6
$\begingroup$

Let $O(n)$ be the $n$-th orthogonal group and $O$ be the direct limit of $O(n)$ with respect to $n$. Let $BO(n)$ and $BO$ be the classifying spaces.

Question: Why $BO$ is an $H$-space? My supervisor said "$BO=\Omega^\infty\mathbb{E}$ where $\mathbb{E}$ is a spectrum." What does this mean?

$\endgroup$
2
  • 1
    $\begingroup$ there's a lot to mention here. Do you know what a spectrum is? If not, it might be good to look up "infinite loop spaces," "extraordinary cohomology theories" or "complex K-theory" to get moving in that direction. $\endgroup$ Nov 18, 2015 at 3:20
  • $\begingroup$ Ah, @Qiaochu has given a much more helpful answer. $\endgroup$ Nov 18, 2015 at 3:20

3 Answers 3

12
$\begingroup$

$BO$ is the connected component of the zeroth space of a spectrum called the real K-theory spectrum. This spectrum represents a cohomology theory, namely real K-theory, and this means that $BO$ has much more structure than an H-space: it is in fact an infinite loop space, which is loosely a homotopy-theoretic version of an abelian group (as opposed to merely a monoid).

If you believe that homotopy classes of maps $X \to BO$ classify stable real vector bundles (ignoring their dimension) on $X$, then the H-space structure on $BO$ comes from direct sum of vector bundles.

$\endgroup$
3
  • 8
    $\begingroup$ Details: a spectrum does not have an ``underlying space'', but it does have a zeroth space, which is an infinite loop space, and the zeroth space of the spectrum representing real K-theory is BO \times Z, not BO; otherwise the answer is on the mark. $\endgroup$
    – Peter May
    Nov 18, 2015 at 4:30
  • 2
    $\begingroup$ May I add that this is related to (real) Bott periodicity, which describes the `additive' delooping of $\mathbb{Z}\times BO$, and hence its base point component $BO$. In fact $\mathbb{Z}\times BO$ is a ring space, as studies by May (together with Quinn, Ray and Tornehave) and the other (infinite) lopp space structure comes from tensor product of vector bundles. Please correct me if I am wrong, but I think, Bott's work has appeared before the machinery for stable homotopy and spectra, as we know today, was established, and later on it was interpreted in the language of stable homtopy theory. $\endgroup$
    – user51223
    Nov 18, 2015 at 5:49
  • $\begingroup$ @Peter: yes, I guess "zeroth space" is a better term. I'll edit. $\endgroup$ Nov 18, 2015 at 6:10
10
$\begingroup$

$BO$ can be defined as the colimit over $(k,n)$ of Grassmanians $G_k(\Bbb R^n)$ of $k$-dimensional linear subspaces of $\Bbb R^n$ (the limit over $n$ is defined by standard inclusions $\Bbb R^n \subset \Bbb R^{n+1}$, and the limit over $k$ is defined by the operation $X \mapsto X\oplus \Bbb R$ which takes a $k$-plane in $\Bbb R^n$ to the $(k+1)$-plane $X\oplus \Bbb R$ inside $\Bbb R^n \oplus \Bbb R = \Bbb R^{n+1}$.

The direct sum operation defines pairings $$ G_k(\Bbb R^n) \times G_{\ell}(\Bbb R^m) \to G_{k+\ell}(\Bbb R^{n+m}) $$ that are compatible with respect to taking colimits. Taking colimits induces a map $BO \times BO \to BO$ defining the $H$-space structure.

$\endgroup$
6
$\begingroup$

If you do not know about spectra and just want to know that $BO$ is a monoid (and hence a H-space), instead of the stronger statement that it is an infinite loop space; you can use that $\Omega_0 U/O \simeq BO$. This is part of the Bott periodicity theorem, $U/O$ is known as the Lagrangian Grassmannian.

$\endgroup$

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.