\tbGraphs with prescribed radius, diameter, and center
Abstract
Among other things, it is shown that for every pair of positive integers , , satisfying , and every finite simple graph there is a connected graph with diameter , radius , and center
Kelly Guest
Tuskegee University, [email protected]
Andrew Johnson
Kennesaw State University, [email protected]
Peter Johnson
Auburn University, [email protected]
William Jones
State University of New York at Binghamton, [email protected]
Yuki Takahashi
Grinnell College, [email protected]
Zhichun (Joy) Zhang
Swarthmore College, [email protected]
Key words and phrases: distance in graph, eccentricity of a vertex, radius/diameter of a connected graph, central vertex, center of a connected graph.
AMS Subject Classification: 05C12
1 Introduction
All graphs referred to will be finite and simple. The vertex and edge sets of a graph will be denoted and , respectively. If is connected and , is the length of a shortest walk in from one of to the other; a geodesic under the shortest-walk metric. As every shortest walk is a path, may also be formulated as the length of a shortest path in with end-vertices and .
If is connected and , the eccentricity of in , denoted , is:
The radius of a connected graph is:
and its diameter is:
Equivalently,
It is easy to see that . It is a standard exercise in a first course in graph theory to show that for any positive integers satisfying , there is a connected graph such that and . (A more challenging, but still elementary, exercise would be to determine, for pairs constrained as above, the values of such that there exists a connected graph with , , and .)
A vertex is a central vertex in if and only if . The center of , denoted , is the subgraph of induced by the set of centers of . (Therefore, that set is .)
The question broached in [1] is: which graphs can be installed as the center of another graph? That is, given a graph , can you find a connected graph such that ?
As reported in [1], this question in full generality was killed at its birth as a question meriting research by a brilliant observation of S. Hedetniemi (Steve? Sandra?), encapsulated in Figure 1.

The authors of [1] resurrect the problem by asking: for a distinguished family of connected graphs, which graphs can be the center of a graph ? And, for such and , how small can be, if ? These questions have borne fruit, but we are going in a different direction.
The graph in Figure 1 has diameter 4 and radius 2. The set of central vertices of is precisely , regardless of what is. If the paths leading away from from and are each lengthened to have length , the result is a graph with center , radius , and diameter .
Our aim here is to answer the question: for which positive integers , satisfying , and graphs , does there exist a connected graph such that , , and ? The extension of the observation of Hedetniemi just above shows that there is such a for every , , and . Our main result, in Section 3, is that there is such a for every , , and . In the next section we deal with extremes, and alternative solutions to that in Section 3, in some cases.
2 Extremes and alternative solutions
2.1
If , then is its own center. Therefore, and if and only and .
2.2
If , then each central vertex of is adjacent to every other vertex of . Therefore, if then must be a complete graph, and each vertex of must be adjacent to each vertex of . Furthermore, since all central vertices of are in , it must be that every has a non-neighbor in in .
Let “” stand for the join of two graphs: is formed by taking disjoint copies of and and then adding in every edge , . By the paragraph above, when , , the only for which a solution can exist are , and the only possible solutions are in which is a graph with and for each , the degree deg of satisfies deg.
Every such satisfies , , and , so we have completely characterized the values of for which our problem with has a solution, and all possible solutions (, as above).
2.3 A standard method
Proposition 1.
Suppose that is a connected graph with , , and ; i.e., there is a single central vertex in . For an arbitrary graph , if is formed by replacing by , with every vertex of adjacent in to every vertex in to which is adjacent, then , , and .
The proof is straightforward. Note that the assumption that plays a role in the proof that .
For instance, the graph in Figure 1 is obtained from , the path on 5 vertices, by the device of Proposition 1. The generalization to the solution of our problem for all when uses the device of Prop. 1 with .
In Figure 2 we have a graph with a single central vertex such that , , for arbitrary . By Proposition 1, this shows that every graph can be the center of a graph of radius and diameter , for every .

For those who enjoy variety, we can vary to the graph shown in Figure 3, which gives another solution to our problem when and arbitrary.

If you have been paying attention, you might exclaim: why do we need this? Hedetniemi’s construction already gives us solutions of our problem in the case . Yes, bur Figure 3 gives a different solution, and different solutions of our problem contribute to the solution of a problem that towers over ours: given positive integers and satisfying , and a graph , find all possible graphs satisfying , , and . In view of Proposition 1, in pursuit of this larger problem, it is appropriate to pose the following: given and as above, find all graphs such that , , and .
Moreover, the alternative solutions to the case provide a related problem: what properties characterize those graphs with and center ? The majority of graphs constructed with center in fact had , and the solution to this problem will considerably narrow down the larger problem.
In Figure 4, we have, for , a graph of radius and diameter , and a graph of radius and diameter , both with a single central vertex.

2.4 A non-standard strategy in special cases
The strategy referred to, applicable only when is connected is: attach pairwise vertex-disjoint paths to the vertices of . This trick appears to be of use only in a special class of cases.
Proposition 2.
Suppose that is connected with . Suppose that is formed by attaching vertex-disjoint paths to the vertices of , with each vertex of being an end of its attached path (when , nothing is attached, and ). Then , , and .
The proof is straightforward.
Corollary 1.
If is as in Proposition 2 then for all integers and there is a graph , obtained as in Prop. 2 with such that , , and .
3 The main result
Lemma 1.
Let be the graph depicted in Figure 5. Suppose that and . Then is the unique central vertex of , , and .

Proof.
Clearly . Checking shows that every other vertex of has eccentricity in . For instance,
Finally, it is easy to see that the vertices , , , have the greatest eccentricity; for instance, . ∎
Theorem 1.
For all integers and satisfying and every graph there is a graph such that , , and . Furthermore, is obtainable from some graph by the method of Proposition 2.
Acknowledgement
We greatly thank Timothy Eller for drawing the graphs for us.
References
- [1] Fred Buckley, Zevi Miller, and Peter J. Slater. On graphs containing a given graph as center. Journal of Graph Theory 5(1981), 427-434.