We are familiar with Künneth theorem:

The Kunneth formula is given by $R$ as a ring, $M,M'$ as the R-modules, $X,X'$ are some chain complex. The Kunneth formula shows the cohomology of a chain complex $X \times X'$ in terms of the cohomology of a chain complex $X$ and another chain complex $X'$. For topological cohomology $H^d$, we have $$ H^d(X\times X', M\otimes_R M') \simeq \Big[\oplus_{k=0}^d H^k(X,M)\otimes_R H^{d-k}(X',M')\Big]\oplus $$ $$\Big[\oplus_{k=0}^{d+1} \text{Tor}_1^R(H^k(X,M),H^{d-k+1}(X',M'))\Big] . $$

$$ H^d(X\times X',M) \simeq \Big[\oplus_{k=0}^d H^k(X,M)\otimes_{\mathbb Z} H^{d-k}(X',\mathbb{Z})\Big]\oplus $$ $$ \Big[\oplus_{k=0}^{d+1} \text{Tor}_1^{\mathbb Z}(H^k(X,{M}),H^{d-k+1}(X',\mathbb Z))\Big]. $$

The above is valid for both topological cohomology $H^d$ and group cohomology $\mathcal{H}^d$ (for $G'$ is a finite group): $$ \mathcal{H}^d(G\times G',M) \simeq \Big[\oplus_{k=0}^d \mathcal{H}^k(G,M)\otimes_{\mathbb Z} \mathcal{H}^{d-k}(G',\mathbb Z)\Big]\oplus $$ $$ \Big[\oplus_{k=0}^{d+1} \text{Tor}_1^{\mathbb Z}(\mathcal H^k(G,M),\mathcal H^{d-k+1}(G',\mathbb Z))\Big]. $$

  • Questions:

[1]. Do we have similar results of Künneth theorem for bordism groups $\Omega_d^{...}(...)$? Schematically, maybe something like $$ \Omega_d^{...}(...) \simeq \oplus_n \Omega_n^{...(1)}(...) \otimes \Omega_{d-n}^{...(2)}(...)? $$

[2]. Can we, and, how can we interpret the decompositions of bordism group generators as manifolds $\Sigma_d$: $$ \Sigma_d \sim \Sigma_{d-n}^{(1)} \times \Sigma_{n}^{(2)}? $$ where $d$-manifold generators are related to the $d-n$-manifold generator and $n$-manifold generator. Are these general, or are they only special cases?

Refs are welcome. Thanks.


The Künneth formula for ordinary homology as you present it works only when $R$ is a PID (or more generally of cohomological dimension 1).

For a general well-behaved homology theory[1] (this includes both ordinary cohomology, K-theory and cobordism) there is a Künneth spectral sequence

$$E^2_{p,q}=\mathrm{Tor}_{p,q}^{E_*}(E_*X,E_*Y)\Rightarrow E_{p+q}(X\times Y)$$

(see for example theorem 4.1 in

Elmendorf, A. D.; Kříž, Igor; Mandell, Michael A.; May, J. P., Rings, modules, and algebras in stable homotopy theory. With an appendix by M. Cole, Mathematical Surveys and Monographs. 47. Providence, RI: American Mathematical Society (AMS). xi, 249 p. (1997). ZBL0894.55001.

with $M=E\wedge \Sigma^\infty_+X$ and $N=E\wedge \Sigma^\infty_+Y$).

When $X$ and $Y$ have the homotopy type of finite CW-complexes, you can use Spanier-Whitehead duality to obtain the same result for cohomology $E^*X=E_{-*}(\mathbb{D}X)$, with a bit of care due to the signs that appear from the duality. So under these hypotheses we obtain a spectral sequence

$$E_2^{p,q}=\mathrm{Tor}^{E_*}_{-p,-q}(E^{-*}X,E^{-*}Y)\Rightarrow E^{p+q}(X\times Y)\,.$$

The case of the Künneth formula corresponds to the degenerate situation in which $\mathrm{Tor}^{E_*}_i=0$ for $i\neq 0,1$.

[1] Precisely, I need the homology theory to be represented by an $E_1$-ring spectrum.


If you work with the unoriented bordism groups $\Omega^O_*(X)=MO_*(X)$ then there is a Künneth isomorphism. This is just because there is a natural isomorphism $MO_*(X)=H_*(X;\mathbb{Z}/2)\otimes MO_*$, where $MO_*$ is a graded polynomial ring over $\mathbb{Z}/2$ with one generator $x_i$ in each degree $i>0$ not of the form $2^j-1$. I think that all of this was already known to Thom. The key point is that $H_*(MO;\mathbb{Z}/2)$ is a cofree comodule over the dual Steenrod algebra, by a fairly concrete and straightforward calculation.

There are also some interesting things to say if you are willing to work with the complex bordism groups $MU_*(X)=\Omega^U_*(X)$ instead of the real ones. Here it is known that $MU_*=\mathbb{Z}[a_1,a_2,a_3,\dotsc]$,so in particular this is a free abelian group. Consider the following conditions:

  1. $H_*(X)$ is free abelian
  2. $MU_*(X)$ is a free $MU_*$-module
  3. $MU_*(X)$ is a flat $MU_*$-module
  4. $MU_*(X)$ is Landweber exact
  5. $MU_*(X)$ is torsion-free
  6. The Künneth map $MU_*(X)\otimes_{MU_*}MU_*(Y)\to MU_*(X\wedge Y)$ is an isomorphism for all $Y$.

First, 1 implies 2. Indeed, there is an Atiyah-Hirzebruch spectral sequence $H_*(X)\otimes MU_*\Rightarrow MU_*(X)$, and the differentials are controlled by homotopy groups of spheres in nonzero dimensions so they are torsion-valued, so they must be zero. The spectral sequence therefore collapses and the claim follows easily. Once you have recalled the definition of Landweber exactness it is easy to see that 2 implies 3 implies 4 implies 5. The Landweber exact functor theorem says that 4 implies 6. By taking $Y$ to be a generalised Moore spectrum $S/(v_0^{i_0},\dotsc,v_n^{i_n})$ one can also check that 6 implies 4.

Because $MU_*(X)$ is actually an $MU_*MU$-comodule rather than just an $MU_*$-module, I think it works out that 5 implies 4. However, I am surprised to find that this very natural question has never occurred to me before.

Now suppose we want to consider the oriented real bordism groups $\Omega^{SO}_*(X)=MSO_*(X)$. The coefficient ring $MSO_*$ is somewhat complicated: it is the direct sum of a polynomial algebra over $\mathbb{Z}$ with a module over $\mathbb{Z}/2$, and does not have any very simple presentation. However, there is a map $MU\to MSO$ of ring spectra, and this gives a map $MU_*\to MSO_*$ of coefficient rings, and thus a natural map $$ MSO_* \otimes_{MU_*} MU_*(X) \to MSO_*(X). $$ If $H_*(X)$ is free abelian then we have seen that $MU_*(X)$ is a free $MU_*$-module. It is then not hard to see that the above map is an isomorphism and so $MSO_*(X)$ is free over $MSO_*$. Assuming that, it is not hard to deduce that the Künneth map $$ MSO_*(X)\otimes_{MSO_*}MSO_*(Y)\to MSO_*(X\wedge Y) $$ is again an isomorphism for all $Y$.

  • $\begingroup$ Thanks! +1, I was not aware of this before $\endgroup$ – wonderich Oct 24 '18 at 19:51

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.