I know very little about algebraic topology, and more about $k$-linear stable $\infty$-categories (i.e. homological algebra).

Given an abelian group $A$, there is the Eilenberg-Mac Lane spectrum $HA$, for which: $\pi_{< 0}(\cdot) = 0$, $\pi_0 (\cdot) = A$, $\pi_{>0}(\cdot) = 0$. There is also the Moore spectrum $SA$, for which $\pi_{<0}(\cdot) = 0$, $\pi_{=0} (\cdot)= A$, $H_{> 0} (\cdot , \mathbb{Z}) = 0$.

I more-or-less have some feeling for $HA$ (under Dold-Kan, it is simply $A$, so one simply gets connected deloopings $BA$, $BBA$, etc. making the infinite loop space structure explicit).

I have no feeling what-so-ever about $SA$.

But reading about Bousfield localization, it seems that most the standard examples are with $SA$. How to understand this for a novice? Say, for $p$-localization and $p$-completion, how do I understand that I "want" to use the Moore spectrum and not the Eilenberg-Mac Lane spectrum?

Or, at least, what is the rationale for introducing Moore spectra?

Thank you

  • 4
    $\begingroup$ Denis's answer is 100% the reason why the specific object S/p appears in the theory of localization. You could still ask whether you could use HZ/p, and here the answer is somewhat unintuitive: for connective spectra, localization at S/p and at HZ/p agree, and so you cannot tell the difference before passing to nonconnective spectra, which are somewhat exotic objects. The spectrum $KU := \Sigma^\infty_+ \mathbb CP^\infty[\beta^{-1}]$ is a concrete example of a spectrum with $H_*(KU; Z)$ rational, hence $H_*(KU; Z/p)$ and $L_{HZ/p} KU$ null, but $L_{S/p} KU$ nontrivial. $\endgroup$ – Eric Peterson Feb 2 '19 at 13:26
  • $\begingroup$ @EricPeterson: Thank you! I will have to think about it. $\endgroup$ – Sasha Feb 2 '19 at 19:40

I'm not sure I can answer to the general question, but I can explain why the Moore spectrum $\mathbb{S}/p$ shows up in the discussion of Bousfield localizations.

This is just the cofiber of multiplication by $p$ from $\mathbb{S}$ to itself. Hence saying that for a spectrum $X$ we have $X\wedge \mathbb{S}/p=0$ means simply that multiplication by $p$ acts on $X$ as an equivalence. So $X$ is $\mathbb{S}[1/p]$-local iff it is $\mathbb{S}/p$-acyclic. From standard localization arguments (plus the fact that $\mathbb{S}/p$ is a finite spectrum) you get the fracture square relating $\mathbb{S}[1/p]$-localization and $p$-completion (i.e. the "complementary localization" of $\mathbb{S}[1/p]$-localization).

So, essentially, the reason it shows up is because the Moore spectrum is the layer of the diagram $$ \mathbb{S} \xrightarrow{p} \mathbb{S}\xrightarrow{p}\mathbb{S}\xrightarrow{p}\mathbb{S}\xrightarrow{p}\cdots $$ that computes $\mathbb{S}[1/p]$-localization. From this it is clear that $\mathbb{S}/p^n$ is also going to be relevant to the study of $p$-completion (since these are just the "higher" layers of the tower), and from there to $$\mathbb{S}/p^\infty = \mathbb{S}\mathbb{Q}_p/\mathbb{Z}_p = \mathrm{colim}\,\mathbb{S}/p^n$$ and $$\mathbb{S}\mathbb{Q}/\mathbb{Z}=\bigvee_p \mathbb{S}/p^\infty$$ is just a small step.

  • $\begingroup$ Thank you! This now makes perfect sense. But for p=2, it seems that the cofiber of multiplication by 2 on the sphere spectrum has $\pi_0$ with four elements, so this should differ from the Moore spectrum as defined in my question, no? $\endgroup$ – Sasha Feb 2 '19 at 19:40
  • $\begingroup$ @Sasha No, $π_0\mathbb{S}/2=H_0\mathbb{S}/2=\mathbb{Z}/2$. Maybe you're confounding it with $[\mathbb{S}/2,\mathbb{S}/2]=\mathbb{Z}/4$? (Yes, it is a bit counterintuitive but this nontrivial extension is secretly coming from $\pi_1\mathbb{S}$) $\endgroup$ – Denis Nardin Feb 2 '19 at 19:41
  • $\begingroup$ Oops, I am used to cohomological convention, so I simply calculated wrongly the long exact sequence of homotopy groups for the cofiber sequence. Thank you! $\endgroup$ – Sasha Feb 2 '19 at 19:46
  • $\begingroup$ @Sasha Well, as I'm fond of saying, the cohomological convention is the tool of the Devil. Don't fall for it! ;) $\endgroup$ – Denis Nardin Feb 2 '19 at 19:49
  • $\begingroup$ I don't quite have a say, since this is so common in homological algebra (representaiton theory etc.) that I can't go against it. But perhaps it makes sense to put $\pi_1$ in negative degree, since it represents "smaller" stuff? (like the cardinality of $BG$ is $|G|^{-1}$ etc.) Anyway, probably one can find justification for each. $\endgroup$ – Sasha Feb 2 '19 at 19:52

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.