Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. 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

Let $M$ and $N$ be two connected manifolds, such that $M$ is a covering space of $N$. Is it possible for $\pi_1(M)$ to be isomorphic to $\pi_1(N)$ in some non-canoncial way?

EDIT: Of course, I'm excluding the trivial case where $M=N$ and the covering map is the identity.

share|cite|improve this question
When you say $M = N$ do you mean the case that $M \to N$ is the trivial cover or do you mean $M \cong N$? (I don't consider the latter case trivial in the sense that the existence of a connected manifold that can cover itself nontrivially isn't a completely trivial fact.) – Qiaochu Yuan Feb 7 '12 at 21:08
A more interesting question is finding an example where $M\not\cong N$. – user641 Feb 7 '12 at 23:12
@QiaochuYuan - Yes, sorry. I meant the trivial cover. – Braindead Feb 8 '12 at 3:37
So, can we have a space with fundamental group $\mathbb Q$ and a covering space with fundamental group $\mathbb Q \oplus \mathbb Q$, which is isomorphic to $\mathbb Q$ but in many different ways...??? – GEdgar May 9 '12 at 13:41
up vote 13 down vote accepted

Take a double cover $z \mapsto z^2$ of the circle $S^1 \to S^1$.

The underlying group-theoretic question is whether there exist groups with subgroups isomorphic to themselves, and of course there are; above I used the subgroup $2\mathbb{Z}$ of $\mathbb{Z}$, but there are also less trivial examples (for example many subgroups of $F_2$ are isomorphic to $F_2$; in fact $[F_2, F_2] \cong F_{\infty}$). So it remains to find manifolds with appropriate fundamental groups, and as it turns out every finitely presented group is the fundamental group of a $4$-manifold (see Wikipedia).

share|cite|improve this answer

Edit As pointed out by Steve D, step (1) of the following is wrong, but easily fixable. First, to fix it, merely change all occurrences of $S^3$ to $S^2$. The key point is that the antipodal map of $S^2$ is orientation reversing, while that of $S^3$ is orientation preserving.

The problem with my calculation where I "prove" $N$ is nonorientable is that I didn't compute in a chart, but rather in the ambient Euclidean space. It's easy to see it's wrong if one simply replaces $S^3$ by $S^1$. Then one can visually see that the corresponding $N$ I construct is not the Klein bottle, but it just $S^1\times S^1$.

End Edit

For a slighty less trivial example, consider $M = S^3\times S^1$ and $N = S^3 \hat{\times} S^1$, the unique nontrivial $S^3$ bundle over $S^1$. (If you want, $N$ can be thought of as the unit sphere bundle in Möbius+rank 3 trivial bundle over $S^1$. Alternatively, I'm thinking of $N$ as $S^3\times [0,\pi]/$~ where we identify $((x,y,z,w),0)$ with $((-x,-y,-z,-w),\pi)$.)

I claim that (1) $N$ is nonorientable (and therefore not even homotopy equivalent to $M$), (2) $M$ double covers $N$, and (3) $\pi_1(M)$ is isomorphic to $\pi_1(N)$.

To see (1), consider the curve $\gamma(t) = ((\cos(t),\sin(t),0,0), t)$ on $S^3\times [0,\pi]$. When projected to $N$, $\gamma$ is a closed curve. The claim is the a basis chosen at $\gamma(0)$ changes orientation coming back to $\gamma(\pi)$. To see this, notice the vector $e_1(t) = ((-\sin(t), \cos(t),0,0),0)$ is always in the tangent space of $S^3\times[0,1]$ at $\gamma(t)$ and likewise so are $e_2(t) = ((0,0,1,0),0)$, $e_3(t) = ((0,0,0,1),0)$ and $e_4(t) = ((0,0,0,0),1)$.

The point is the the differential of the gluing map sends $e_i(\pi)$ to $-e_i(0)$ for $i=1,2,3$ and sends $e_4(\pi)$ to itself. This is a negative determinant transformation, so the orientation has reversed.

Now on to (2). The motivation comes from the picture of the torus double covering the Klein bottle.

Thinking of $M$ as $S^3\times [0,2\pi]/$~ where we identify $((x,y,z,w),0)$ with $((x,y,z,w),2\pi)$, define the map $f:M\rightarrow N$ by identifying $((x,y,z,w),t)$ with $((-x,-y,-z,-w),t+\pi)$ for $t\leq \pi$. Check that $f$ is well defined, really maps onto $N$, is continuous, and really does double cover it. (I'll admit I haven't checked all the details myself).

Finally, (3). The long exact sequence for the homotopy groups of a fibration immediately imply (since $\pi_1(S^3) = 0$), that $\pi_1(N)\rightarrow\pi_1(S^1)$ is an isomorphism (and likewise for $\pi_1(M)$). Thus, $\pi_1(M)\cong \pi_1(N)\cong\mathbb{Z}$.

share|cite|improve this answer
This is similar to $S^2\times S^1$ covering $S^2 \hat{\times} S^1$, which is slightly easier to handle since these are the only two closed 3-manifolds with infinite cyclic $\pi_1$. – user641 May 9 '12 at 2:08
@SteveD: Agreed, but I wasn't sure how to show $S^2 \hat{\times} S^1$ and $S^2\times S^1$ were not homotopy equivalent. I think a similar argument to the above shows $S^2\hat{\times} S^1$ is orientable, so that won't work. – Jason DeVito May 9 '12 at 3:42
It is not orientable; it comes from the antipodal map on $S^2$. Explicitly, take $S^2\times I$ and identify the boundary spheres via the antipodal map. – user641 May 9 '12 at 4:51
In fact, I think this case is easier to handle, because $S^2$ is even-dimensional. In particular, if $S^2\hat{\times} S^1$ was orientable, it would be bundle-isomorphic to $S^2\times S^1$. But with $S^2\hat{\times} S^1$, the generator of $\pi_1(S^1)$ acts via inversion on $H_2(S^2)$ (this is by design). – user641 May 9 '12 at 5:11
I got it - the problem is that I didn't use a chart for my computations. Thanks for catching my error - now comes a big edit. – Jason DeVito May 9 '12 at 13:29

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.