31
$\begingroup$

The existence of a 4-chromatic unit distance graph (e.g., the Moser spindle) establishes a lower bound of 4 for the chromatic number of the plane (see the Nelson-Hadwiger problem).

Obviously, it would be nice to have an example of a 5-chromatic unit distance graph. To the best of my knowledge, the existence of such a graph is open. Has there been any (documented) attempt to find such a graph through a computer search? For instance, has every $n$-vertex possibility been checked up to some $n$?

$\endgroup$
4
  • $\begingroup$ An interesting variation, which seems more amenable to computer search, is to look for non-4-colorable almost-unit distance graph. You can indeed ask for the chromatic number $\chi_\epsilon$ of the graph whose vertices are the points of the plane, and an edge joins two vertices if their distance is between $1$ and $1+\varepsilon$. Obviously $\chi_\varepsilon$ decreases when $\varepsilon\to 0$, but the limit is not known. It is not known if the limit is the chromatic number of the plane either. $\endgroup$ Apr 22, 2016 at 20:07
  • $\begingroup$ @Benoit Do you have some references for almost-unit distance graphs? $\endgroup$
    – domotorp
    Oct 12, 2017 at 8:26
  • $\begingroup$ @domotorp: no, I don't think it has been actually looked at (I have some lose ideas myself, but nothing written). $\endgroup$ Oct 12, 2017 at 14:30
  • 3
    $\begingroup$ @Benoit I've just realized that $\chi_\varepsilon\ge 6$, as shown by the following simple argument. Divide the plane into a grid of length $\varepsilon$ and color each small square to the color of its center. If the original coloring avoided distances $1\pm \varepsilon$, then this new coloring will still avoid unit distances. Since each region is nice, at least 6 colors are needed because of this result: sciencedirect.com/science/article/pii/0097316573900204 $\endgroup$
    – domotorp
    Aug 3, 2018 at 7:57

3 Answers 3

44
$\begingroup$

As of this morning there is a paper on the ArXiv claiming to show that there exists a 5-chromatic unit distance graph with $1567$ vertices. The paper is written by non-mathematician Aubrey De Grey (of anti-aging fame), but it appears to be a serious paper. Time will tell if it holds up to scrutiny.

EDIT: in fact, it must be the one with 1585 vertices, according to checkers, see https://dustingmixon.wordpress.com/2018/04/10/the-chromatic-number-of-the-plane-is-at-least-5/

$\endgroup$
9
  • 2
    $\begingroup$ You have to be kidding me! $\endgroup$
    – Nik Weaver
    Apr 9, 2018 at 18:48
  • 5
    $\begingroup$ It is computer checkable if the graph (including vertex locations) is available. I don't think it should be dismissed out of hand, even though the result is quite surprising. $\endgroup$ Apr 10, 2018 at 0:19
  • 5
    $\begingroup$ The graph is now available in multiple formats (including DIMACS) here: dustingmixon.wordpress.com/2018/04/10/… $\endgroup$ Apr 10, 2018 at 14:39
  • 3
    $\begingroup$ This is an amazing result! @Juho: It wouldn't hurt to run a different SAT solving program on it, as an independent check. $\endgroup$ Apr 10, 2018 at 15:18
  • 4
    $\begingroup$ @TimothyChow treengeling finished with UNSAT result as well, as expected (or hoped). I can upload the files later on too if anyone is interested. $\endgroup$
    – Juho
    Apr 11, 2018 at 10:40
10
$\begingroup$

It depends how serious you require the search to be. ☺

When writing this note, I made a few attempts at experimenting in this direction, but I quickly came to the conclusion that either I didn't know how to approach the experimental problem, or that it was just too large to be feasible, or both.

I tried to concentrate on a particular set of graphs, namely the minimal $5$-chromatic subgraphs of $(\mathbb{F}_p)^2$ (with an edge between $(x,y)$ and $(x',y')$ iff $(x-x')^2+(y-y')^2=1$) for small $p$, because some of the remarks in the aforementioned note (esp. around prop. 5.3) suggest that this might be a good place to look. But even there, I obviously got nowhere (although I can't say that I tried extremely hard).

$\endgroup$
8
$\begingroup$

It is at least known that there is no 5-chromatic unit distance graph on at most 12 vertices [1, Theorem 4]. I don't know if something similar is known for larger values of $n$.


[1] Pritikin, Dan. "All unit-distance graphs of order 6197 are 6-colorable." Journal of Combinatorial Theory, Series B 73.2 (1998): 159-163.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.