# Embeddings of Cubic Halin Graphs A Surface-by-Surface Inventory

Jonathan L. Gross∗ Columbia University, New York, NY 10027
Abstract We derive an O(n2 )-time algorithm for calculating the sequence of numbers g0 (G), g1 (G), g2 (G), ... of distinct ways to embed a 3-regular Halin graph G on the respective orientable surfaces S0 , S1 , S2 , ... . Key topological features are a quadrangular decomposition of plane Halin graphs and a new recombinant-strands reassembly process that ﬁts pieces together three-at-a-vertex. Key algorithmic features are reassembly along a post-order traversal, with just-in-time dynamic assignment of roots for quadrangular pieces encountered along the tour.

1. Introduction 2. Some preexisting results 3. Quadrangulating a plane Halin graph 4. Reassembling from quadrangles 5. Partials and productions 6. Productions for Halin graphs 7. Sample calculation 8. Conclusions and open problems
* gross@cs.columbia.edu; http://www.cs.columbia.edu/∼gross/. 1

Embeddings of Cubic Halin Graphs: An Inventory

2

1

Introduction

We begin with a quick explanation of the title. Halin graphs A Halin graph is constructed as follows: 1. Let T be a plane tree • with at least 4 vertices • no vertices of degree 2 2. Draw a cycle thru the leaves of the tree • in the order they occur in a preorder traversal.

Figure 1.1: A Halin graph for a 14-vertex tree with 8 leaves.

Embeddings of Cubic Halin Graphs: An Inventory

3

Inventories of Embeddings Def. The genus distribution for graph G is the sequence γdist (G) : g0 (G), g1 (G), g2 (G), · · ·

EXAMPLE: K2 × K3 a.k.a. CL3

Figure 1.2: g0 = 2, g1 = 38, g2 = 24

Notation Si denotes the orientable surface of genus i. Notation gi (G) denotes # embeddings G → Si . Notation γmin (G) is the minimum genus of G. Notation γmax (G) is the maximum genus of G.

Embeddings of Cubic Halin Graphs: An Inventory

4

Counting Embeddings Def. Let G be a labeled graph and S an oriented surface. The embeddings ι : G → S and ι : G → S are equivalent iﬀ there exists an autohomeomorphism h : S → S that • maps ι(G) isomorphically to ι (G); • preserves every label of G; • preserves the orientation of S.

0 2 1 3

0 2 3 1

Figure 1.3: Looking alike is not enough.

Embeddings of Cubic Halin Graphs: An Inventory

5

Graph Amalgamations In general, amalgamating two graphs means identifying a subgraph in one of them to an isomorphic subgraph in the other. Vertex-amalgamation and edge-amalgamation are the two simplest kinds of amalgamation of two graphs.

*
V

=

Figure 1.4: Amalgamation of two graphs at a vertex.

*
E

=

Figure 1.5: Amalgamation of two graphs on an edge.

Embeddings of Cubic Halin Graphs: An Inventory

6

Genus Distributions of Graph Amalgamations

u

v

*

=

Figure 1.6: Bar-amalgamation of two graphs.

¯ γdist (G ∗ H) = deg(u) · deg(v) · γdist (G) ◦ γdist (H) here ◦ means convolution; [GrFu87]

g-dist of vertex amalg [GrKhPo10, KhPoGr10, Gr11]. g-dist of edge-amalg [PoKhGr10a, PoKhGr10b].

Embeddings of Cubic Halin Graphs: An Inventory

7

2

Some Preexisting Results

Thm 2.1 Calculating γmin (G) is NP-hard. [Th89] γmin (G) is known for relatively few graphs: • complete graphs Kn [RiYo68] [Ri65]

• complete bipartite graphs Km,n • hypercubes Qn [Ri55]

• a bunch of less prominent families of graphs

K6

K3,4

Q3

Figure 2.1: Three families solved by Ringel.

Remark By way of contrast, there are combinatorial characterizations of maximum genus by [Xu79] and [Ne81] and a polynomial-time algorithm by [FuGrMc88]. There are maximum genus determinations (e.g., see [Sk91]) for families of graphs for which minimum genus calculation would seem to be formidably hard. An up-to-date overview of maximum genus is given by [ChHu09].

Embeddings of Cubic Halin Graphs: An Inventory

8

Minimum Genus of Graph Amalgamations Thm 2.2 (Additivity of minimum genus) γmin (G ∗V H) = γmin (G) + γmin (H) Proof [BHKY62] ♦

By way of contrast, consider the edge-amalgamation K5 ∗E K5 .

Figure 2.2: Amalgamation of two copies of K5 on an edge.

Obviously, the minimum genus is at least 1 and at most 2. But is it really 2?

Embeddings of Cubic Halin Graphs: An Inventory

9

NO!
Y

γmin (K5 ∗E K5 ) = 1.
0 2 X 1 Y

2

X

1

Y

0

2

K5 -> S1
Y 0 2 X 1 Y

2

X

1

Y

0

2

K5 -> S1
Y 0 2 X 1 Y

2

X

1

Y

0

2

K5 *E K5 -> S1
Figure 2.3: K5 ∗E K5 −→ S1

Embeddings of Cubic Halin Graphs: An Inventory

10

About Genus Distributions Calculating γdist (G) requires calculating γmin (G). The graphs whose genus distributions are known are planar or one edge away from planarity. Closed-end ladders [FuGrSt89] (derived 1984).

Figure 2.4: The closed-end ladder L4 .

Figure 2.5: Circular ladder CL4 . and M¨bius ladder M L4 . o

Figure 2.6: Ringel ladder RL4 .

Embeddings of Cubic Halin Graphs: An Inventory

11

Some g-dist deriviations use [Ja87] (group representations). Bouquets [GrRoTu89].

Figure 2.7: Bouquets B1 , B2 , and B3 .

Dipoles [Ri90], [KwLe93].

Figure 2.8: Dipoles D1 , D2 , and D3 .

Fans [ChMaZo11].

Figure 2.9: Fans F3 and F5 .

Embeddings of Cubic Halin Graphs: An Inventory

12

Cubic outerplanar graphs [Gr11].

Figure 2.10: A cubic outerplanar graph.

4-regular outerplanar graphs [PoKhGr11].

Figure 2.11: A 4-regular outerplanar graph.

Remark A survey of genus distributions, including average genus, is given by [Gr09].

Embeddings of Cubic Halin Graphs: An Inventory

13

3

We decompose a plane Halin graph H → S0 into atomic parts that will facilitate a genus distribution calculation. We regard the vertices and the edges of the Halin graph itself as black. Step 1. In each edge on the exterior cycle, insert a red midpoint.

Figure 3.1: Halin graph with red midpoints on exterior cycle.

Since H is a Halin graph, there is exactly one exterior edge on each polygonal face of the plane embedding.

Embeddings of Cubic Halin Graphs: An Inventory

14

Step 2. Join each red vertex v to all of the non-leaf vertices on the boundary of the face in whose boundary v lies.

Figure 3.2: Halin graph with all of its red edges.

Prop 3.1 The red and black edges together triangulate the region inside the exterior cycle of the Halin graph. ♦

Prop 3.2 Every black tree edge lies on two of the triangles formed by Steps 1 and 2. ♦

Embeddings of Cubic Halin Graphs: An Inventory

15

Step 3a. For each black tree edge, we pair the two incident triangles into a quadrangle. Step 3b. We assign (unseen) colors blue, green, and brown the to tree edges so as to form a proper edge 3-coloring. Step 3c. We visibly color each quadrangle with the unseen color of the tree edge that bisects it.

Figure 3.3: Quadrangulation of a plane Halin graph.

Remark Any tree of maximum degree 3 is edge-3-colorable (via greedy algorithm).

Embeddings of Cubic Halin Graphs: An Inventory

16

Figure 3.4: Separated quadrangles of a plane Halin graph.

Remark This quadrangulation works for any Halin graph, regardless of the degrees of its vertices. What’s Next?? §4 a puzzle: reassembling quadrangles into the Halin graph. §5 how this leads to a genus distribution. §6 the new set of productions for Halin graphs. §7 application to our running example of a Halin graph. §8 conclusions and open problems

Embeddings of Cubic Halin Graphs: An Inventory

17

4

u v w x

y

z

Quadrangulation puzzle for a plane cubic Halin graph H → S0 : 1. Each quadrangle Q is regarded as an initial fragment. 2. An RR-path on a fragment boundary is a • 2-path with two red edges, • from a red vertex • through a black vertex • to another red vertex. 3. Initially, all RR-paths are said to be live.

Embeddings of Cubic Halin Graphs: An Inventory

18

4. A legal move is initiated by choosing a vertex v such that • v is previously unchosen; • at least one fragment at v is a quadrangle; • all three RR-paths through v are live RR-paths. Then the three fragments that meet at v are merged into a single (larger) fragment. • If there is more than one live RR-path on the boundary of the merged fragment, then all but one of the live RRpaths are deemed to be dead. 5. You LOSE if you run out of legal moves before the map is fully reassembled.

This happens whenever there occurs an unmerged vertex w such that • there is a dead RR-path through w; • none of the fragments meeting at w is a quadrangle. 6. You WIN the game by reassembling the plane map.

Embeddings of Cubic Halin Graphs: An Inventory

19

Attempt 1. Start with a merger at v. Analysis. There are three live RR-paths on the boundary of the merged fragment.

u
live RR-path

y v x z

w

Figure 4.1: Attempt #1.

Result. LOSE, because RR-paths through two of the unmerged vertices u, w, x become dead.

Embeddings of Cubic Halin Graphs: An Inventory

20

Attempt 2. First choose u and then choose v. Analysis. There are two live RR-paths on the boundary of the merged fragment.

y u v
live RR-path

x z

w

Figure 4.2: Attempt #2.

Result. LOSE, because the RR-path through one of the unmerged vertices w, x becomes dead.

Embeddings of Cubic Halin Graphs: An Inventory

21

u v w x

y

z

Figure 4.3: Attempt #3.

Result. LOSE, since after there is a merger at v or x, there will be no quadrangle at the remaining unmerged vertex.

Embeddings of Cubic Halin Graphs: An Inventory

22

Solution: post-order traversal post-order traversal for a plane tree: trace the boundary of the only region and call out the name of a vertex whenever you see it for the last time.

u v w x

y

z

Figure 4.4: Post-order traversal.

z y x u v w

Embeddings of Cubic Halin Graphs: An Inventory

23

u v w x

y

z

1. As a root for the inscribed tree of the Halin graph, choose any leaf-vertex. (Must be a leaf to win.) 2. Choose vertices in the order in which they occur on a postorder traversal of the tree. SOLUTION: for a root at the lower left corner

SOLUTION: z y x u v w

Embeddings of Cubic Halin Graphs: An Inventory

24

z y x u v w

u v w x z

y

u v w x

y

z

u

u y v w x z
w v x

y

z

u v x

y

u v x

y

w

z

w

z

Figure 4.5: Solving the puzzle with post-order traversal.

Embeddings of Cubic Halin Graphs: An Inventory

25

5

Partials and Productions
Partial Genus Distributions

In a single-rooted graph (G, r) with deg(r) = 2, the root r lies either on two distinct fb-walks (type di ) or twice on the same fbwalk (type si ). We partition gi (G, r) into single-root partials: di (G, r) = the number of embeddings of type-di , and si (G, r) = the number of embeddings of type-si .
r

r

type-d0

type-s1

Figure 5.1: The types of single-root partials for a 2-valent root.

Example 5.1 d0 (B2 , r) = 4 s0 (B2 , r) = 0 g0 (B2 , r) = 4 Example 5.2 d0 (K4 , r) = 2 s0 (K4 , r) = 0 g0 (K4 , r) = 2 d1 (K4 , r) = 8 s1 (K4 , r) = 6 g1 (K4 , r) = 14 d1 (B2 , r) = 0 s1 (B2 , r) = 2 g1 (B2 , r) = 2

Embeddings of Cubic Halin Graphs: An Inventory

26

Double-Rooted Graphs When two single-rooted graphs are pasted together across their single roots, there is no surviving root. However, if two double-rooted graphs are pasted together with one root of one graph matched to one root of the other graph, then the merged graph has two roots. (G, b, c) ∗ H(d, e) −→ (X, b, e) This permits iterated amalgamations.

Figure 5.2: An iterated edge-amalgamation.

Example 5.3 By following iterated amalgamation with a selfamalgamation, we can obtain graphs like circular ladders and M¨bius ladders. o

Embeddings of Cubic Halin Graphs: An Inventory

27

Whereas a single-edge-rooted graph (G, c) has only two partials, namely di (G, c) and si (G, c) a double-edge-rooted graph (G, c, d) has the four partials ddi (G, c, d), dsi (G, c, d), sdi (G, c, d), and ssi (G, c, d) Example 5.4 (double-root partial g-dist for K4 )

Figure 5.3: K4 with two roots at midpoints of two edges.

i ddi dsi sdi ssi 0 2 0 0 0 1 4 4 0 2

gi 2 14

Table 5.1: Partial genus distribution for (K4 , u, v).

Remark The number of partials increases exponentially according to the number of roots. It follows that the number of roots employed by a method of genus distribution calculation must be limited.

Embeddings of Cubic Halin Graphs: An Inventory

28

There are three sub-partials within the dd-type: dd the two roots sharing two fb-walks; dd the two roots sharing one fb-walk; dd0 no shared fb-walks for the two roots.

type dd"

type dd'

type dd0
Figure 5.4: The three types of dd-subpartials.

Remark There are 10 double-root sub-partials for two edgeroots that both have 2-valent endpoints.

Remark The number of partials increases exponentially (also) with the degrees of vertex-roots and with degrees of the endpoints of the edge-roots, corresponding to a larger number of possible conﬁgurations of fb-walks at the root.

Embeddings of Cubic Halin Graphs: An Inventory

29

Deriving Productions with Recombinant Strands Example 5.5 ddi (G, b, c) ∗ ddj (H, d, e) −→ 4ddi+j (X, b, e) + 2ss2 i+j+1 (X, b, e)

b

b

b

dd"i

b c

e

e

e

dd'i+j

ss2 i+j

ss2 i+j

d

dd"j

e

b

b

b

e

e

e

dd'i+j

dd'i+j

dd'i+j

Figure 5.5: Derivation of a production for vertex-amalgamation.

Embeddings of Cubic Halin Graphs: An Inventory

30

6

Productions for Halin Graphs

We now introduce a new form of graph amalgamation, in which several graphs are merged in a cycle around a vertex. For cubic Halin graphs, we merge three graphs at a time, exactly as for the puzzle, so that one of them is a quadrangle Q = K4 − e.
A
r r' s v v t s' v t'

X Q
v

z

y

B

Figure 6.1: A three-way pie-merge (A, B, Q) → X.

Prop 6.1 In a pie-merge (A, B, Q) → X, each rotation system ρ for X is consistent with exactly two rotation systems for A and exactly two for B.

Remark In this oral presentation, we use the picture to deﬁne a pie-merge.

Remark A series of fundamental papers ([GrKhPo10], [Gr10a], [KhPoGr10], [PoKhGr10a], [PoKhGr10b], and [Gr10b]) derives the productions that correspond to various ways of synthesizing graphs from graphs whose partial genus distributions are known.

Embeddings of Cubic Halin Graphs: An Inventory

31

Six Relevant Partials For Halin graphs, we split gi into six double-rooted partials:
dd' dd" ds' sd' ss1 ss2

u

u

u

u

u

u

v

v

v

v

v

v

Figure 6.0: The 6 double-rooted partials for 3-way pie-merge.

Here is what they count: dd Each of the roots u and v lies on two distinct fb-walks. One and only one of these fb-walks traverses both roots. dd Each of the roots u and v lies on two distinct fb-walks. Both of these fb-walks traverse both roots. ds Root u lies on two distinct fb-walks. One of these fb-walks traverses root v twice. sd Root v lies on two distinct fb-walks. One of these fb-walks traverses root u twice. ss1 A single fb-walks traverses roots u and v twice. The occurrences of each root are consecutive. ss2 A single fb-walks traverses roots u and v twice. The occurrences of the two roots alternate.

Embeddings of Cubic Halin Graphs: An Inventory

32

Computational Complexity Thm 6.2 (a) The values of the six partials for the amalgamands A and B completely determine those for the result X of a pie-merge (A, B, Q) → X. (b) For |VA | = k and |VB | = m, there is an O(km)-time algorithm for calculating the partial genus distribution of the resulting graph X of a pie-merge (A, B, Q) → X. Cor 6.3 The post-order traversal using the 36 productions corresponding to the six partials yields an O(n2 ) algorithm for the genus distribution of a cubic Halin graph with n vertices.

Practical Calculation with Spreadsheet Express each of the six partial genus distributions of X as a sum of convolutions of partials of A and B, as prescribed by the productions. (See §7.)

Embeddings of Cubic Halin Graphs: An Inventory

33

Thirty-Six Productions 1. ddi ∗ ddj −→ ddi+j + 2ddi+j+1 + ss2 i+j+1 . 2. ddi ∗ ddj −→ 2ddi+j + 2ss2 i+j+1 . 3. ddi ∗ dsj −→ 2ddi+j + 2ss2 i+j+1 . 4. ddi ∗ sdj −→ 2sdi+j + 2ss1 i+j+1 . 5. ddi ∗ ss1 −→ 4sdi+j . j 6. ddi ∗ ss2 −→ 2dsi+j + 2sdi+j . j 7. ddi ∗ ddj −→ 2ddi+j + 2ss2 i+j+1 . 8. ddi ∗ ddj −→ 4ddi+j . 9. ddi ∗ dsj −→ 4dsi+j . 10. ddi ∗ sdj −→ 4sdi+j . 11. ddi ∗ ss1 −→ 4ss1 . i+j j 12. ddi ∗ ss2 −→ 2ddi+j−1 + 2ss2 . i+j j 13. dsi ∗ ddj −→ 2dsi+j + 2ss1 i+j+1 . 14. dsi ∗ ddj −→ 4dsi+j . 15. dsi ∗ dsj −→ 4dsi+j . 16. dsi ∗ sdj −→ 4ss1 . i+j 17. dsi ∗ ss1 −→ 4ss1 . j i+j 18. dsi ∗ ss2 −→ 2dsi+j−1 + 2ss1 . j i+j

Embeddings of Cubic Halin Graphs: An Inventory

34

19. sdi ∗ ddj −→ 2ddi+j + 2ss2 i+j+1 . 20. sdi ∗ ddj −→ 4sdi+j . 21. sdi ∗ dsj −→ 2ddi+j−1 + 2ss2 . i+j 22. sdi ∗ sdj −→ 4sdi+j . 23. sdi ∗ ss1 −→ 4sdi+j−1 . j 24. sdi ∗ ss2 −→ 2ddi+j−1 + 2ss2 . j i+j 25. ss1 ∗ ddj −→ 4dsi+j . i 26. ss1 ∗ ddj −→ 4ss1 . i+j i 27. ss1 ∗ dsj −→ 4dsi+j−1 . i 28. ss1 ∗ sdj −→ 4ss1 . i i+j 29. ss1 ∗ ss1 −→ 4ss1 i+j−1 . j i 30. ss1 ∗ ss2 −→ 4dsi+j−1 . j i 31. ss2 ∗ ddj −→ 2dsi+j + 2sdi+j . i 32. ss2 ∗ ddj −→ 2ddi+j−1 + 2ss2 . i+j i 33. ss2 ∗ dsj −→ 2ddi+j−1 + 2ss2 . i+j i 34. ss2 ∗ sdj −→ 2sdi+j−1 + 2ss1 . i i+j 35. ss2 ∗ ss1 −→ 4sdi+j−1 . i j 36. ss2 ∗ ss2 −→ 2ddi+j−1 + ddi+j−2 + ss2 i j i+j−1 .

Embeddings of Cubic Halin Graphs: An Inventory

35

Derivations of Pie-Merge Productions

Figure 6.1: Prod #1: ddi ∗ ddj −→ ddi+j + 2ddi+j+1 + ss2 i+j+1 .

Figure 6.2: Prod #18: dsi ∗ ss2 −→ 2dsi+j−1 + 2ss1 . i+j j

Embeddings of Cubic Halin Graphs: An Inventory

36

7

Sample Calculation
Merger at z

Graph A (K4 − e): i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 0 0 0 0 0 2 2 Graph B (K4 − e): i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 0 0 0 0 0 2 2 Use Productions 1, 6, 31, and 36. Merged Graph K4 : i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 0 4 4 4 0 2 14

Merger at y Just like the merger at z. Merged Graph: K4 i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 0 4 4 4 0 2 14

Embeddings of Cubic Halin Graphs: An Inventory

37

Merger at x Graph A (result from merger at z): i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 0 4 4 4 0 2 14 Graph B (result from merger at y): i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 0 4 4 4 0 2 14 Use 25 productions (all those without the partial ss1 ). Merged Graph: i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 40 4 12 12 0 2 70 2 0 16 48 48 32 40 184

Merger at u Just like the merger at z. Merged Graph: K4 i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 0 4 4 4 0 2 14

Embeddings of Cubic Halin Graphs: An Inventory

38

Merger at v Graph A (result from merger at x): i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 40 4 12 12 0 2 70 2 0 16 48 48 32 40 184 Graph B (result from merger at u): i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 0 4 4 4 0 2 14 Use 30 productions. Merged Graph: i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 112 4 28 12 0 2 158 2 544 96 544 352 80 112 1728 3 0 64 448 448 704 544 2208

Embeddings of Cubic Halin Graphs: An Inventory

39

Merger at w Graph A (result from merger at v): i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 112 4 28 12 0 2 158 2 544 96 544 352 80 112 1728 3 0 64 448 448 704 544 2208 Graph B (K4 − e): i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 0 0 0 0 0 2 2

Merged Graph: ﬁnal result i ddi ddi dsi sdi ss1 ss2 gi i i 0 2 0 0 0 0 0 2 1 144 4 60 4 0 2 214 2 1440 224 1632 224 56 144 3720 3 1024 1088 4800 1088 1088 1440 10528 4 0 0 0 0 896 1024 1920

Embeddings of Cubic Halin Graphs: An Inventory

40

8

Conclusions and Open Problems

New features of this solution: • quadrangulation • pie-merge (invented for application to Halin graphs) • post-order traversal (also used in [Gr11] and [PoKhGr11]) • dynamic assignment of roots to graphs • calculation for graphs of tree-width 3 Some open genus-distribution calculation problems: 1. wheels 2. planar cubic Hamiltonian graphs 3. truncated polyhedra

Embeddings of Cubic Halin Graphs: An Inventory

41

Some High-Level Theoretical Questions 1. shape of genus distributions • unimodality of genus distributions • unimodality of the partials • approximation by Stirling cycle numbers 2. average genus (without calculating the entire g.d.) The method presented here works for Halin graphs of bounded valence. Halin graphs are known [Wi87] to be of treewidth 3. Conjecture 8.1 There is a quadratic-time algorithm for the genus distribution of any cubic graph of treewidth 3.

Embeddings of Cubic Halin Graphs: An Inventory

42

References
[AHU83] A. V. Aho, J. E. Hopcroft, and J. D. Ullman, Data Structures and Algorithms, Addison-Wesley, 1983. [BHKY62] J. Battle, F. Harary, Y. Kodama, and J. W. T. Youngs, Additivity of the genus of a graph, Bull. Amer. Math. Soc. 68 (1962), 565–568. [BWGT09] L. W. Beineke, R. J. Wilson, J. L. Gross, and T. W. Tucker (editors), Topics in Topological Graph Theory, Cambridge University Press, 2009. [ChHu09] J. Chen and Y. Huang, Maximum genus, Chapter 2 of Topics in Topological Graph Theory (editors L. W. Beineke, R. J. Wilson, J. L. Gross, and T. W. Tucker), Cambridge University Press, 2009. [ChMaZo11] Y. Chen, T. Mansour, and Q. Zou, On the total embeddings for some types of graphs, manuscript. [FuGrMc88] M. Furst, J. L. Gross, and L. A. McGeoch, Finding a maximum-genus umbedding, J. ACM 35 (1988), 523–534. [FuGrSt89] M. L. Furst, J. L. Gross, and R. Statman, Genus distribution for two classes of graphs, J. Combin. Theory (B) 46 (1989), 22–36. [Gr09] J. L. Gross, Distribution of embeddings, Chapter 3 of Topics in Topological Graph Theory (editors L. W. Beineke, R. J. Wilson, J. L. Gross, and T. W. Tucker), Cambridge University Press, 2009. [Gr10a] J. L. Gross, Genus distribution of graph amalgamations, II. Self-pasting at 2-valent co-roots, Preprint (2010), 21pp.

Embeddings of Cubic Halin Graphs: An Inventory

43

[Gr10b] J. L. Gross, Genus distribution of graphs under surgery: adding edges and splitting vertices, New York J. Math. 16 (2010), 161–178. [Gr11] Genus distributions of cubic outerplanar graphs, preprint. [GrFu87] J. L. Gross and M. L. Furst, Hierarchy for imbeddingdistribution invariants of a graph, J. Graph Theory 11 (1987), 205–220. [GrKhPo10] J. L. Gross, I. Khan, and M. Poshni, Genus distribution of graph amalgamations, I: Pasting two graphs at 2-valent roots, Ars Combinatoria 94 (2010), 33–53. [GrKlRi93] J. L. Gross, E. W. Klein, and R. G. Rieper, On the average genus of a graph, Graphs and Combinatorics 9 (1993), 153–162. [GrRoTu89] J. L. Gross, D. P. Robbins, and T. W. Tucker, Genus distributions for bouquets of circles, J. Combin. Theory (B) 47 (1989), 292–306. [GrTu87] J. L. Gross and T. W. Tucker, Topological Graph Theory, Dover, 2001; original edn. Wiley, 1987). [Ja87] D. M. Jackson, Counting cycles in permutations by group characters, with an application to a topological problem, Trans. Amer. Math. Soc. 299 (1987), 785–801. [KhPoGr10] I. Khan, M. Poshni, and J. L. Gross, Genus distribution of graph amalgamations, III: Pasting when one root has arbitrary degree, Ars Mathematica Contemporanea 3 (2010), 121–138.

Embeddings of Cubic Halin Graphs: An Inventory

44

[KwLe93] J. H. Kwak and J. Lee, Genus polynomials of dipoles, Kyungpook Math. J. 33 (1993), 115–125. [KwLe94] J. H. Kwak and J. Lee, Enumeration of graph embeddings, Discrete Math. 135 (1994), 129–151. [McG87] L. A. McGeoch, Algorithms for two graph problems: computing maximum-genus imbedding and the two-server problem, PhD thesis, Carnegie-Mellon University, 1987. [MoTh01] B. Mohar and C. Thomassen, Graphs on Surfaces, Johns Hopkins University Press, 2001. [Ne81] L. Nebesk´, A new characterization of the maximum y genus of a graph, Czech. Math. J. 31 (1981), 604–613. [PoKhGr10a] M. Poshni, I. Khan, and J. L. Gross, Genus distribution of graphs under edge amalgamations, Ars Mathematica Contemporanea 3 (2010), 69–86. [PoKhGr10b] M. Poshni, I. Khan, and J. L. Gross, Edgeoperation eﬀect on genus distribution, II. Selfamalgamation on a pair of edges, Preprint (2010), 21 pages. [PoKhGr11] M. Poshni, I. Khan, and J. L. Gross, Genus distributions of 4-regular outerplanar graphs, Preprint (2011), 22 pages. [Ri90] R. G. Rieper, The enumeration of graph embeddings, Ph.D. thesis, Western Michigan University, 1990. ¨ [Ri55] G. Ringel, Uber drei kombinatorische Probleme am n-dimensionalen W¨rfel und W¨rfelgitter, u u Abh. Math. Sem. Univ. Hamburg 20 (1955), 10–19.

Embeddings of Cubic Halin Graphs: An Inventory

45

[Ri65] G. Ringel, Das geschlect des vollst¨ndingen paaren a Graphen, Abh. Math. Sem. Univ. Hamburg 28 (1965), 139– 150. [RiYo68] G. Ringel and J. W. T. Youngs, Solution of the Heawood map-coloring problem, Proc. Nat. Acad. Sci. USA 60 (1968), 438–445. [RoSe84] N. Robertson and P. D. Seymour, Graph minors, III. Planar tree-width, J. Combin. Theory 36 (1984), 40–64. [Sk91] M. Skoviera, The maximum genus of graphs of diameter two, Discrete Math. 87 (1991), 175–180. [Te00] E. H. Tesar, Genus distribution of Ringel ladders, Discrete Math. 216 (2000), 235–252. [Th89] C. Thomassen, The graph genus problem is NPcomplete, J. Algorithms. 10 (1989), 568–578. [Wh01] A. T. White, Graphs of Groups on Surfaces, Elsevier, 2001. [Wi87] T. V. Wimer, Linear algorithms on k-terminal graphs, Ph.D. Thesis in Computer Science, Clemson Univ., 1987. [Xu79] N. H. Xuong, How to determine the maximum genus of a graph, J. Combin. Theory (B) 26 (1979), 217–225.

Embeddings of Cubic Halin Graphs: An Inventory

46

9

Appendix
Derivations of the Productions

Figure 9.1: ddi ∗ ddj −→ ddi+j + 2ddi+j+1 + ss2 i+j+1 .

Figure 9.2: ddi ∗ ddj −→ 2ddi+j + 2ss2 i+j+1 .

Figure 9.3: ddi ∗ dsj −→ 2ddi+j + 2ss2 i+j+1 .

Embeddings of Cubic Halin Graphs: An Inventory

47

Figure 9.4: ddi ∗ sdj −→ 2sdi+j + 2ss1 i+j+1 .

Figure 9.5: ddi ∗ ss1 −→ 4sdi+j . j

Figure 9.6: ddi ∗ ss2 −→ 2dsi+j + 2sdi+j . j

Embeddings of Cubic Halin Graphs: An Inventory

48

Figure 9.7: ddi ∗ ddj −→ 2ddi+j + 2ss2 i+j+1 .

Figure 9.8: ddi ∗ ddj −→ 4ddi+j .

Figure 9.9: ddi ∗ dsj −→ 4dsi+j .

Embeddings of Cubic Halin Graphs: An Inventory

49

Figure 9.10: ddi ∗ sdj −→ 4sdi+j .

Figure 9.11: ddi ∗ ss1 −→ 4ss1 . i+j j

Figure 9.12: ddi ∗ ss2 −→ 2ddi+j−1 + 2ss2 . i+j j

Embeddings of Cubic Halin Graphs: An Inventory

50

Figure 9.13: dsi ∗ ddj −→ 2dsi+j + 2ss1 i+j+1 .

Figure 9.14: dsi ∗ ddj −→ 4dsi+j .

Figure 9.15: dsi ∗ dsj −→ 4dsi+j .

Embeddings of Cubic Halin Graphs: An Inventory

51

Figure 9.16: dsi ∗ sdj −→ 4ss1 . i+j

Figure 9.17: dsi ∗ ss1 −→ 4ss1 . i+j j

Figure 9.18: dsi ∗ ss2 −→ 2dsi+j−1 + 2ss1 . i+j j

Embeddings of Cubic Halin Graphs: An Inventory

52

Figure 9.19: sdi ∗ ddj −→ 2ddi+j + 2ss2 i+j+1 .

Figure 9.20: sdi ∗ ddj −→ 4sdi+j .

Figure 9.21: sdi ∗ dsj −→ 2ddi+j−1 + 2ss2 . i+j

Embeddings of Cubic Halin Graphs: An Inventory

53

Figure 9.22: sdi ∗ sdj −→ 4sdi+j .

Figure 9.23: sdi ∗ ss1 −→ 4sdi+j−1 . j

Figure 9.24: sdi ∗ ss2 −→ 2ddi+j−1 + 2ss2 . i+j j

Embeddings of Cubic Halin Graphs: An Inventory

54

Figure 9.25: ss1 ∗ ddj −→ 4dsi+j . i

Figure 9.26: ss1 ∗ ddj −→ 4ss1 . i+j i

Figure 9.27: ss1 ∗ dsj −→ 4dsi+j−1 . i

Embeddings of Cubic Halin Graphs: An Inventory

55

Figure 9.28: ss1 ∗ sdj −→ 4ss1 . i i+j

Figure 9.29: ss1 ∗ ss1 −→ 4ss1 i+j−1 . j i

Figure 9.30: ss1 ∗ ss2 −→ 4dsi+j−1 . j i

Embeddings of Cubic Halin Graphs: An Inventory

56

Figure 9.31: ss2 ∗ ddj −→ 2dsi+j + 2sdi+j . i

Figure 9.32: ss2 ∗ ddj −→ 2ddi+j−1 + 2ss2 . i+j i

Figure 9.33: ss2 ∗ dsj −→ 2ddi+j−1 + 2ss2 . i+j i

Embeddings of Cubic Halin Graphs: An Inventory

57

Figure 9.34: ss2 ∗ sdj −→ 2sdi+j−1 + 2ss1 . i i+j

Figure 9.35: ss2 ∗ ss1 −→ 4sdi+j−1 . j i

Figure 9.36: ss2 ∗ ss2 −→ 2ddi+j−1 + ddi+j−2 + ss2 i+j−1 . j i