3
$\begingroup$

For $X$ a paracompact space, I am trying to classify all locally trivial fibration with base the suspension $SX = X \times [-1,1]\, /\, (X \times \{-1\} \cup X \times \{1\})$, and fiber-type a space $F$ such that $G_F = Homeo(F)$ with the C.O. topology is a topological group.

I duplicate the reasonning for classification of fibration with fiber-type $F$ over the sphere $S^n$. In that case, denoting the set of isomorphism classes of such fibrations by $[FB(S^n,F)]$, we get that $$ [FB(S^n,F)] \simeq \pi_{n-1}(G_F) / \pi_{0}(G_F), $$ where $\pi_{n-1}(G_F)$ is the set of pointed homotopy classes $(S^{n−1}, ∗) → (G_F , Id_F )$, and the action of $\pi_{0}(G_F)$ on $\pi_{n-1}(G_F)$ is induced by the action of $G_F$ on itself by conjugaison.

In the case of base $SX$, we have $SX = CX_+ \cup CX_-$ and $X \simeq CX_+ \cap CX_-$, with the cones $CX_{\pm}$ being paracompact and contractile, so every fibration over $CX_+$ or $CX_-$ is trivial. If $\Phi_{\pm}$ are the corresponding trivializations, $f_0 \in G_F$ a fixed homeomorphism of $F$, and $x_0 \in X$ an arbitrary base-point, we can always choose $\Phi_{\pm}$ s.t. $\Phi_+(x_0) = \Phi_-(x_0) = f_0$, and so the transition function from $\Phi_+$ to $\Phi_-$ determine an application $$\phi \colon (X,x_0) \to (G_F,Id_F)$$ As in the case of $S^n$, the action of $G_F$ on itself by conjugaison induces an action on the homotopy classes $[\phi]$ which depends only of the path-component of the element in $G_F$. Lets denote the set of isomorphism classes of fibration with fiber $F$ by $[FB(SX,F)]$, and the set of equivalence classes of homotopy class under the action of $\pi_{0}(G_F)$ by $$[(X,x_0),(G_F,Id_F)]_* = [(X,x_0),(G_F,Id_F)]\,/\,\pi_{0}(G_F)$$ For $[\xi] \in [FB(SX,F)]$ and $[\phi]_* \in [(X,x_0),(G_F,Id_F)]_*$, we can define an application $$\theta \colon [FB(SX,F)] \to [(X,x_0),(G_F,Id_F)]_*, \quad [\xi] \mapsto [\phi]_*$$ This is well-defined : choosing another set of trivializations $\Psi_{\pm}$ gives a function $\psi$ s.t. $[\psi]_* = [\phi]_*$, changing the choice of the value at $x_0$ from $f_0$ to $g_0$ also gives applications in the same equivalence class ; it is also easy to show that two isomorphic fibration $\xi$ and $\xi'$ gives function in the same equivalence class.

Now, $\theta$ is a bijection : it is surjective because from an application $\phi\colon (X,x_0) \to (G_F,Id_F)$ we can reconstruct a fibration over $SX$ with $\phi$ as its transition function unique up to iso, an injective because if $\xi$ gives $\phi$ and $\xi'$ gives $\phi'$ and $[\phi]_* = [\phi']_*$, we can modify $\phi$ s.t. $\phi$ and $\phi'$ are homotopic by an homotopy $H$ and construct a fibration $\xi_H$ over $SX \times [0,1]$, and $SX$ being paracompact we get $$\xi \simeq \xi_{\psi}|_{SX \times {0}} \simeq \xi_{\psi}|_{SX \times {1}} \simeq \xi′$$ So we apparently get that $$ [FB(SX,F)] \simeq [(X,x_0),(G_F,Id_F)]_*$$

But here $[FB(SX,F)]$ cannot depend of the choice of the base-point, but it seems to me that $[(X,x_0),(G_F,Id_F)]_*$ does depend on that choice if $X$ is not a homogeneous space like $S^n$.

So I think there must be a mistake somewhere, but where ??

Or contrary to the intuition, does $[(X,x_0),(G_F,Id_F)]_*$ is independant of the choice of $x_0$ ?

$\endgroup$
1
$\begingroup$

It is independent of the choice of base point.

Let $Map((X,x_0), (G_F,id))$ be the based mapping space (based at $x_0$). Let $Map(X, G_F)$ be the free mapping space. Then we have a split short exact sequence of topological groups:

$$Map((X,x_0), (G_F,id)) \to Map(X, G_F) \stackrel{ev_{x_0}}{\to} G_F$$

(These are groups using pointwise multiplication). The last map, which evaluates at $x_0$, is split by identifying $G_F$ with the constant maps. Note that the inclusion of the constant maps doesn't depend on the choice of basepoint.

In any case this induces a split short exact sequence on the corresponding groups of path components: $$\pi_0Map((X,x_0), (G_F,id)) \to \pi_0Map(X, G_F) \stackrel{ev_{x_0}}{\to} \pi_0G_F$$

Now the group structure on $\pi_0Map((X,x_0), (G_F,id))$ might depend on identifying it as the kernel of the map $ev_{x_0}$, which of course depends on the base point.

However we only care about it as a $\pi_0G_F$-set. From the above split exact sequence it follows that as sets with $\pi_0G_F$-action $[(X,x_0), (G_F,id)] = \pi_0Map((X,x_0), (G_F,id))$ is isomorphic to the $\pi_0G_F$-set of, say, left cosets: $$\pi_0Map((X,x_0), (G_F,id)) \cong \pi_0Map(X, G_F)/ \pi_0G_F$$ The right-hand side is independent of the choice of basepoint $x_0$, which tells us, perhaps surprisingly, that already $[(X,x_0), (G_F,id)]$ is independent of basepoint.

It further follows that the quotient of this set by $\pi_0G_F$, which is your $[(X,x_0), (G_F,id)]_*$, is also independent of basepoint.

$\endgroup$
  • $\begingroup$ Nice ! thx a lot. $\endgroup$ – ychemama Mar 19 at 14:55

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.