The utility problem posits three houses and three utility companies--say, gas, electric, and water--and asks if each utility can be connected to each house without having any of the gas/water/electric lines/pipes pass over any other. This is equivalent to the equation "Can a planar graph be constructed from each of three nodes ('houses') to each of three other nodes ('utilities')?" This problem was first posed in this form by H. E. Dudeney in 1917 (Gardner 1984, p. 92).
The answer is that no such planar graph exists, and the proof can be effected using the Jordan curve theorem, while a more general result encompassing this one is the Kuratowski reduction theorem. The utility graph is the graph showing the relationships described above, also known as the Thomsen graph (e.g., Coxeter 1950) and, in the more formal parlance of graph theory, is known as the complete bipartite graph (and is also equivalent to the circulant graph ).
It is implemented in the Wolfram Language as GraphData["UtilityGraph"].
The utility graph has graph crossing number 1, with a minimal crossing embedding illustrated as the rightmost figure above.
A simple proof of the nonplanarity of the utility graph can be effected by noting that the graph consists of a graph cycle , to which the three edges , , and must be added. Now, for each of the edges, we have choose whether to draw the edge inside or outside the graph cycle, and so for two of the edges, we must make the same choice. But two lines can't be drawn on the same side without crossing, hence the graph is not planar.
It can be represented using LCF notation as . The utility graph is an integral graph with graph spectrum .
The plots above show the adjacency, incidence, and graph distance matrices for the utility graph.
Excising an edge of the utility graph gives the tetrahedral graph.
The utility graph has graph genus and so it can be drawn on a torus with no crossings, as illustrated above (M. Malak, pers. comm., Feb. 15, 2006).