The idea "Formal deformation theory in characteristic zero is controlled by a differential graded Lie algebra (dgla)" goes back to Goldman-Millson, Deligne, Drinfeld among others; see Lurie's ICM talk.

What is the analogue of this, over positive characteristic? Specifically, what replaces the Maurer-Cartan equation $$dx + \frac{1}{2}[x,x] =0$$ in positive characteristic?

Do formal groups enter in an essential way, as a substitute for Lie algebras, in deformation theory in positive characteristic? This point is not answered in the responses to a similar question Extended Deformation Theory (dg-Lie algebra principle in positive characteristic?) .

  • $\begingroup$ The results of Lurie and Pridham have now been extended to positive and mixed characteristics by Brantner and Mathew using so-called "partition Lie algebras", see arxiv.org/abs/1904.07352v2. $\endgroup$ Oct 30, 2021 at 17:41

1 Answer 1


Since the philosophy that deformation theory is controlled by DGLAs long predates the abstract characterisation of formal moduli problems, I'll break the answer in two. My answer will also hold in mixed characteristic, where things are complicated by the absence of a basepoints for the deformation functor (objects over $\mathbb{F}_p$ don't tend to have canonical deformations over $\mathbb{Z}_p$).

A) Deformation problems arising naturally tend to be governed by the SDCs of P-, Deformations of schemes and other bialgebraic structures, Trans. AMS 2008, which is mostly contained in arXiv:math.AG/0311168. An SDC $E^{\bullet}$ is a diagram of formal schemes with an associative product $$ E^i \times E^j \xrightarrow{*} E^{i+j}, $$ together with a unit $1 \in E^0$ and all the structure of a cosimplicial diagram except the outer coface maps $\partial^0, \partial^{n+1}$ (up to relabelling, this is the same structure as an augmented simplicial diagram); these satisfy various compatibility conditions. Cosimplicial formal groups give rise to SDCs, but do not exhaust them.

The Maurer-Cartan space of an SDC is then $$ \{\omega \in E^1 ~:~ \partial^1\omega=\omega*\omega\}, $$ on which $E^0$ acts by conjugation. In characteristic $0$, the homotopy categories of SDCs and of non-negatively graded DGLAs are equivalent, with equivalent Deligne groupids.

The constructions of P-, The homotopy theory of strong homotopy algebras and bialgebras, HHA 2010, arXiv:0908.0116 show how SDCs give rise to formal derived moduli stacks. Since SDCs have no negative terms, they can only give rise to derived $1$-stacks, but higher stacks can be obtained by incorporating additional simplicial structure.

B) In the proof P-, Unifying derived deformation theories, Advances 2010, arXiv:0705.0344 of the equivalence between DGLAs and formal derived stacks (later dubbed "formal moduli problems" by Lurie), most of the steps hold in positive or mixed characteristic.

This gives an equivalence between formal derived stacks and certain simplicial cosimplicial formal schemes up to tangent quasi-isomorphism. The relation with DGLAs is that in characteristic $0$, the simplicial and cosimplicial structures can be replaced with dg structures, and these become equivalent to Hinich's dg coalgebras, and hence to $L_{\infty}$-algebras via the bar construction.

  • $\begingroup$ Thanks! This is a great answer! Could you please also point me to some worked out examples illustrating the difference between SDC's and DGLA and also the SDC's coming from formal groups? Many thanks for proving these results! $\endgroup$
    – guest
    Sep 17, 2017 at 21:20
  • $\begingroup$ As in math.AG/0311168 5.4.1, SDCs come from cosimplicial groups just by setting * to be the Alexander-Whitney cup product. That preprint and the sequel "Deforming l-adic representations of the fundamental group" JAG 2006 give a few examples. There are more in "Constructing derived moduli stacks", which works with global moduli. The differences arise outside characteristic 0, where DGLAs don't give Deligne groupoids because the Lie algebra won't exponentiate. In mixed characteristic, cosimplicial groups always give a distinguished point in the deformation functor, so can't model everything. $\endgroup$ Sep 18, 2017 at 7:58

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.