Let $M^3$ denote the Poincare homology sphere. I am wondering what the possible orders of (smooth) automorphisms of $M$ are (I'm not sure if allowing arbitrary homeomorphisms changes things?). By presenting the space as $-1$-surgery on the trefoil, there is a automorphism of order 3 given by the 3-fold symmetry of the trefoil. Similarly, the framed link description of $M$ obtained by presenting $M$ as the 5-fold cyclic branched cover of $S^3$ branched over the trefoil has a 5-fold symmetry. Both of these symmetries can also be seen from the dodecahedron/icosahedron definition of $M$.

I believe that the $5k$-fold cyclic branched covers of $S^3$ branched over the trefoil are all $M$ for any $k \geq 1$ and therefore, by drawing the resulting framed link representations, we obtain automorphisms of order $5k$ for any $k$. I am not sure if any of these automorphisms are isotopic to the identity.

What are the possible orders of automorphisms of $M$? Which possible orders have an automorphism of that order that is not isotopic to the identity?

  • 2
    $\begingroup$ The answer to your question for all spherical 3-manifolds is given by the Rubinstein-McCullough paper. In that paper they compute the isometry groups of spherical 3-manifolds. You can use geometrization to argue any finite-order automorphism is conjugate to such an isometry. $\endgroup$ Commented Sep 21, 2018 at 4:00

1 Answer 1


Let $G$ act smoothly and orientably. The quotient $M/G$ is a spherical 3-orbifold. Therefore, by the elliptization theorem, it may be given a metric of constant curvature 1; pulling back, then $M$ is given a constant curvature metric for which $G$ acts by isometries. (This was known to Thurston when the action of $G$ carried fixed points, and due to Perelman when $G$ acts freely.)

So we may just ask what the finite groups of isometries of $S^3/I$ is, where $I$ is the binary dihedral group. Consider the subgroup $\text{Isom}_I(S^3)$ of isometries of $S^3$ which normalize the (right) action of $I$, in the sense that $f(xi) = f(x)i'$ for some $i' \in I$. It is easy to see that every such function is a lift of an isometry of $S^3/I$, and there is a short exact sequence $1 \to I \to \text{Isom}_I(S^3) \to \text{Isom}(S^3/I) \to 1$.

Now recall that $\text{Isom}(S^3) = SO(4) = S^3 \times_{\pm 1} S^3$, the two factors acting by $(q_1, q_2) \cdot v = q_1 v q_2^{-1}$. The condition above on the action of $(q_1, q_2)$ becomes "For fixed $i$, there is an $i'$ so that $q_1 v i q_2^{-1} = q_1 v q_2^{-1} i'$ for all $v$." This is true if and only if $q_2 i q_2^{-1} = i'$ for every $i$. This implies $q_2 \in N_{SU(2)}(I) = I$.

Now we see that $\text{Isom}(S^3/I) = (S^3 \times_{\pm 1} I)/(1 \times I)$; the quotient is $S^3/\pm 1 = SO(3)$.

So every finite group of automorphisms is conjugate to the action of some subgroup of $SO(3)$, acting on $S^3/I = SO(3)/(I/\pm 1)$ by left multiplication. One concludes by listing all finite subgroups of $SO(3)$; this includes, of course, all cyclic groups.

The Smale conjecture (certainly known now for $\Sigma(2,3,5)$, but I don't know the history well enough to give the correct attribution) dictates that $\text{Isom}^+(M) \to \text{Diff}^+(M)$ is a homotopy equivalence when $M$ is a spherical 3-manifold. So $M$ has trivial mapping class group.

  • $\begingroup$ I would be surprised if you couldn't extend this to the case of compact positive-dimensional group actions (it should be easier...), but I have put no effort into this. $\endgroup$
    – mme
    Commented Sep 20, 2018 at 20:17

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.