3-trees with diameter 2

Allan Bickle1
1Department of Mathematics, Purdue University 610 Purdue Mall, West Lafayette, IN 47907 USA

Abstract

A k-tree is a graph that can be formed by starting with Kk+1 and iterating the operation of making a new vertex adjacent to all the vertices of a k-clique of the existing graph. A structural characterization of 3-trees with diameter at most 2 is proven. This implies a corollary for planar 3-trees which leads to a description of their degree sequences.

Keywords: k-tree, Diameter, Planar graph, Degree sequence

1. Introduction

In this paper, we seek a structural (constructive) characterization of 3-trees with diameter at most 2.

Definition 1.1. A k-tree is a graph that can be formed by starting with Kk+1 and iterating the operation of making a new vertex adjacent to all the vertices of a k-clique (the root) of the existing graph. A deletion sequence of a graph G is an ordering v1,,vn of V(G) such that each vi has minimum degree in the induced subgraph G[{vi,vi+1,,vn}].

A k-leaf is a degree k vertex of a k-tree.

See [5] for a survey of results on k-trees. There are many results describing and characterizing the structure of k-trees. Graphs with diameter 2 have been studied in relation to many other graph classes, such as cages and planar graphs [11].

Definition 1.2. The distance between vertices u and v, d(u,v), is the length of a shortest uv path. The diameter of a graph G is the maximum distance between any pair of vertices in G.

A k-tree has diameter 1 if and only if it is Kk+1. For 2-trees, the following theorem was proved in [6].

Proposition 1.3. [6]The following are equivalent for a 2-tree G:

1. G is has diameter at most 2.

2. G does not contain P62.

3. G is T+K1 for any tree T, or any graph formed by adding any number of vertices adjacent to pairs of vertices of K3.

Note that 12 is a forbidden subgraph characterization, while 13 is a structural (constructive) characterization. In [4], Proposition 1.3 was generalized to a structural characterization of maximal 2-degenerate graphs with diameter 2. In [2], a forbidden subgraph characterization was found for k-trees with diameter d2. Thus the next natural questions are to find a structural characterization of 2-trees with diameter 3 and 3-trees with diameter 2.

Definitions of terms and notation not defined here appear in [3]. In particular, n(G) is the number of vertices of a graph G. The neighborhood of a vertex v is denoted N(v), and the closed neighborhood is denoted N[v]. The square G2 is formed by adding all edges between pairs of vertices with distance 2 in G. The join of graphs G and H is denoted G+H.

2. Preliminaries

One way for a k-tree to have diameter at most 2 is for there to be a vertex adjacent to all other vertices.

Definition 2.1. A dominating vertex of a graph is a vertex adjacent to all other vertices. When constructing a k-tree, we duplicate a k-leaf by adding another k-leaf with the same neighborhood.

The following observations should be immediate.

Lemma 2.2. Let T be a k-tree with diameter at least 2.

a. Adding a k-leaf to T cannot reduce the diameter.

b. Duplicating a k-leaf arbitrarily many times will not change the diameter.

Proposition 2.3. A k-tree has diameter at most 2 if and only if any two k-leaves of G have a common neighbor.

Proof. A k-tree G has diameter at most 2 if and only if the distance between any two vertices of G is at most 2. By Lemma 2.2, this will be the case if and only if any two k-leaves are at distance at most 2. This will hold if and only if any two k-leaves of G have a common neighbor. ◻

Definition 2.4. A k-path graph G is an alternating sequence of distinct k– and k+1-cliques e0,t1,e1,t2,,tp,ep, starting and ending with a k-clique and such that ti contains exactly two k-cliques ei1 and ei.

For order n>k+1, k-paths are just the k-trees with exactly two k-leaves [10]. See Figures 1 and 2 for examples of k-paths.

Figure 1 The 2-trees of order 5 and 6 are shown above. Those in the rst column are 2-paths. The one in the second column is outerplanar but not a 2-path. The rest are not outerplanar
Figure 2 The 3-trees with order 7. The leftmost two are 3-paths, and the leftmost three are maximal planar
Lemma 2.5. A graph T of order n>k+1 is a k-tree if and only if T+K1 is a (k+1)-tree. Moreover, T is a k-path if and only if T+K1 is a (k+1)-path.

Proof. () Any k-tree T has a deletion sequence v1vn so that d(vi)=max{k,ni} when vi is deleted. Joining a vertex x to T results in a graph T with a deletion sequence v1vnx so that d(vi)=max{k+1,n+1i} when vi is deleted. Thus T is a (k+1)-tree.

() Let T+K1 have the K1 denoted x. Then T+K1 has a deletion sequence v1vnx so that d(vi)=max{k+1,n+1i} when vi is deleted. Thus T has a deletion sequence v1vn so that d(vi)=max{k,ni} when vi is deleted, so T is a k-tree.

The proof for k-paths is essentially the same. ◻

3. 2-trees with diameter 3

In this section, we characterize 2-trees with diameter at most 3.

Definition 3.1. A dominating triple is three vertices {x,y,z} that form a triangle of a 2-tree T so that any 2-leaf of T is adjacent to at least one of them. A private neighbor of x (in a dominating triple) is adjacent to x, but not y or z.

A common triple is three vertices {x,y,z} that form a triangle of a 2-tree T so that any 2-leaf of T is adjacent to at least two of them.

Theorem 3.2. A 2-tree T has diameter at most 3 if and only if T has a dominating triple.

Proof. () If T has a dominating triple, then there is a path of length at most 3 between any two vertices of T.

() Suppose that T has diameter at most 3. The result is obvious for diameter 1 or 2, so suppose T has diameter 3.

We use induction on n. Assume the result holds for 2-trees with order n, and let T have order n+1, and 2-leaf v. Now Tv has diameter at most 3, so it has a dominating triple t={x,y,z}. If v is adjacent to any of its vertices, T also has a dominating triple and we are done. Thus we assume that T has no dominating triple with a vertex adjacent to v.

Deleting two vertices of t (say x and y) will disconnect v from the third (z). Thus there is a vertex w adjacent to x and y in the same component of Txy as v. We may assume that z has no private neighbor a, since else d(v,a)=4. But then {x,y,w} is also a dominating triple. Thus by our assumption, v is not adjacent to w. Say d(v,x)=2. Then y has no private neighbor b, since else d(v,b)=4. But then x is a dominating vertex of Tv. Let N(v)={u1,u2}. Then T has a dominating triple {x,u1,u2}◻

A fan is Pr+K1, where Pr is a path. Call K1 the center of the fan.

Proposition 3.3. A 2-tree T has a dominating triple t={x,y,z} if and only if it has a covering by fans centered at the three vertices of t.

Proof. () If this holds, any vertex of T is adjacent to a vertex of t.

() Let T have a dominating triple t={x,y,z}. Let v be a vertex not in t, so v is adjacent to a vertex x of t. If v is adjacent to two vertices of t, it is contained in a fan centered at x. Else v is adjacent to x and a vertex u not in t. Now u is adjacent to x, and the argument can be repeated, producing a fan centered at x◻

4. 3-trees with diameter 2

To characterize 3-trees with diameter 2, we use the strategy of starting with a 3-tree with a dominating vertex, and then considering what can be added while maintaining diameter 2.

Definition 4.1. A k-fan Fk,r is Kk1+Pr. Call the K2 in a 3-fan its base.

Thus a 2-fan is just a fan. Any k-fan is a k-path, and hence also a k-tree. Any 3-fan is maximal planar (see Figure 2).

We may be able to identify a triangle of a 3-fan with a triangle of a 3-tree (with the base as one of the identified edges) while maintaining diameter 2. Call this operation fan overlapping. Fan overlapping produces only 3-trees since identifying k-cliques of two k-trees produces another k-tree.

Theorem 4.2. Let T be 3-tree. Then T has diameter at most 2 if and only if it is formed in one of the following ways.

1. T=H+K1, where H is a 2-tree.

2. Let K4 have vertices {u,x,y,z}. Then T is formed by fan overlapping, where the base of the fan must be ux, uy, or xy, and adding 3-leaves with root {x,y,z}.

3. Let uxy be the K3 in K3+K¯r, r1. Then T is formed by fan overlapping, where the base of the fan must be ux, uy, or xy.

Proof. () Clearly each construction produces a 3-tree. In Case 1, there is a dominating vertex. In Case 2, every pair of vertices not in {u,x,y,z} has a neighbor in {u,x,y,z}. In Case 3, every pair of vertices not in {u,x,y} has a neighbor in {u,x,y}. Thus each 3-tree has diameter at most 2.

() Assume the hypotheses. Let u have maximum degree in T, S=V(T)N[u], and H=N(u). Now H is a 2-tree [7], so TS is a 3-tree. Thus if u is a dominating vertex, Tu is a 2-tree by Lemma 2.5. Thus we assume T has no dominating vertex, so S is nonempty.

Clearly, every vertex in S neighbors a vertex in H. Let R be all vertices in H with neighbors in S. Every vertex in R is contained in a triangle of H, and each pair of these triangles must have a have a nonempty intersection, since else two vertices of S have no common neighbor. Then R is a union of triangles, and the graph induced by R has diameter 2. It is contained in a minimal 2-tree T which has diameter 2, so by Proposition 1.3, T has a dominating vertex or a common triple.

Suppose T has a dominating vertex x. Now H must have diameter 2, since else some vertex in S would be more than 2 away from a vertex in H. If x is dominating in H, x is also dominating in T. By assumption, we can exclude this case. Thus H has a common triple, so T does also.

Next we assume that T=K3, whose vertices are {x,y,z}, none of which is dominating in H. There is at least one vertex v in S whose neighbors are T. Then any other vertex in H is adjacent to a vertex in T and u. Then T is formed by fan overlapping with bases ux, uy, or xy, and adding at least one 3-leaf with root {x,y,z}.

Next we assume that T=K2+K¯s, the vertices of K2 are {x,y}, neither of which is dominating in H. Then for each vertex w in the K¯s, there is at least one vertex in S not adjacent to it (else we return to the previous case). Then every vertex of T not in D={u,x,y} is adjacent to at least two vertices in D. Thus every vertex not in D is part of a 3-fan with base ux, uy, or xy, and there is at least one vertex of T adjacent to all vertices of D.

Finally, we assume T contains a triangle xyz, any other vertex of T is adjacent to exactly two of {x,y,z}, and each pair (xy, xz, and yz) has at least one additional neighbor in T. Now each vertex of T is adjacent to at least one vertex in S, so H=T. Thus each vertex in S is adjacent to at least two vertices in {x,y,z}. Thus each vertex in T is adjacent to at least two vertices in {x,y,z}. But then we can return to the previous case by giving {x,y,z} the roles of {u,x,y}◻

This characterization allows us to evaluate or bound parameters of 3-trees with diameter 2. In the following results, we refer to the three graph classes in the statement of Theorem 4.2 as Cases 1, 2, and 3.

Corollary 4.3. A 3-tree with diameter 2 and order n5 and maximum degree Δ has n5Δ53.

Proof. In Case 1, a 3-tree with a dominating vertex has Δ=n1, so n=Δ+1.

In Case 2, for each vertex v{u,x,y,z}, let Sv be the set of vertices not adjacent to v. To have Δ(T)=n1r, each Sv must contain at least r vertices. Now vertices in Su are adjacent to triangle xyz, so they are only in Su.

The vertices in Sx may be in Sy or Sz (not both), and similarly for the vertices in Sy or Sz. However, there must be at least one vertex only in one of the sets Sx, Sy, or Sz. If there is exactly one vertex adjacent to (say) {u,y,z}, then Δ(T)=n1r requires at least r vertices each in Sy and Sz, and none in both. Thus n4+3r+1=3r+5 and Δ(T)n1n53=23n+23, so n3Δ22.

Suppose there are two vertices in only one of the sets Sx, Sy, or Sz, say one each in sets Sx and Sy. Any other vertex can be in any two of the three sets. Then s vertices in SxSySz yield |Sx||Sy||Sz|2s2, so r2s23. Now n=4+s+r4+3r+22+r=5r+102, so r2n105. Then Δ(T)n12n105=3n5+1. Thus n5Δ53.

In Case 3, K3+K¯r has r1, with uxy being the K3. There are at most n4 vertices with exactly two neighbors in K3+K¯r. These vertices split into three sets based on which of {u,x,y} they are not adjacent to. When Δ is minimum, one of these sets contains at least n43 vertices. Then Δn1n43=2n+13, so n3Δ12◻

The smallest possible Δ for a 3-tree with diameter 2 and order n is n1 for 3n7 and 3n5+1 for n5.

5. Planar 3-trees

Next we consider an important special class of k-trees.

Definition 5.1. A simple k-tree is defined recursively by starting with Kk+1 and iteratively adding a vertex adjacent to all vertices of a k-clique Q not previously used as the neighborhood of a k-leaf.

A plane drawing of a graph is a drawing in the plane that has no crossings. A graph is outerplanar if it has a plane drawing with all vertices on the boundary of the exterior region. A graph is a maximal outerplanar graph (MOP) if no edge can be added so that the resulting graph is still outerplanar.

An Apollonian network is a planar 3-tree.

The MOPs are exactly the simple 2-trees, and the planar 3-trees are exactly the simple 3-trees [10]. See Figures 1 and 2 for examples of these graphs.

Corollary 5.2. Let T be planar 3-tree. Then T has diameter at most 2 if and only if it is formed in one of the following ways.

1. T=H+K1, where H is a MOP.

2. Let K4 have vertices {u,x,y,z}. Then T is formed by fan overlapping with bases ux, uy, uz, and only triangles of K4 are used for overlapping, each at most once. A single 3-leaf may be added with root {x,y,z}.

3. Let uxy be the K3 in K3+K¯r, 1r2. Then T is formed by fan overlapping with bases ux, uy, or xy. Only triangles of K3+K¯r are used for overlapping, and each at most once.

Proof. In Case 1, for T to be planar, H must be outerplanar.

In Cases 2 and 3, for T to be planar, it must be a simple 3-tree, so each root is used at most once. Thus each triangle of K4 or K3+K¯r can be used at most once for overlapping, and no other triangle can be used for overlapping. In Case 3, r2, since K3+K¯3 is not planar. ◻

Seyffarth [11] studied maximal planar graphs with diameter 2. Seyffarth showed that such graphs have n32Δ+1 and found two infinite classes of maximal planar graphs that show this bound is sharp. This is not claimed to be a complete characterization.

Of course, maximal planar graphs with diameter 2 need not be 3-trees. For example, the double wheel K¯2+Cn2 has minimum degree 4 and diameter 2. Seyffarth’s two classes both contain subgraphs with minimum degree 4. Thus it appears that the bound on n can be improved when we only consider planar 3-trees.

Corollary 5.3. A planar 3-tree with diameter 2 with order n4 and maximum degree Δ has n32Δ12.

Proof. In Case 1, a 3-tree with a dominating vertex has Δ=n1, so n=Δ+1.

In Case 2, there can only be one vertex not adjacent to u, so Δn2, and nΔ+2.

In Case 3, K3+K¯r has 1r2, with uxy being the K3. There are at most n4 vertices with exactly two neighbors in K3+K¯r. These vertices split into three sets based on which of {u,x,y} they are not adjacent to. When Δ is minimum, one of these sets contains at least n43 vertices. Then Δn1n43=2n+13, so n3Δ12◻

Thus no planar 3-tree can be an extremal graph for Seyffarth’s theorem.

We may be interested to characterize the degree sequences of planar 3-trees with diameter 2. Note that in Case 1, G has a dominating vertex u if and only if Gu is a MOP. Consequently, we can determine whether a list of numbers is a degree sequence of a planar 3-tree with a dominating vertex if and only if we can determine whether a corresponding list is the degree sequence of a MOP. However, no characterization of degree sequences of MOPs is known. See [1,8,9] for partial results. Thus we instead consider graphs for Cases 2 and 3 that are not covered by Case 1.

Case 2: To avoid a dominating vertex, we assume there is a vertex rooted on each triangle of the K4 with vertex set {u,x,y,z}. We designate the six triangles A-F in order around u (see the graph above). We can break down the cases by how many vertices are in each of the 6 triangles. Note that if there are no vertices in D, we can move the vertices in C to B without changing the degree sequence.

We organize cases based on how many degree 5 vertices there are rooted on the K4. We can reduce the cases to possibly adding vertices inside ACE, ABD, ABCD, or ABCDEF. Suppose a vertices are added inside A, and similarly for the other triangles. We require a,b1 when A and B are both listed in a case, and similarly for the pairs {C,D} and {E,F}, but not otherwise. We obtain the following possible degree sequences (dr indicates r vertices of degree d).

Case 3: To avoid a dominating vertex, there must be fans attached to at least one of each of the three pairs {A,D}, {B,E}, and {C,F} (see the graph above). Thus the triangles where fans are attached are (up to symmetry) ABC, ABF, ABCD, ABDF, ABCDE, or ABCDEF. Suppose a1 vertices are added inside A, and similarly for the other triangles. We obtain the following possible degree sequences.

Triangles Degree Sequence
ABC (4+a+b)1(4+a+c)1(4+b+c)1614a+b+c334
ABF (4+a+b)1(4+a+f)1(4+b+f)1514a+b+f233
ABCD (4+a+b+d)1(4+a+c+d)1(4+b+c)1614a+b+c+d334
ABDF (4+a+b+d)1(4+a+d+f)1(4+b+f)1524a+b+d+f434
ABCDE (4+a+b+d+e)1(4+a+c+d)1(4+b+c+e)161514a+b+c+d+e535
ABCDEF (4+a+b+d+e)1(4+a+c+d+f)1(4+b+c+e+f)1624a+b+c+d+e+f636

For example, suppose we have the degree sequence S=8261524134. We see n=10 and di=48=2(3n6), so we have the right degree sum for a 3-tree [5]. The 34 and 52 shows it must fall under Case 3, subcase ABDF. Then a+b+d+f=5, so a=2 and b=d=f=1. Thus S is the degree sequence of a 3-tree with diameter 2.

Acknowledgements

Thanks to Zhongyuan Che for discussing this problem with me, and particularly for helping to develop Proposition 2.3, Lemma 2.5, and the argument in the proof of Theorem 4.2.

References:

  1. A. Bar-Noy, T. Böhnlein, D. Peleg, Y. Ran, and D. Rawitz. Approximate realizations for outerplanaric degree sequences. Journal of Computer and System Sciences, 148:103588, 2025. https://doi.org/10.1016/j.jcss.2024.103588.
  2. A. Bickle. k-paths of k-trees. In Springer Proc. Math. Stat., Volume 388, pages 287–291, 2020.
  3. A. Bickle. Fundamentals of Graph Theory. AMS, 2020.
  4. A. Bickle. Maximal k-degenerate graphs with diameter 2. International Journal of Mathematical Combinatorics, 2:67–78, 2021.
  5. A. Bickle. A survey of maximal k-degenerate graphs and k-trees. Theory and Applications of Graphs, 0(1):Article 5, 2024. 10.20429/tag.2024.000105.
  6. A. Bickle and Z. Che. Wiener indices of maximal k-degenerate graphs. Graphs and Combinatorics, 37(2):581–589, 2021. https://doi.org/10.1007/s00373-020-02264-8.
  7. R. Froberg. A characterization of k-trees. Discrete Mathematics, 104(3):307–309, 1992. https://doi.org/10.1016/0012-365X(92)90452-L.
  8. D. Y. Li and J. Z. Mao. Degree sequences of maximal outerplanar graphs. J Central China Normal Univ Natur Sci, 26(3):270–273, 1992.
  9. Z. Li and Y. Zuo. On the degree sequences of maximal outerplanar graphs. Ars Combinatoria, 140:237–250, 2018.
  10. L. Markenzon, C. M. Justel, and N. Paciornik. Subclasses of k-trees: characterization and recognition. Discrete Applied Mathematics, 154(5):818–825, 2006. https://doi.org/10.1016/j.dam.2005.05.021.
  11. K. Seyffárth. Maximal planar graphs of diameter two. Journal of Graph Theory, 13(5):619–648, 1989.