![](/files/happy5.png)
For a graph , we let
denote the crossing number of
, and we let
denote the size of the largest complete subgraph of
.
![$ G $](/files/tex/b8e7ad0330f925492bf468b5c379baec88cf1b3d.png)
![$ cr(G) \le 5 $](/files/tex/a403effa182f63b08ce8a87b127d4400b86db161.png)
![$ \omega(G) \le 5 $](/files/tex/2320b0f36ab8acdec343b9d613e16e4bab300f1f.png)
In a nifty paper [OZ], Oporowski and Zhao get a little more mileage out of the classic Kempe-chain proof of the 5-color theorem, proving the following result.
![$ G $](/files/tex/b8e7ad0330f925492bf468b5c379baec88cf1b3d.png)
![$ cr(G) \le 3 $](/files/tex/935874c3c12a4f74fff847bda1778a9e5752ccc7.png)
![$ \omega(G) \le 5 $](/files/tex/2320b0f36ab8acdec343b9d613e16e4bab300f1f.png)
In particular, this shows that every graph with crossing number at most 2 is 5-colorable. The graph can be drawn in the plane with three crossings, so it is clear that the assumption
is necessary in the highlighted theorem. However, it is possible that the assumption
might be relaxed. The most extreme example known is the graph
obtained from
by removing the edges of a 5-cycle. This graph has
,
and
, showing that the
bound in the question is best possible.
Erman, Havet, Lidicky and Pangrac [EHLP] further improve this to
![$ G $](/files/tex/b8e7ad0330f925492bf468b5c379baec88cf1b3d.png)
![$ cr(G) \le 4 $](/files/tex/55acc05cc4b0f1175bd09913b9923a0e36d0a5fa.png)
![$ \omega(G) \le 5 $](/files/tex/2320b0f36ab8acdec343b9d613e16e4bab300f1f.png)
On the other hand, the graph with ,
and
(disproving the original conjecture) can be constructed in the following way: take two copies of
with non-edges
and
, respectively. Let
and
be two triangles in these copies, and identify the corresponding vertices in these triangles. Finally, add the edge
.
Bibliography
*[OZ] B. Oporowski and D. Zhao, Coloring graphs with crossings.
[EHLP] R. Erman, F. Havet, B. Lidicky and O. Pangrac, 5-colouring graphs with 4 crossings, in preparation.
* indicates original appearance(s) of problem.