Let us consider real vector bundles, and denote by $V_k$ the tautological bundle $V_k\to BO(k)$. From $$ Thom(V\oplus 1_{\mathbb{R}}\to X)=\Sigma Thom(V\to X) $$ and from $j^*V_{k+1}=V_k\oplus1_{\mathbb{R}}$, where $j\colon BO(k) \to BO(k+1)$ is the canonical embedding, one sees that writing $MO_k=\Sigma^{-k}Thom(V_k\to BO(k))$ one gets a sequence of morphisms $MO_k\to MO_{k+1}$ and one can define $$ MO=\lim_{\to} MO_k $$ This way (and taking infinite suspension) one defines the Thom spectrum $MO$, whcih can therefore informally be thought as the infinite desuspension of the Thom spectrum of the tautological infinite rank vector bundle over $BO(\infty)$. Direct sum of vector bundles and the fact that $j_{m,n}^*V_{m+n}=V_m\oplus V_n$, where $j_{m,n}\colon BO(m)\times BO(n)\to BO(m+n)$ is the canonical embedding makes $MO$ a ring spectrum.

Apparently, one should be able to do the very same construction by "doubling" the $V_k$'s, i.e., by considering vector bundles $$ W_k=V_k\oplus V_k $$ noticing that $j^*W_{k+1}=W_k\oplus 1_{\mathbb{R}^2}$ and so $Thom(j^*W_{k+1}\to BO(k)) = \Sigma^2 Thom(W_{k}\to BO(k))$, and then considering the pointed spaces $M_2O_k=\Sigma^{-2k}Thom(W_k\to BO(k))$ and then the colimit $$ M_2O=\lim_{\to} M_2O_k $$ This again should be a ring spectrum by the same reason as for $MO$.

As, for any real vector space $V$, its double $V\oplus V$ carries a canonical comple structure $J\colon V\oplus V \to V\oplus V$, and since the defining representation $\mathbb{C}^n$ of $U(n)$ restricted to $O(n)$ splits as the sum of two copies of the defining representation $\mathbb{R}^n$ of $O(n)$, one should get a morphism of ring spectra $M_2O \to MU$.

By looking at how the $\hat{A}$ polynomial is derived from the Todd polynomial, one would suspect that the analogous morphism $$ M_2Spin \to MU $$ and a compatibility between this, the standard complex orientation of $KU$, the Atiyah-Bott-Shapiro orientation of $KO$ and complexification of vector bundles is what secretely lies behind the Atiyah-Singer formula for the index of a twisted spin complex, making Atiyah-Singer formula a version of the general kind of Hirzebruch-Riemann-Roch formulas one has when dealing with morphisms of generalized cohomology theories and pushforwards (see, e.g., Panin-Smirnov).

However, I've not been able so far to locate something resembling $M_2O$ or $M_2Spin$ in the literature I have searched in, so I could possibly be on a false track here.

Is the construction sketched above possibly correct? What is a reference to it? has it a more canonical name than $M_2O$? is it really related to the construction and relevance of $\hat{A}$ along the lines sketched above?

  • $\begingroup$ For the construction you consider above, the only existing example I can think of are the spectra $\mathbb{R} P^\infty_k$ at the Thom spectrum of $kV_1$ as well as $\mathbb{C} P^\infty_k$ which fit into a nice cofibre sequence of spectra. The other examples that I can think of are Thom spectra of $k\rho_n$ where $\rho_n$ is the reduced representation of $(\mathbb{Z}/2)^{\times n}$. For this, you can look at work of Takayasu. The existing examples somehow show that the spectra $kV_n$ where $k$ is any integer could be/are very complicated. $\endgroup$
    – user51223
    Nov 6, 2019 at 12:27
  • 2
    $\begingroup$ @user51223 OP is asking about "Thom spectra over $BO$", and not Thom spectra over $BO(n)$ for fixed $n$. So something that would be more closely related to his question is $MTO$, which would become $M_{-1}O$ in his notation. $\endgroup$
    – user43326
    Nov 6, 2019 at 16:41
  • 2
    $\begingroup$ Recall that $MU$ is the Thom spectrum of the bundle over $BU$ given by realification $BU\to BO$. The spectrum $M_2 O$ is the Thom spectrum of the map $BO\to BU\to BO$, where the first map is complexification. Upon Thomification, this gives the desired map $M_2 O\to MU$. $\endgroup$
    – skd
    Nov 9, 2019 at 15:44

1 Answer 1


I think that both of your examples, $M_2Spin$ and $M_2O$, arise naturally in the context of Thom spectra induced by $(B,f)$-structures. Given a $(B,f)$-structure $\mathcal{B}= \{f_n: B_n \to BO(n)\}$, the associated Thom spectrum $M\mathcal{B}$ is defined componentwise as: $$ M\mathcal{B}_k = Thom(f_k^*V_k\to B_k), $$ where the maps $\Sigma M\mathcal{B}_k \to M\mathcal{B}_{k+1}$ are given by looking at the pullback square $\require{AMScd}$ $$ \begin{CD} \mathbb{R}\oplus f_k^*V_k @>>> f_{k+1}^*V_{k+1}\\ @VVV @VVV \\ B_k @>>> B_{k+1}. \end{CD} $$ One can 'double' this construction replacing the maps $f_k$ with $\tilde{f_k}$ defined to be the composition: $$ \tilde{B_{2k}}:=B_k\overset{f_k}{\to} BO(k) \overset{\Delta}{\to} BO(k) \times BO(k) \overset{j_{k,k}}{\to} BO(2k) $$ and get a $S^2$-$(B,f)$-structure, a $(B,f)$-structure indexed only on even natural numbers, denoted by $2\mathcal{B}$. By definition, the Thom spectrum $M_2\mathcal{B}$ associated to this new $S^2$-$(B,f)$-structure, is

$$ (M_2\mathcal{B})_{2k} = Thom(\tilde{f_k^*}V_{2k}\to \tilde{B_{2k}}) = Thom(f_k^*V_k\oplus f_k^*V_k \to B_k) $$ $$ (M_2\mathcal{B})_{2k+1} = \Sigma(M_2\mathcal{B})_{2k}. $$ In your case, $M_2Spin$ and $M_2O$, are (as sequential spectra) the Thom spectra associated to the 'doubled' $(B,f)$-structures that classically define $MSpin$ and $MO$, i.e. the $(B,f)$-structures respectively given by the maps $BSpin(k)\to BO(k)$ and ${{\rm id}}_{BO(k)}$.

All the details and references can be found in the nlab pages Thom spectrum and G-structure.


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.