Thickness and Colorability of Geometric Graphs.
Proc. 39th International Workshop on Graph-Theoretic Concepts in Computer Science, 2013.

D. Eppstein.
Separating geometric thickness from book thickness.

D. Eppstein.
Separating thickness from geometric thickness.
10th Int. Symp. Graph Drawing, Irvine, 2002.
Lecture Notes in Comp. Sci. 2528, 2001, pp. 150-161.
In Towards a Theory of Geometric Graphs, AMS, Contemporary Math 342, J. Pach, ed., pp. 75-86, 2004.

D. Eppstein.
Testing bipartiteness of geometric intersection graphs.
Proc. 15th Symp. Discrete Algorithms, New Orleans, pp. 853-861, 2004.
ACM Trans. Algorithms 5(2):15, 2009.

D. Eppstein.
Algorithms for drawing media.
Proc. 12th Int. Symp. Graph Drawing, New York, 2004.

J. P. Hutchinson, T. Shermer, and A. Vince.
On representation of some thickness-two graphs.
3rd Int. Symp. Graph Drawing, Passau, Germany, 1995.
Lecture Notes in Comp. Sci. 1027, pp. 324-332, 1995.
Computational Geometry: Theory and Applications 13(3):161-171, 1999.

P. C. Kainen.
Thickness and coarseness of graphs.
Abhandlungen aus dem Mathematischen Seminar der Univ. Hamburg 39:88-95, 1973.

A. Mansfield.
Determining the thickness of a graph is NP-hard.
Math. Proc. Cambridge Phil. Soc. 93(9):9-23, 1983.

P. Mutzel, T. Odenthal, and M. Scharbrodt.
The thickness of graphs: a survey.
Graphs Combin. 14(1):59-73, 1998.

D. R. Wood.
Geometric thickness in a grid of linear area.
Proc. Euroconf. Combinatorics, Graph Theory, and Applications, pp. 310-315, 2001.
Electronic Notes in Discrete Mathematics 10, 2001.

From the Geometry Junkyard, computational and recreational geometry.
David Eppstein, Theory Group, ICS, UC Irvine.