TOPICS
Search

Graph Bandwidth


The bandwidth of a connected graph G is the minimum matrix bandwidth among all possible adjacency matrices of graphs isomorphic to G. Equivalently, it is the minimum graph dilation of a numbering of a graph. Bandwidth is variously denoted bw(G), B(G), or phi(G).

The bandwidth of the singleton graph is not defined, but the conventions bw(K_1)=0 or bw(K_1)=1 (Miller 1988) are sometimes adopted.

The bandwidth of a disconnected graph is the maximum of the bandwidths of its connected components.

The bandwidth bw(G) of a connected graph G satisfies the inequalities

 [(n-1)/(diam(G))]<=bw(G)<=n-diam(G)
(1)

(Chinn et al. 1982), where n=|G| is the vertex count of G and G is the graph diameter and

 bw(G)>=chi(G)-1,
(2)

where chi(G) is the chromatic number.

Every cubic graph that is also a triangle-free graph has bandwidth at least 4. This can be shown by contradiction. Suppose that G has an ordering of bandwidth at most 3, and let v be its first vertex. The three neighbors of v must occupy positions 2, 3, and 4. If u is the vertex in position 2, triangle-freeness implies that u is adjacent to neither of the vertices in positions 3 and 4. Apart from v, the two remaining neighbors of u must therefore both occupy position 5, which is impossible.

The wheel graph W_n on n>=4 vertices has bandwidth

 bw(W_n)=max(3,|_n/2_|).
(3)

For the lower bound, placing the hub in position p gives bandwidth at least max(p-1,n-p)>=|_n/2_|. This bound is at least 3 for n>=6, while W_4=K_4 has bandwidth 3. If W_5 had bandwidth at most 2, its hub would occupy position 3. The rim vertex in position 1 would then have only the rim vertex in position 2 within distance 2, contradicting its two neighbors on the rim. For the upper bound, take a bandwidth-2 ordering of the rim cycle graph and insert the hub in position [n/2]. Hub edges then have length at most |_n/2_|, while rim edges have length at most 3.

Computing the bandwidth of a graph is NP-hard.

Bounds for the bandwidth of a graph have been considered by (Harper 1964), and the bandwidth of the k-cube was determined by Harper (Harper 1966, Wang and Wu 2007, Harper 2010).

Special cases are summarized in the following table.


See also

Bandwidth, Broadcast Time, Gossiping, Graph Dilation, Pathwidth, Treewidth

Explore with Wolfram|Alpha

References

Böttcher, J.; Preussmann, K. P.; Taraz, A.; and Würfl, A. "Bandwidth, Expansion, Treewidth, Separators and Universality for Bounded-Degree Graphs." Eur. J. Combin. 31, 1217-1227, 2010.Chinn, P. Z.; Chvátalová, J.; Dewdney, A. K.; and Gibbs, N. E. "The Bandwidth Problem for Graphs and Matrices--A Survey." J. Graph Th. 6, 223-254, 1982.Chvátalová, J. "Optimal Labelling of a Product of Two Paths." Disc. Math. 11, 249-253, 1975.Harper, L. H. "Optimal Assignments of Numbers to Vertices." J. Soc. Indust. Appl. Math. 12, 131-135, 1964.Harper, L. H. "Optimal Numberings and Isoperimetric Problems on Graphs." J. Combin. Th. 1, 385-393, 1966.Harper, L. H. Global Methods for Combinatorial Isoperimetric Problems. Cambridge, England: Cambridge University Press, 2010.Miller, Z. "A Linear Algorithm for Topological Bandwidth with Degree-Three Trees." SIAM J. Comput. 17, 1018-1035, 1988.Wang, X. and Wu, X. "Recursive Structure and Bandwidth of Hales-Numbered Hypercube." 27 Aug 2007. https://arxiv.org/abs/0708.3628.West, D. B. Introduction to Graph Theory, 2nd ed. Englewood Cliffs, NJ: Prentice-Hall, p. 390, 2000.

Referenced on Wolfram|Alpha

Graph Bandwidth

Cite this as:

Weisstein, Eric W. "Graph Bandwidth." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/GraphBandwidth.html

Subject classifications