3
$\begingroup$

I asked the same question on math.stackexchange recently (https://math.stackexchange.com/questions/2134978/is-it-possible-to-orbifold-torus-td-into-a-sphere-sd-using-mathbbz-2), but it didn't receive much attention there, so I decided to move the question to this forum.

Consider a torus $T^d$ constructed as a hypercube $[-1,1]^d$ with identified opposite faces. Then additionally identify pairs points $\boldsymbol{r}$ and $-\boldsymbol{r}$ within this cube. For $d>1$, what is the result of this orbifold quotient? Specifically, is the underlying space homeomorphic to sphere $S^d$ or not?

For $d=2$ it is easy to visualize this procedure simply by taking half of a square $[-1,1]^2$ and by properly "gluing" the edges. It turns out that indeed $T^2/\mathbb{Z}_2\cong S^2(2,2,2,2)$, which has underlying space $S^2$. I am wondering whether such a statement (or a similar one) also generalizes to $d\geq 3$.

$\endgroup$
11
  • 6
    $\begingroup$ Are you sure it's true for d=1? That seems to me homeomorphic to the interval, not the circle. By the way, this question seems very related mathoverflow.net/questions/213001 $\endgroup$
    – j.c.
    Feb 23, 2017 at 14:13
  • 1
    $\begingroup$ It's not going to be homeomorphic to any manifold. The origin will be locally a cone over a projective space, and this is a disc only in dimension 2. $\endgroup$ Feb 23, 2017 at 14:28
  • 1
    $\begingroup$ Do you have $2^{2d}$ fixed points, right? My impression is that for $d$ even this is homeomorphic to the quotient of a $d$-dimensional complex torus by the Kummer involution $(-1)$, whose quotient is not a sphere for $d >2$. For instance, when $d=4$ the quotient is homeomorphic to the quotient of a $2$-dimensional complex torus by the action of $(-1)$, and the quotient is a Kummer surface, namely a complex $K3$ surface with sixteen conical singularities, which is obviously not homeomorphic to $S^4$ (for instance, because the Euler number is $8$). $\endgroup$ Feb 23, 2017 at 14:29
  • $\begingroup$ @j.c. You are perfectly right, I screwed up the $d=1$ case... so there is perhaps nothing to generalize to higher dimensions. $\endgroup$ Feb 23, 2017 at 14:41
  • $\begingroup$ @FrancescoPolizzi: I think it's only $2^d$ stable points. E.g. for $d=2$ these are $[0,0]$, $[0,1]$, $[1,0]$ and $[1,1]$. $\endgroup$ Feb 23, 2017 at 14:43

1 Answer 1

6
$\begingroup$

Given the specific quotient described in the question, the answer is no. Under this quotient there is a fixed point at the origin and the neighborhood of this point quotients to a cone over $RP^{d-1}$. For $d=2$, $RP^1=S^1$ so the cone has underlying space a disk, however for $d>2$, the underlying cone will not be homeomorphic a $d$-ball and so it is not even a manifold (a necessary condition to be $S^d$).

For $d\geq 3$, even if we allow more interesting fixed points sets (say knots and links), we still cannot obtain $S^d$ as the 2-fold quotient (or in fact cyclic) of $T^d$. This is more commonly said in reverse that $T^d$ is not the cyclic-branched cover over $S^d$, which follows from:

Hirsch, U., & Neumann, W. D. (1975). On cyclic branched coverings of spheres. Mathematische Annalen, 215(3), 289-291.

$\endgroup$

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.