5
$\begingroup$

Ravenel and Wilson showed that $K(\mathbb Z / p^j,q)$ is $K(n)$-acyclic for any $q \geq n+1$, and that $K(\mathbb Z, q)$ is $K(n)$-acyclic for $q \geq n+2$. It follows that $K(A,q)$ is $K(n)$-acyclic for $q \geq n+2$ when $A$ is finitely-generated.

From here, a Serre spectral sequence argument reveals that the map $\tau_{\leq m} X \to \tau_{\leq n+1} X$ (where $\tau$ is Postnikov truncation) is a $K(n)$-local equivalence for all $m \geq n+1$ when $X$ has finitely-generated homotopy groups. (For $\pi$-finite spaces, $\tau_{\leq m} X \to \tau_{\leq n} X$ is in fact a $K(n)$-local equivalence, as observed by Carmeli,Schlank, and Yanovski).

It's tempting to conclude that $X \to \tau_{\leq n+1} X$ is a $K(n)$-local equivalence for any $X$ with finitely-generated homotopy groups, but this can't possibly be true. If it were true, then in particular $X \to \tau_{\leq n+1} X$ would be an equivalence for all simply-connected finite spaces $X$. Then we could conclude $K(n)_\ast(S^2) = K(n)_{\ast+n+1}(\Sigma^{n+1} S^2) = K(n)_{\ast+n+1}(pt)$, which is false.

Indeed, according to Bauer, the convergence of the spectral sequence for the Postnikov tower of $X$ only holds when $X$ is $n$-truncated. This leads to my

Question: If $X$ is a space with infinitely many nontrivial homotopy groups, is there any meaningful relationship between $K(n)_\ast(X)$ and $K(n)_\ast(\tau_{\leq n} X)$ or $K(n)_\ast(\tau_{\leq n+1} X)$? (Beyond the mere existence of a map -- for all I know, this map is zero!) How about if $X$ is finite? Or perhaps, what if $X$ is $(n-1)$-connected?

$\endgroup$
  • 1
    $\begingroup$ You do have the Serre Spectral sequence, but I'm not sure how helpful it is $\endgroup$ – Denis Nardin Apr 17 at 18:22
  • 2
    $\begingroup$ I wouldn't think that there is any meaningful relationship. For $E$ stable, there is an equivalence $L_{K(n)} \tau_{\ge k} E \simeq L_{K(n)} E$, so that all of the information that $L_{K(n)}$ (or $K(n)_*$) sees is attached to the "germ at infinity" of $E$ (a memorable slogan credited to Dwyer). Since the parts of $\Sigma^\infty_+ X$ and $\Sigma^\infty_+ \tau_{\le k} X$ near infinity can be arbitrarily far apart without very strong hypotheses on $X$, their $K(n)$-homologies are not likely to bear any particular relation. $\endgroup$ – Eric Peterson Apr 17 at 20:31
  • 1
    $\begingroup$ You can make pretty strong statements about what $K(n)_* \tau_{\le k} X$ looks like, though, which you might find separately interesting: math.jhu.edu/~wsw/papers2/math/… . $\endgroup$ – Eric Peterson Apr 17 at 20:31
  • 2
    $\begingroup$ I don't know a good answer, but I suspect that this could be a fruitful area to investigate. One possible approach: the above comments tell us about $K(n)_*(\Omega^\infty X)$ for $X$ in the thick subcategory generated by $H=BP\langle 0\rangle$, and some of the proofs can be given in terms of the Bousfield-Kuhn functor $\Phi_n$ which satisfies $\Phi(\Omega^\infty X)=L_{K(n)}X$. So you can proceed to investigate $K(n)_*(\Omega^\infty X)$ when $X$ is in the thick subcategory generated by $BP\langle m\rangle$ with $m>0$, using the theory of Wilson spaces. $\endgroup$ – Neil Strickland Apr 17 at 21:44
  • 2
    $\begingroup$ @NeilStrickland Neil and others: I wrote a long paper on the Morava K-theory of infinite loopspaces [Adv. Math. 201 (2006), 318-378]. It is quite definitive, and, yes, does use the $\Phi_n$ functors. Folks are encouraged to look at this to get a sense of weird stuff that can happen. $\endgroup$ – Nicholas Kuhn Apr 17 at 22:16
6
$\begingroup$

Let me make more concrete a comment I already wrote. In [Adv. Math. 201 (2006), 318-378], my example 2.22, illustrating a theorem just before it when $n=1$, says that, for any spectrum $Y$, there is a short exact sequence of Hopf algebras over $K(1)_*$ as follows: $$ K(1)_*(\mathbb PY) \rightarrow K(1)_*(\Omega^{\infty}Y) \rightarrow K(1)_*(\tau_{\leq 2}\Omega^{\infty}Y).$$ Here $\mathbb PY$ is the free $E_\infty$--algebra generated by $Y$: the wedge of all the extended powers of $Y$.

So if $X = \Omega^{\infty}Y$, then $K(1)_*(X) \rightarrow K(1)_*(\tau_{\leq 2}X)$ is onto. One could wonder if one has an epimorphism for other $X$. Okay, lets try something else: Bousfield gave a cute short argument that if $X$ is $E_*$--acyclic, so is $K(\pi_j(X), j)$ for any $j$. So if $X$ is $K(1)_*$--acyclic, so is $\tau_{\leq 2}X$, and thus our map is still an epi.

This is enough fun for one answer, so I'll leave things here.

$\endgroup$
  • $\begingroup$ Great, thanks! It's good to at least know that this map is decidedly nonzero for a good class of nontrucated spaces. Btw, where can I find Bousfield's cute argument? $\endgroup$ – Tim Campion Apr 18 at 21:50
  • 2
    $\begingroup$ @TimCampion See [Bousfield, A. K. On homology equivalences and homological localizations of spaces. Amer. J. Math. 104 (1982), no. 5, 1025–1042.] The cute idea: $SP^{\infty}$ preserves homology isomorphisms. $\endgroup$ – Nicholas Kuhn Apr 19 at 2:46

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.