Let $M$ be manifold, $n = 4$. Is $w_2$ special in in the regard it's the only thing of $H^2(M, \mathbb{Z}_2)$ where $w_2 \cup \tau = \tau \cup \tau$, $\tau \in H^2(M, \mathbb{Z}_2)$ or not? I wondered this question after reading this excerpt from a book on $4$-manifolds by Donaldson and Kronheimer and this excerpt from a book on $4$-manifolds by Scorpan.
- 104,029
-
1Welcome to math.SE: since you are new, I wanted to let you know a few things about the site. In order to get the best possible answers, it is helpful if you write what your thoughts are on the problem and include your efforts (work in progress) in this and future posts and in what context you have encountered the problem; this will prevent people from telling you things you already know, and help them give their answers at the right level. – JKnecht Feb 10 '16 at 00:09
-
1Yes, it's true, and yes, it's unique. Which one of those two facts are you asking about? – Feb 10 '16 at 00:48
1 Answers
Let $M$ be a topological space. For each $n \geq 0$ the is a linear map $\operatorname{Sq}^n : H^{\bullet}(M; \mathbb{Z}_2) \to H^{\bullet}(M; \mathbb{Z}_2)$ satisfying the following properties:
- $\operatorname{Sq}^0(x) = x$,
- $\operatorname{Sq}^m(x) = 0$ for $|x| < m$ and $\operatorname{Sq}^m(x) = x\cup x$ for $|x| = m$, and
- $\operatorname{Sq}^m(x\cup y) = \sum_{i+j = k}\operatorname{Sq}^i(x)\cup\operatorname{Sq}^j(y)$.
These maps are called the Steenrod squares.
Now suppose $M$ is a closed connected $n$-dimensional manifold. Restricting $\operatorname{Sq}^m$ to $H^{n-m}(M; \mathbb{Z}_2)$, we obtain a linear map $\operatorname{Sq}^m \colon H^{n-m}(M; \mathbb{Z}_2) \to H^n(M; \mathbb{Z}_2)$, and therefore an element of
\begin{align*} &\ \operatorname{Hom}(H^{n-m}(M; \mathbb{Z}_2), H^n(M; \mathbb{Z}_2))\\ \cong &\ \operatorname{Hom}(H^{n-m}(M; \mathbb{Z}_2), \mathbb{Z}_2) && \text{($M$ is a closed connected manifold)}\\ \cong &\ \operatorname{Hom}(\operatorname{Hom}(H_{n-m}(M; \mathbb{Z}_2), \mathbb{Z}_2), \mathbb{Z}_2) && \text{(Universal Coefficient Theorem)}\\ \cong &\ H_{n-m}(M; \mathbb{Z}_2) && \text{($H_{n-m}(M; \mathbb{Z}_2)$ is a finite-dimensional}\\ &&& \text{vector space over $\mathbb{Z}_2$)}\\ \cong &\ H^m(M; \mathbb{Z}_2) &&\text{(Poincaré Duality)}. \end{align*}
Let $\nu_m$ denote the unique element of $H^m(M; \mathbb{Z}_2)$ corresponding to $\operatorname{Sq}^m$ under these natural isomorphisms. Unwinding the isomorphisms, we see that $\operatorname{Sq}^m(x) = \nu_m\cup x$ for all $x \in H^{n-m}(M; \mathbb{Z}_2)$. We call $\nu_m$ the $m^{\text{th}}$ Wu class of $M$.
If $M$ is also assumed to be smooth, then the Wu classes of $M$ can be computed in terms of its Stiefel-Whitney classes (see this answer for example). In particular, $\nu_2 = w_2 + w_1\cup w_1$, so in the case $n = 4$, for $x \in H^2(M; \mathbb{Z}_2)$, $x\cup x = \operatorname{Sq}^2(x) = \nu_2\cup x = (w_2 + w_1\cup w_1)\cup x$. That is, $w_2 + w_1\cup w_1$ is the unique class in $H^2(M; \mathbb{Z}_2)$ such that its cup product with $x \in H^2(M; \mathbb{Z}_2)$ is $x\cup x$. In the case of $M$ simply connected (or even just orientable), $w_1 = 0$ so the Wu class reduces to $w_2$.
As Mike Miller points out below, for $M$ a simply connected, closed, smooth four-manifold, $w_2$ is the unique class with the stated property. To see this, let $u \in H^2(M; \mathbb{Z}_2)$ be an element satisfying $u\cup x = x\cup x$ for every $x \in H^2(M; \mathbb{Z}_2)$, then
$$(u - w_2)\cup x = u\cup x - w_2\cup x = x\cup x - x\cup x = 0,$$
so by Poincaré Duality, $u - w_2 = 0$ and hence $u = w_2$.
For more details on Steenrod squares and Wu classes, see this note.
- 104,029
-
2Such a $c$ is unique because if $d$ was another, then $(c-d) \smile x = 0$ for all $x$, which by Poincare duality is absurd. – Feb 10 '16 at 01:06