4
$\begingroup$

I was going through this paper by Tanaka. I am actually stuck at Lemma 5.2, page 365, given below also

here

The argument he gives above works, in particular for $\operatorname{Sq}^{2^r-2^j}$ but I am not sure how to prove the result for any Steenrod operation i.e. a monomial of the form $\operatorname{Sq}^i\operatorname{Sq}^j\dots \operatorname{Sq}^n$ with total degree ${2^r-2^j}$. I am also aware of the fact that Steenrod algebra has generators of the form $\operatorname{Sq}^{2^i}$ as well as the admissible monomials. Will proving the result for these generators suffices? If so, how can we do it?

Any suggestions or hints will be appreciated and will be of great help.

$\endgroup$

2 Answers 2

7
$\begingroup$

Let me add some details to Nicholas Kuhn's answer (mostly for my own edification).

First, since Steenrod squares are stable cohomology operations, it is enough to show that the maps $\operatorname{Sq}^I : H^{2^j-2k}(\mathbb{CP}^n) \to H^{2^r-2k}(\mathbb{CP}^n)$ must all be zero.

Lemma: The subspace $F_p = \operatorname{span}\{y^m \mid \alpha(m) \leq p\}$ is a sub-$A$-module of $H^*(\mathbb{CP}^n)$; here $A$ denotes the Steenrod algebra.

Proof: We just need to check that $F_p$ is preserved by each Steenrod square. For degree reasons, we have $\operatorname{Sq}^{2\ell+1}(y^m) = 0 \in F_p$, while $\operatorname{Sq}^{2\ell}(y^m) = \binom{m}{\ell}y^{m+\ell}$; see this answer. If $\binom{m}{\ell} \equiv 0 \bmod 2$, then $\operatorname{Sq}^{2\ell}(y^m) = 0 \in F_p$. On the other hand, if $\binom{m}{\ell} \equiv 1 \bmod 2$, then $\alpha(m + \ell) \leq \alpha(m) \leq p$ as pointed out by Tanaka, so $\operatorname{Sq}^{2\ell}(y^m) = y^{m+\ell} \in F_p$. $\quad\square$

Now let $p = \alpha(2^{j-1}-k)$.

Note that $y^{2^{j-1}-k} \in H^{2^j-2k}(\mathbb{CP}^n)$ is an element of $F_p$, while $y^{2^{r-1}-k} \in H^{2^r-2k}(\mathbb{CP}^n)$ is not because $\alpha(2^{r-1}-k) > \alpha(2^{j-1}-k) = p$. By the lemma, we see that $\operatorname{Sq}^I(y^{2^{j-1}-k}) \in F_p$, so $\operatorname{Sq}^I(y^{2^{j-1}-k}) \neq y^{2^{r-1}-k}$ and hence $\operatorname{Sq}^I(y^{2^{j-1}-k}) = 0$, i.e. $\operatorname{Sq}^I : H^{2^j-2k}(\mathbb{CP}^n) \to H^{2^r-2k}(\mathbb{CP}^n)$ is the zero map.

$\endgroup$
4
  • $\begingroup$ Yes, Tanaka's sentence "It is easy to see that ..." implies that these subspaces are submodules, and this is what he was implying. $\endgroup$ Jun 24, 2022 at 21:04
  • $\begingroup$ @NicholasKuhn: I can see that now after reading your answer. Prior to that, I was unaware that he was referring to the subspace $F_p$. $\endgroup$ Jun 24, 2022 at 22:13
  • $\begingroup$ So, $Sq^I(y^{2^{j-1}-k})\in F_p$ because $F_p$ is an A-submodule, right? $\endgroup$ Jun 25, 2022 at 8:02
  • 1
    $\begingroup$ Exactly, which is equivalent to $F_p$ being closed under Steenrod operations. $\endgroup$ Jun 25, 2022 at 8:17
8
$\begingroup$

Recall that $H^*(\mathbb CP^{\infty};\mathbb Z/2)= \mathbb Z/2[y]$ with $|y|=2$. Let $F_k \subset \mathbb Z/2[y]$ be the span of $y^m$ such that $\alpha(m)\leq k$. It is well known, and easy to check, that $F_k$ is a sub-$A$-module of $\mathbb Z/2[y]$. Thus Tanaka's inequality $\alpha(2^{r-1}-k) > \alpha(2^{j-1}-k)$ suffices to show his lemma.

(The $F_k$'s also make up the `primitive filtration' of $\mathbb Z/2[y]$.)

$\endgroup$
1
  • 3
    $\begingroup$ I hope you don't mind, I added some details in a separate answer. $\endgroup$ Jun 24, 2022 at 19:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.