4
$\begingroup$

Let $O$ be the infinite orthogonal group. By taking a colimit of the diagram of topological groups $O(1) \to O(2) \to O(3) \to \ldots$, we know $O$ has a continuous group operation given by matrix multiplication.

On the other hand, $O$ is an infinite loop space. The $E_\infty$ structure is often written as Whitney sum of matrices. Probably the cleanest model is to take the linear isometries operad as your model for the $E_\infty$ operad.

Here is my question: Is there a sense in which the two operations above are equivalent? In other words, is $O$ under Whitney sum equivalent as a group-like $E_\infty$ space to $O$ under matrix multiplication? (It is not even obvious that the latter is $E_\infty$ as opposed to $E_1$.)

I tried looking through May's works but could not decipher whether an answer is in his papers.

$\endgroup$
2
  • 3
    $\begingroup$ The block sum of matrices extends to make $O$ an $E_\infty$-algebra in topological groups, and forgetting this operad action leaves you with the group $O$ under matrix multiplication. Reformulated after applying the bar construction: The Whitney sum of vector bundles extends to make $BO$ an $E_\infty$-algebra in spaces, and forgetting this operad action leaves you with the space $BO$. The Eckmann-Hilton argument corresponds to the block matrix with entries A, 0; 0; B being conjugate to the matrix with entries AB, 0; 0; I. $\endgroup$ Commented Dec 5, 2023 at 12:04
  • $\begingroup$ So, by Dunn additivity (which says an $E_n$-algebra in $E_1$-algebras is an $E_{n+1}$-algebra) and the Eckmann-Hilton argument (which says the $n+1$ multiplications of an $E_{n+1}$ algebra given by Dunn additivity are all equivalent) we can conclude that O (with matrix multiplication) is equivalent to O (with direct sum of block matrices) as an $E_1$-algebra; and in fact, as an $E_\infty$-algebra. Is that correct? $\endgroup$ Commented Dec 5, 2023 at 14:03

1 Answer 1

6
$\begingroup$

First note that if $f\colon\mathbb{R}^\infty\to\mathbb{R}^\infty$ is a linear isometric embedding and $u\in O(n)\leq O$ we can define $f_*(u)\in O$ by $f_*(u)(f(x))=f(u(x))$ when $x\in\mathbb{R}^n$ and $f_*(u)(y)=y$ when $y\in f(\mathbb{R}^n)^\perp$. This gives a map $f_*\colon O\to O$.

Let $\mathcal{L}$ be the linear isometries operad, so $\mathcal{L}(n)$ can be described as the space of $n$-tuples $(f_1,\dotsc,f_n)$, where each $f_i$ is a linear isometric embedding $\mathbb{R}^\infty\to\mathbb{R}^\infty$ such that $f_i(\mathbb{R}^\infty)$ is orthogonal to $f_j(\mathbb{R}^\infty)$ whenever $i\neq j$. Let $\mathcal{L}'$ be the similar operad defined without the orthogonality condition. Then both $\mathcal{L}(n)$ and $\mathcal{L}'(n)$ are contractible, and $\Sigma_n$ acts freely on $\mathcal{L}(n)$, but not on $\mathcal{L}'(n)$. There is a natural map $$ \alpha\colon \mathcal{L}'(n) \times O^n \to O $$ given by $$ \alpha(f_1,\dotsc,f_n,u_1,\dotsc,u_n)= (f_1)_*(u_1) \cdot \dotsb (f_n)_*(u_n). $$ Because multiplication in $O$ is not commutative, this does not factor through $\mathcal{L}'(n)\times_{\Sigma_n}O^n$. It gives an action of $\mathcal{L}'$ on $O$ in the sense of non-$\Sigma$-operads but not in the world of genuine operads. The operation $\alpha(\text{id},\text{id})\colon O^2\to O$ is just the multiplication map for the group structure of $O$.

If we restrict to the supoperad $\mathcal{L}<\mathcal{L}'$ we find that the maps $(f_i)_*(u_i)$ have orthogonal support in an evident sense, so they commute, so the map $\alpha\colon\mathcal{L}(n)\times O^n\to O$ factors through $\mathcal{L}(n)\times_{\Sigma_n}O^n$, so we have a genuine operad action. This is of course the standard operad action used to regard $O$ as an infinite loop space.

$\endgroup$
2
  • $\begingroup$ Okay, let me try to parse this another way. Let Ass be the non-symmetric associative $\infty$-operad, which I can identify with $\mathcal{L}'$. Likewise I can identify Comm with $\mathcal{L}$ as a symmetric $\infty$-operad, and also as a non-symmetric $\infty$-operad (by forgetting the symmetric group actions). Through your $\alpha$, we witness maps of non-symmetric $\infty$-operads Comm --> Ass --> End($O$) where the composite arises from a map of symmetric $\infty$-operads. $\endgroup$ Commented Dec 5, 2023 at 14:20
  • $\begingroup$ In particular, this says that the $E_1$-algebra structure on $O$ described as matrix multiplication lifts to an $E_\infty$-algebra structure described as block sum. Is this a correct interpretation of your answer? $\endgroup$ Commented Dec 5, 2023 at 14:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .