Let $S^3-K$ be the complement of the figure eight knot complement. Thurston, in his Lecture Notes, constructed a hyperbolic structure, which comes from a discrete, faithful representation $\pi_1(S^3-K)\to SL(2,{\mathcal O}_3)\subset SL(2,C)$. This representation is not conjugate into $SU(2)$, because traces are not real. But, of course, one can get nontrivial $SU(2)$-representations factoring over the abelianization $Z$ of $\pi_1(S^3-K)$.

Thurston goes on to prove that almost all Dehn surgeries at the figure eight knot complement admit a hyperbolic structure, hence a faithful, discrete representation $\pi_1(S^3-K)\to PSL(2,C)$, which by Culler lifts to $SL(2,C)$ and again is not conjugate into $SU(2)$.

On the other hand, Kronheimer and Mrowka show that for Dehn surgery coefficients $\vert\frac{p}{q}\vert\le 2$, the Dehn surgered manifold admits a noncyclic representation to $SU(2)$. This is, in a sense, complementary to Thurston‘s result, who proved his result for sufficiently large Dehn surgery coefficients.

Question: is it known whether Dehn surgeries at the figure eight knot complement, with sufficiently large Dehn surgery coefficients, admit nontrivial (not necessarily faithful) representations to $SU(2)$?


All Dehn surgeries on the figure eight knot $K$ admit irreducible $SU(2)$ representations. This can be proved using Corollary 4.8 of my paper with John Baldwin, "Stein fillings and $SU(2)$ representations", arXiv:1611.05629. The case of 0-surgery follows from Kronheimer and Mrowka's work, as you mention, so I'll omit it.

Suppose for some $\frac{p}{q}$ that there are no irreducible representations $\pi_1(S^3_{p/q}(K)) \to SU(2)$; we can take $\frac{p}{q}>0$ without loss of generality since $K$ is amphichiral. Then the corollary says that either some root of $\Delta_K(t^2) = -t^2+3-t^{-2}$ is a $p$th root of unity, which by inspection doesn't happen, or $S^3_{p/q}(K)$ is an "instanton L-space", meaning that the framed instanton homology $I^\#(S^3_{p/q}(K))$ has rank $p$.

If $S^3_{p/q}(K)$ is an instanton L-space, then Theorem 4.20 of this paper asserts that $I^\#(S^3_n(K))$ likewise has rank $n$ for all sufficiently large integers $n$. At this point, an argument involving the surgery exact triangle and the adjunction inequality (see Proposition 9.4 of arXiv:1710.01957) says that since $K$ has slice genus 1, we can in fact take $n=1$. But $S^3_1(K)$ is the Brieskorn sphere $\Sigma(2,3,7)$, whose framed instanton homology has rank 3, so none of the surgeries on $K$ are L-spaces after all.

  • $\begingroup$ You are saying that manifolds, which are not instanton L-spaces, must admit a nontrivial representation of their fundamental group to SU(2)? $\endgroup$ – ThiKu Mar 7 at 10:34
  • $\begingroup$ I'm claiming this fact for any rational homology sphere $Y$ with the additional hypothesis that $\pi_1(Y)$ is "cyclically finite", a condition which ensures that the Chern-Simons functional used to define $I^\#(Y)$ is Morse-Bott at the reducible representations. (See Theorem 4.6 of that paper for the precise statement.). When $Y$ is $p/q$-surgery on a knot $K$, Boyer--Nicas proved this hypothesis equivalent to the condition that no root of $\Delta_K(t^2)$ is a $p$th root of unity. $\endgroup$ – Steven Sivek Mar 7 at 11:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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