7
$\begingroup$

Edit : There are several satisfying proofs in the case each $L^i$ is finite-dimensional. It is proven (for example, Hinich DG coalgebras as formal stacks) that for $A$ : local Artin ring then $\operatorname{Def}_L(A)=\operatorname{Hom}(A^*, H^0(\mathscr{C}(L)))$. So when everything is finite-dimensional, just use basic linear algebra and duality to get $\operatorname{Hom}(H^0(\mathscr{C}(L))^*, A)$. How could one reduce to that case ? If one can not, what happens ? How to obtain pro-representability ?


I work with deformation theory and DG Lie algebras, motivation coming from Goldman-Millson theory. Here DG Lie algebras are always assumed to have $L^i=0$ for $i<0$ and $H^i(L)$ finite-dimensional.

Context : To such a DG Lie algebra $L$ over a field $\mathbf{k}$ of characteristic zero is associated a functor $\operatorname{Def}_L$ on local Artin rings, sending $(A,\mathfrak{m}_A)$ so the set of solutions of the Maurer-Cartan equation $dx+\frac{1}{2}[x,x]=0$ in $L^1\otimes\mathfrak{m}_A$ modulo the exponential action of $L^0\otimes\mathfrak{m}_A$ by gauge transformations. It is known by classical methods in deformation theory that $\operatorname{Def}_L$ is pro-representable if $H^0(L)=0$ (Manetti, Deformation theory via differential graded Lie algebras, section 4 on the Kuranishi map).

On the other hand I'm reading about more modern deformation theories (Kontsevich, Hinich, Manetti...) They describe a functor which has several different names: Quillen $\mathscr{C}$ functor, bar construction, Chevalley-Eilenberg complex... obtained as follows : take the symmetric algebra on $L$ shifted by $1$, $\operatorname{Sym}(L[1])$, and turn it into a DG coalgebra $\mathscr{C}(L)$ (the codifferential has one part coming from the differential in $L$ and one part coming from the Lie bracket). Its dual $\mathscr{C}(L)^*$ is a DG complete local algebra.

Question : Now I would like the following theorem to be true but I never find it stated so explicitly and each time I try to extract it from the existing litterature (Kontsevich, Hinich, Manetti) there seems to be some subtle points (about duality algebras-coalgebras, finite-dimensionality, or algebras up to quasi-isomorphism) that I may not understand.

Let $L$ be a DG Lie algebra ($L^i=0$ for $i<0$ and $H^i(L)$ finite-dimensional). If $H^0(L)=0$ then $\operatorname{Def}_L$ is pro-represented by $H^0(\mathscr{C}(L))^*$.

Is this theorem true ? I would like some help to understand it (I work at the most "concrete" possible level of deformation theory, with the least possible amount of $\infty$ / simplicial / derived techniques). I'm suprised I never see this so stated nor used. It means, that $\operatorname{Def}_L$ is pro-represented by some very explicit and functorial object, which is much better than choosing a non-canonical splitting of $L^1$ as is done in Manetti's lecture notes.

$\endgroup$
6
  • $\begingroup$ Do you work with dg Artin rings or just Artin rings concentrated in degree $0$? I'm think all the results you stated fail in the dg case, but there is a good chance that everything works out more or less as you want in the non-dg case. $\endgroup$ Mar 7, 2017 at 13:23
  • $\begingroup$ Exactly I want to work in the non-DG case ("classical deformation theory" à la Schlessinger) and it seems that to get this theorem, one has to go through higher deformation theories (DG Artin rings, coalgebras, simplicial sets...) That's why I'm having such a hard time coming back to the classical case ! $\endgroup$ Mar 7, 2017 at 13:34
  • $\begingroup$ I think I have something close to what you would like, but not exactly that. I can write it as an answer in the next couple of days if you want. I just need to know: how familiar are you with the language of operads? $\endgroup$ Mar 7, 2017 at 15:18
  • $\begingroup$ I'm sorry I'm not familiar with operads. But anyway I need to understand this theorems and what could go wrong. What result do you obtain ? $\endgroup$ Mar 14, 2017 at 16:12
  • $\begingroup$ Never mind, I would basically have recovered Hinich's result you put in your edit (using more general results). I really think you need some kind of finite-dimensionality assumption if you want to dualize. I would advise working with coalgebras. It can be a bit hard to get used to it, but once you're familiar with the objects then everything becomes very natural. $\endgroup$ Mar 16, 2017 at 1:18

1 Answer 1

3
$\begingroup$

Edit: The answer below only works for the case where all the $L^i$'s are finite dimensional.

The statement is true. In fact, under the assumptions of the question, it is also true that ${\scr C}(L)^*$ pro-represents the associated derived moduli problem for commutative artinian dg-algebras. Recall that the derived deformation problem determined by $L$ associates to a commutative artinian dg-algebra $A$ the derived mapping space ${\rm Map}_{{\rm dg-Lie}}({\scr D}A,L)$, where ${\scr D}A$ is the dg-lie algebra Koszul dual to $A$. On the other hand, the commutative dg-algebra ${\scr C}(L)^*$ is the Koszul dual of $L$. Furthermore, under the given assumptions $L$ is also the Koszul dual of ${\scr C}^*(L)$ (see Proposition 13.3.1.1 of http://www.math.harvard.edu/~lurie/papers/SAG-rootfile.pdf). One then obtains a natural equivalence ${\rm Map}_{{\rm dg-Lie}}({\scr D}A,L) \simeq {\rm Map}_{{\rm dg-Lie}}({\scr D}A,{\scr D}{\scr C}(L)^*) \simeq {\rm Map}_{{\rm dg-com}}({\scr C}(L)^*,A)$. To get the pro-representability in the classical setting use the fact if $A$ is concentrated in degree 0 then ${\rm Map}_{{\rm dg-com}}({\scr C}(L)^*,A) \simeq {\rm Map}_{{\rm com}}(H_0({\scr C}(L)^*),A)$. Note that if the cohomology groups $H^i(L)$ are not finite dimensional but $H^0(L)$ is still 0 then then the moduli problem associated to $L$ is still pro-representable (see Proposition 13.3.3.1 of http://www.math.harvard.edu/~lurie/papers/SAG-rootfile.pdf), but just not usually by ${\scr C}(L)^*$, since in this case $L$ is generally not the Koszul dual of ${\scr C}(L)^*$.

$\endgroup$
2
  • 1
    $\begingroup$ First, thank you for answering and confirming this is true. However, Lurie's theory is not an easy thing ! I would like to understand it with more down-to-earth arguments. In particular in section 13.3 you quote, Koszul duality seems very abstract but it is written that it can be be more explicitly obtained via $L^\infty$ algebras.. ? By the way the theorem you quote is for $L^i$ finite-dimensional. In that case I already have a satisfying proof (use Hinich + duality algebras-coalgebras). But I have $H^i(L)$ finite-dimensional, how to reduce to that ? $\endgroup$ Mar 10, 2017 at 14:23
  • $\begingroup$ You're right, there is a gap here. I'm not sure how to reduce from $H^i(L)$ finite to $L^i$ finite. Maybe one can try to argue by comparing the lie-commutative Koszul duality with the associative-associative Koszul duality (and then use Corollary 14.1.3.3.of SAG). $\endgroup$ Mar 11, 2017 at 20:49

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.