Is there an easy way to see the equivalence of the two statements of Bott periodicity. $$BU \times \mathbb{Z} \simeq \Omega^2BU$$ and $$K(X)\otimes K(S^2) \cong K(X\times S^2)$$


The easy way for me to think about this things is via the Yoneda lemma. This works better with reduced $K$-theory. This is defined for a pointed space $X$ as $$ \tilde K^0 (X) := ker(K^0(X)\to K^0(*))$$ It is ``easy'' to see that your second statement reduces to the fact that multiplication by the generator of $\tilde K^0(S^2)$ induces an isomorphism $\tilde K^0(X) \cong \tilde K^0(X\wedge S^2)$ (this can be seen in many ways, one is noting that the projections of a product to two factors have canonical pointed sections and so induce direct summands in $K$-theory).

The key fact about reduced $K$-theory is that it is representable. That is there is a natural isomorphism $$ \tilde K^0(X) \cong [X,BU\times \mathbb{Z}]_*$$ where $[,]_*$ are the homotopy classes of maps. So your second statement says that there is a natural isomorphism $$[X,BU\times \mathbb{Z}]_* \cong [\Sigma^2 X,BU\times\mathbb{Z}]_* $$ By using the suspension-loopspace adjunction you get a natural isomorphism $$ [X, BU\times \mathbb{Z}]_* \cong [X,\Omega^2(BU\times\mathbb{Z})]_*$$ Finally the Yoneda lemma tells us that this must correspond to an homotopy equivelence $BU\times \mathbb{Z}\sim \Omega^2(BU\times\mathbb{Z})$, which is your first statement.

If you look closely all steps are reversible, so the two statements are in fact equivalent.

EDIT: For future reference I am putting the steps to recover the unreduced statement from the reduced statement

I am adding the steps on how to prove the reduced formulas and the unreduced formulas to be equivalent. First note that the basepoint inclusion $*\to X$ has a retraction $X\to *$, so $$ K(X) = \tilde K(X)\oplus \mathbb{Z}$$ Moreover, if $X,Y$ are pointed spaces the inclusion $X=X\times * \to X\times Y$ has a retraction, so the induced map on $K$-theory $$ \tilde K^0(X\times Y) \to \tilde K^0(X) $$ has a section. Since the same is true for $Y$ from the additivity of $K$-theory it follows that the map induced by the inclusion $X\vee Y\to X\times Y$ $$ \tilde K^0(X\times Y)\to \tilde K^0(X\vee Y) = \tilde K^0(X)\oplus \tilde K^0(Y)$$ splits. Now consider the cofiber sequence $$ X\vee Y \to X\times Y \to X\wedge Y$$ This induces an exact sequence $$ \tilde K^0(\Sigma(X\times Y)) \to \tilde K^0(\Sigma X\vee \Sigma Y) \to \tilde K^0(X\wedge Y) \to \tilde K^0(X\times Y) \to \tilde K^0(X\vee Y)$$ We now that the rightmost arrow splits and the rightmost arrow splits too by a similar argument (in fact it splits even at the space level). So we have the desired isomorphism $$ K(X\times Y) = \mathbb{Z}\oplus\tilde K^0(X\times Y) = \mathbb{Z}\oplus\tilde K^0(X) \oplus \tilde K^0(Y) \oplus \tilde K^0(X\wedge Y) $$

| cite | improve this answer | |
  • $\begingroup$ This works directly only if one has $K(X)\otimes K(S^2) \cong K(X\times S^2)$ for all spaces $X$ and not only for compact Hausdorff spaces. In the latter case one gets a priori only a weak homotopy equivalence. Depending on the model for BU (CW-complex or not) one might get really a homotopy equivalence. I guess, the standard Grassmannian model for $BU$ has at least the homotopy type of a CW-complex? $\endgroup$ – Lennart Meier Apr 3 '14 at 22:58
  • $\begingroup$ Yeah $BU$ has the homotopy type of a CW-complex ($BU_n$ has an explicit cell decomposition and I think that it passes to the limit, even if I never really checked). Anyway in my experience trying too hard to distinguish between homotopy equivalences and weak equivalences is not very useful. $\endgroup$ – Denis Nardin Apr 3 '14 at 23:02
  • $\begingroup$ Thanks for answering. The problem I am have with this is that $[-,BU\times\mathbb{Z}]$ represents the unreduced $K$ functor. Also since $K(S^2)=\mathbb{Z}\oplus\mathbb{Z}$, I think $K(X)\cong K(X\times S^2)$ is false. $\endgroup$ – Rene Schipperus Apr 3 '14 at 23:03
  • $\begingroup$ $BU\times \mathbb{Z}$ represents the reduced $K$-theory in pointed spaces. Also I wrorte $\tilde K^0(X)\cong \tilde K^0(X\wedge S^2)$, where I used the smash product not the cartesian product. $\endgroup$ – Denis Nardin Apr 3 '14 at 23:04
  • $\begingroup$ I cannot read very well your last formula, if you want I'll edit in the details on how to get the unreduced result from the reduced result. It essentially comes from $K(X\times S^2) = \tilde K(X\wedge S^2) \oplus \tilde K(X) \oplus \tilde K(S^2) \oplus \mathbb{Z}$ $\endgroup$ – Denis Nardin Apr 4 '14 at 0:36

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.