7
$\begingroup$

I'm trying to figure out what an ideal of a, say, dg-algebra (or, if you prefer, $\mathbb E_1$-ring) $R$ is in a homotopically relevant fashion, but I can't actually figure it out. I can assume that $R$ is concentrated in cohomologically nonpositive degrees (or homologically nonnegative degrees). I have stumbled upon a few possibilities:

  1. There is a notion of monomorphism in an $\infty$-category; hence, I would consider the derived category $\mathsf D(R)$ of $R$-dg-modules as an $\infty$-category and say that $I \to R$ is an "ideal" if it is a monomorphism according to that notion. This notion is also used in Spectral algebraic geometry (Remark C.2.3.4. page 1965) in the framework of Grothendieck prestable $\infty$-categories.
  2. On the other hand, I find other sources such as this and this. From the first one, I quote:

    There are other things that are weird for commutative ring spectra. Quite often, we end up working with ideals in the graded commutative ring of homotopy groups, but as we saw above, this is not a suitable notion of ideal.There is a notion of an ideal in the context of (commutative) ring spectra [53] due to Jeff Smith, but still several algebraic constructions do not have an analogue in spectra.

Given that, I'm pretty confused. Perhaps the notion of monomorphism (1) is fine, but in the case of commutative ring spectra it does not work really well, hence the issues I found (2)? I've tried to skim through some literature on derived algebraic geometry, but still I couldn't find any satisfying answer...

$\endgroup$
2
  • 1
    $\begingroup$ Homotopy fibre of an $H_0$-surjective morphism of algebras is probably the best you'll manage, and that only works if the algebras are connective. In the derived setting, ideals have much more structure than modules - see for instance Ciocan-Fontanine and Kapranov's papers on derived Hilb and derived Quot. $\endgroup$ May 28, 2020 at 13:37
  • $\begingroup$ I characteristic zero, I have suggested a definition of "ideal" in my arxiv.org/abs/1608.08598. Namely, I suggest that the definition of "idea" should be "homotopy fibre of a homomorphism", and the paper gives a universal description of the algebraic structure carried by such "ideals" for homomorphisms of any (di)operad. I should warn that that paper needs some rewriting, and my interests have shifted a bit, so I'm not sure when I will do it. I also have not done a good enough job comparing my proposals with the rest of the literature. $\endgroup$ May 28, 2020 at 22:50

2 Answers 2

3
$\begingroup$

In (2), you linked to Mark Hovey's paper on Smith ideals, and mentioned "the commutative framework." But Hovey explicitly writes "we have not dealt with the commutative situation at all," so I don't know what you mean. However, if you do want a theory of commutative Smith ideals, you can find this in my first paper. Also, if you want a theory of Smith $O$-algebras, for an operad $O$ (e.g., $O = E_n$), then you can find this in a paper of mine with Donald Yau.

One of the crucial aspects of the story is that the algebraic structure on the morphism $f: I\to R$, viewed as an object in the arrow category, matches the algebraic structure on the cofiber of $f$ (at least, in stable settings where taking the cofiber makes sense). This is proven in section 4 of Hovey's paper, and in Theorem 4.4.1 of my paper with Donald Yau (with lots of examples occupying the rest of the paper). I'd like to write more, but have to run off to a Zoom meeting now for the rest of the day. Hopefully this observation gets you started. It's an important justification for the approach of both (1) and (2).

$\endgroup$
3
  • $\begingroup$ This is even more confusing, namely, here: math.uni-hamburg.de/home/richter/commutativeringspectra.pdf I quote "There is a notion of an ideal in the context of (commutative) ring spectra [53] due to Jeff Smith, but still several algebraic constructions donot have an analogue in spectra." That's why I thought it was a notion meant for the "commutative framework". Also, I don't see how the approaches of (1) and (2) relate. $\endgroup$ May 28, 2020 at 14:08
  • $\begingroup$ Probably Birgit had in mind that I'd already sorted out the commutative case, and I was Hovey's student (graduating in 2014). So, she might have just kind of given him credit there for the part of my first semester that was about commutative smith ideals. Not a big deal. Sorry if that caused confusion. $\endgroup$ May 28, 2020 at 15:10
  • 1
    $\begingroup$ My previous comment should have said "the part of my first paper" - sorry, this is what I get for typing on mathoverflow during a Zoom meeting about the fall semester. $\endgroup$ May 28, 2020 at 17:27
3
$\begingroup$

At least in commutative situations, I would argue that a good notion is simply an ideal in $H^0(R)$.

For example, the theory of local cohomology works just as well as it does for commutative rings, as long as you do it with respect to ideals in $H^0(R)$.

Similarly, you can take "derived quotients" with respect to a finite sequence of elements in $H^0(R)$, by taking the Koszul complex with respect to such a sequence.

See for example my very recent paper from last week:

"Koszul complexes over Cohen-Macaulay rings"

https://arxiv.org/abs/2005.10764

$\endgroup$
2
  • $\begingroup$ Welcome to MathOverflow! I tend to disagree with your answer for the case of spectra. The philosophy of Jeff Smith and Mark Hovey, and the applications to K-theory Smith envisioned, really need some kind of object at the point-set level, not just in the homotopy category. That said, Proposition 1.7 in Hovey's paper unwinds the definition of a Smith ideal into the language of R-bimodules, and might help bridge between your answer and mine. $\endgroup$ May 28, 2020 at 17:27
  • $\begingroup$ @DavidWhite, thanks! my answer was about DG-algebras. I have no experience with spectra. $\endgroup$ May 28, 2020 at 17:31

Your Answer

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