6
$\begingroup$

Assume $M$ is a topological space and $f\in \operatorname{Homeo}(M)$, then $f$ obviously plays a significant role in the torus bundle

$$M_f = M\times I/\{(x,0)\sim (f(x),1)\mid x\in M\}.$$

Hence there should be some general result of the following type: $M_f$ and $M_g$ are bundle isomorphic (resp. diffeomorphic) if and only if "W", where W is a relation between $f$ and $g$.

Can someone help give W and explain? Thank you!

$\endgroup$
4
  • $\begingroup$ I'm not sure "torus bundle" is right, I've heard "mapping torus" used for this construction. Do you want your maps (between mapping tori or whatever) to necessarily preserve the I-coordinate, or at least take each M x {t_0} to some M x {t_1}? $\endgroup$ Apr 22, 2010 at 8:07
  • $\begingroup$ @ Aaron Mazel-Gee.sorry,i made a mistake.as you point out,what i mean is mapping torus. $\endgroup$
    – sara
    Apr 22, 2010 at 9:00
  • 2
    $\begingroup$ @Sara - It would be nice if you edited your question to fix the title. $\endgroup$
    – Sam Nead
    Apr 22, 2010 at 10:25
  • $\begingroup$ At least in 3-manifold topology, the term 'torus bundle' is usually used in the special case of this construction when M is a 2-torus. $\endgroup$
    – HJRW
    Apr 22, 2010 at 18:25

2 Answers 2

8
$\begingroup$

Your $M_f$ is usually called the "mapping torus" of $f$, as Aaron points out. It comes with a map to the circle $S^1 = I/(0 \sim 1)$ and this map is a fibre bundle with fibres isomorphic to $M$.

First of all, the bundle isomorphism type of the bundle $M_f \to S^1$ only depends on the the isotopy class of $f$; i.e. if $f_0$ and $f_1$ are homotopic through homeomorphisms $f_t$ then $M_{f_0}$ and $M_{f_1}$ will be isomorphic bundles. This reduces you to working with the mapping class group, which you probably already knew since I see the mapping-class-groups tag included.

Now, it's an easy exercise to check that $M_f$ and $M_g$ are isomorphic bundles if and only if $f$ and $g$ are conjugate in the mapping class group of $M$.

$\endgroup$
3
  • $\begingroup$ I wonder if there is a subtle point lurking here... if $f_0$ and $f_1$ are isotopic (homotopic through homeomorphisms) then the two bundles will be homeomorphic. However, the question asked about diffeomorphism. This is not my area, so I don't see that a diffeotopy (homotopic through diffeomorphisms?) gives a diffeomorphism of bundles... It feels like the "quality" of the derivative of the time direction will be important? $\endgroup$
    – Sam Nead
    Apr 22, 2010 at 10:32
  • $\begingroup$ For a diffeotopy it is exactly the same short argument; topologically isotopic maps give homeomorphic bundles, and smoothly isotopic maps give diffeomorphic bundles. And in fact, reducing to the mapping class group is unnecessary. If f and g are isotopic then they are conjugate. $\endgroup$ Apr 22, 2010 at 11:31
  • 2
    $\begingroup$ For the second question (in brackets): if the bundles are just diffeomorphic and not bundle-isomorphic, then the mapping classes need not be conjugate. See mathoverflow.net/questions/241822/… $\endgroup$
    – ThiKu
    Oct 10, 2016 at 5:00
5
$\begingroup$

Mapping tori come equipped with a projection to the circle: $p\colon M_f\to S^1$. Bundle isomorphism is isomorphism of the total space that commutes with the projection. A weaker notion of isomorphism for mapping tori is an isomorphism that commutes with the projection only up to homotopy. Pseudoisotopy is the name for the relation on isomorphisms of $M$ that replaces isotopy in this coarser notion of equivalence of mapping tori.

If $M$ is a simply connected manifold of high dimension, then pseudoisotopy implies isotopy and preserving the homotopy class of the map to the circle isn't a big deal, so isomorphism of mapping tori is pretty much isotopy of mapping classes. But if $M$ is not simply connected, there are two ways that mapping tori may be isomorphic without being bundle isomorphic: there may be more pseudoisotopies than isotopies and there may be isomorphisms that do not preserve the homotopy class of the map to the circle.

$\endgroup$
3
  • $\begingroup$ I'm not sure about the high-dimensional situation, but in two dimensions it's worth noting that pseudoisotopy and isotopy are the same (by a non-obvious theorem). $\endgroup$ Oct 10, 2016 at 5:31
  • $\begingroup$ @DylanThurston Right, in 2d pseudoisotopy implies homotopy implies isotopy. In both low and high dimensions (but not smooth 4), the answer is that the difference between isotopy and pseudoisotopy is a higher Whitehead group, which should vanish for aspherical spaces (Borel conjecture). Waldhausen proved the pseudoisotopy=isotopy for Haken 3-manifolds, suggesting that the K-theory gadget vanishes. Hence his seemingly sharp turn into K-theory. (I imagine that for some 3d lens space pseudoisotopy is coarser than isotopy.) $\endgroup$ Oct 13, 2016 at 2:43
  • 1
    $\begingroup$ @DylanThurston Actually, I was wrong. In high dimensions even if $Wh_2$ vanishes, as for free abelian groups, there is an additional obstruction for pretty much any space that is not simply connected. For the nicest spaces around, $S^n\times S^1$ and $(S^1)^n$, the group of homeomorphisms pseudo-isotopic to the identity but not isotopic is the $\mathbb Z/2$ vector space on the set of nontrivial conjugacy classes of elements in the fundamental group (or maybe the subspace invariant under inversion?). Same for diffeomorphisms. $\endgroup$ Feb 27, 2017 at 18:43

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.