Let $w(M)$ be the total Stiefel-Whitney classes of a manifold $M$. If we know $w(M)$ and $w(N)$, can we obtain $w(M\# N)$? The following simple relation: $$ w(M \# N) = w(M) + w(N) . $$ seems not right, at lest for the leading term 1. But for $w_1$, it seems right.

==== Added ===

According to the answer below, we have a simple result: $$ w(M \# N) = w(M) + w(N) -1. $$


This is true except in bottom degree.

1) Let $M^\circ = M \setminus \{pt\}$. Then the inclusion induces an isomorphism $H^*(M^\circ) \cong H^*(M)$ in degrees $* < n$. Because $TM^\circ = TM\big|_{M^\circ}$, we get that $w_k(M^\circ) = w_k(M)$ for $k<n$. By Mayer-Vietoris one gets an isomorphism $H^k(M \# N) \cong H^k(M^\circ) \oplus H^k(N^\circ)$ in degrees less than $n$. Because again $TM^\circ = T(M \# N)\big|_{M^\circ}$ and similarly with $N$, we see that $w_k(M^\circ)$ is the projection of $w_k(M \# N)$ in the above isomorphism. This is precisely what it means to say that $w_k(M \# N) = w_k(M) + w_k(N)$.

2) In top degree, $w_n(M) = \chi(M) \mod 2$, so $$\begin{align}w_n(M \# N) &= \chi(M \# N) \mod 2\\&= \chi(M^\circ) + \chi(N^\circ) - \chi(S^{n-1}) \mod 2\\ &= \chi(M) - 1 + \chi(N) - 1 - (1+ (-1)^{n-2})\mod 2\\&= \chi(M)+\chi(N) \mod 2\\&=w_n(M)+w_n(N).\end{align}$$ This could be obtained cohomologically as in (1) by being more careful about the Mayer-Vietoris sequence in top degree.

  • $\begingroup$ Are you assuming that $M$ and $N$ are closed? If not, how does one make sense of the argument for $w_n$? $\endgroup$ – Michael Albanese Dec 31 '16 at 8:59
  • $\begingroup$ Yes, I am. Of course, if they're not closed, that class is automatically zero (since the cohomology group it lives in is zero). $\endgroup$ – user98602 Dec 31 '16 at 9:00
  • $\begingroup$ Yeah, I just realized that, thanks. Great answer by the way. $\endgroup$ – Michael Albanese Dec 31 '16 at 9:08
  • $\begingroup$ Weird. It doesn't hold if one of them is closed and the other isn't, e.g. $\mathbb{RP}^2\#\mathbb{R}^2$. $\endgroup$ – Michael Albanese Dec 31 '16 at 9:18
  • 1
    $\begingroup$ I know it makes sense, we just don't have $w_2(\mathbb{RP}^2) + w_2(\mathbb{R}^2) = w_2(\mathbb{RP}^2\#\mathbb{R}^2)$ - one side is zero, the other isn't. Of course, that's because equality is not literal here, it just being used to mean 'corresponds to' under a certain map. $\endgroup$ – Michael Albanese Jan 1 '17 at 12:35

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.