MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Is it the case that, for any given knot $K$, there exists some graph $G$ whose every embedding into $\mathbb{R}^3$ (or into $\mathbb{S}^3$) contains a cycle that realizes $K$?

I know the famous Conway-Gordon result that the complete graph on seven vertices $K_7$ is intrinsically knotted in that every embedding contains a knotted cycle. And Joel Foisy proved that $K_{3,3,1,1}$ is also intrinsically knotted (J. Graph Theory, 2002, ACM link).

Instead, I am asking a Ramsey-like question: Whether it's known that, for every particular knot, there exists some graph for which that particular knot is unavoidable.

share|cite|improve this question
Trivial observation: since every graph is a subgraph of a complete graph, it suffices to consider complete graphs for the purposes of this question. – John Pardon Jul 15 '14 at 13:20
up vote 12 down vote accepted

Yes. See this paper of Negami. The main result is that for any fixed knot (or link) of type $k$, there is a constant $R(k)$ such that every straight line embedding of $K_{R(k)}$ in $\mathbb{R}^3$ contains a knot (or link) equivalent to $k$.

share|cite|improve this answer
Very nice! But usually the notion of intrinsic knottedness relies on tame embeddings, rather than specifically straight-line embeddings. – Joseph O'Rourke Jul 15 '14 at 13:16
@JosephO'Rourke a generalisation of Fary's theorem should be able to show these equivalent. – Jan Dvorak Jul 15 '14 at 13:20
Certainly you cannot allow arbitrary embeddings into $\mathbb{R}^3$, because then you could embed each edge as a local knot and the theorem would be false. – Tony Huynh Jul 15 '14 at 13:29
Does it make sense to ask whether, allowing arbitrary embeddings, there must be a cycle realising a knot which is the connected sum of $K$ and some other knot? – Sean Eberhard Jul 15 '14 at 14:05
Negami obtains an upper bound on $R(k)$, but, as he says in his final sentence, this bound "is not however so worthy" because it depends upon as-yet unknown values of Ramsey numbers. – Joseph O'Rourke Jul 15 '14 at 18:02

The answer is no for the following silly reason. Take any graph $G$ and some embedding into $\mathbb R^3$. To any edge, we may perform the connected sum with a fixed knot $K'$ (that is, take a small straight part of the edge and connect sum to it $K'$ there). Do this operation separately to all of the edges. Then any cycle in $G$ has knot type of the form $K'\#(\text{something})$. We just need to pick a knot $K'$ so that the given $K$ is not of this form.

We could modify the question and ask whether a fixed knot $K$ exists in some weaker sense in any given embedding of $G$. One such (very) weak sense could be that there exists a knotted torus $T$ in knot type $K$ and a cycle in $G$ which is contained in $T$ and cannot be made disjoint from a compressing disk for $T$. I suspect this may also have a negative answer, however. It seems one can probably always find embeddings of $G$ which are arbitrarily complicated, and in particular which are not related in any way to the given knot $K$.

share|cite|improve this answer
Tony Huynh said it first. – John Pardon Jul 15 '14 at 13:34

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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