Symplectic and Poisson derived geometry
and deformation quantization

T. Pantev, G. Vezzosi
Abstract

We review recent results and ongoing investigation of the symplectic and Poisson geometry of derived moduli spaces, and describe applications to deformation quantization of such spaces.

Introduction

From the vantage point of the timeline of the AMS Summer Institutes, this contribution is a continuation and an update of B. Toën’s 2005 overview [To1]. Our goal here is to highlight some of the remarkable developments in derived geometry that we witnessed in the past ten years.

Our main topics - symplectic and Poisson geometry on derived moduli spaces - are among the latest topics in the area. Even though the study of these topics is still in an early stage, it has already led to some exciting applications. Among the moduli theoretic applications, we might mention the fact that (1)1(-1)-shifted symplectic structures induce symmetric perfect obstruction theories in the sense of [Be-Fa] (and actually all the geometrically interesting examples of symmetric obstruction theories arise this way), and the related important result that the Donaldson-Thomas moduli space is (1)1(-1)-shifted symplectic and Zariski locally isomorphic to the critical locus of a potential [BBJ].

Section 1 explains and summarizes the main results from [PTVV], while sections 2 and 3 delve into the substance of [CPTVV]. The aim of our review has been twofold: on one hand to convey the intuition behind definitions, constructions, and proofs of the main results, and on the other hand, to explain and motivate the slight change of point of view going from [PTVV] to [CPTVV]. The upshot is that shifted Poisson geometry and its applications to deformation quantization of derived categories require a new broader perspective and new technical tools, i.e. differential calculus in an extremely general setting (Section 2.1), and formal localization (Section 2.2). Even though these powerful tools were created in order to solve our specific problems related to shifted Poisson structures, they constitute also a conceptual advance, very likely to become relevant in other contexts and to different problems in derived algebraic geometry.

Note that there is a parallel theory of shifted quadratic forms on derived moduli spaces, but we will not review it here (see [Vez2] for a first investigation).

We now describe, section by section, the mathematical contents of this paper in more details.

Shifted symplectic structures. A shifted symplectic structure on a derived stack X𝑋X with a perfect cotangent complex 𝕃Xsubscript𝕃𝑋\mathbb{L}_{X} is a structured self-duality of 𝕃Xsubscript𝕃𝑋\mathbb{L}_{X} up to a shift, i.e. a quasi-isomorphism 𝕋X𝕃X[n]similar-to-or-equalssubscript𝕋𝑋subscript𝕃𝑋delimited-[]𝑛\mathbb{T}_{X}\simeq\mathbb{L}_{X}[n] induced by a closed n𝑛n-shifted 222-from on X𝑋X. The idea is an obvious generalization of the classical definition of symplectic form, but with an important additional, purely derived algebro-geometrical feature: for a shifted form on X𝑋X, being closed is not a property but rather a datum. In other words, there is a canonical map, called the underlying-form map, from the space 𝒜p,cl(X,n)superscript𝒜𝑝cl𝑋𝑛\mathcal{A}^{p,\textrm{cl}}(X,n) of n𝑛n-shifted closed p𝑝p-forms on X𝑋X, to the space 𝒜p(X,n)superscript𝒜𝑝𝑋𝑛\mathcal{A}^{p}(X,n) of n𝑛n-shifted p𝑝p-forms on X𝑋X, but this map is not, in general, “injective” in any reasonable sense of the word (e.g. not injective on the connected components of these spaces). The space 𝖲𝗒𝗆𝗉𝗅(X,n)𝖲𝗒𝗆𝗉𝗅𝑋𝑛\mathsf{Sympl}(X,n) is exactly the subspace of 𝒜2,cl(X,n)superscript𝒜2cl𝑋𝑛\mathcal{A}^{2,\textrm{cl}}(X,n) of closed 222-forms whose underlying 222-from is non-degenerate, i.e. such that the induced map 𝕋X𝕃X[n]subscript𝕋𝑋subscript𝕃𝑋delimited-[]𝑛\mathbb{T}_{X}\to\mathbb{L}_{X}[n] is a quasi-isomorphism. Shifted symplectic structures abound, in the sense that many moduli spaces of interest to algebraic geometers and topologists, such as the moduli spaces of principal bundles or perfect complexes on algebraic Calabi-Yau manifolds or compact orientable topological manifolds, have derived enhancements carrying natural shifted symplectic structures. In Section 1 we give three general existence results for shifted symplectic structures on derived moduli stacks, leading to a long list of examples.

Shifted Poisson structures. Having at our disposal a theory of shifted symplectic structures, it is natural to look for a more general theory of shifted Poisson structures on derived moduli stacks. Actually, our original motivation for such a general theory came from the expected link between a shifted Poisson structure on a derived stack and an induced deformation quantization of its dg-derived category of perfect complexes. We will say more about this motivation-application below. While classically, setting up a theory of Poisson varieties does not present more difficulties than setting up a theory of symplectic varieties, in derived algebraic geometry the situation is radically different. The usual bad functoriality properties of shifted polyvectors (as opposed to the good functoriality of closed shifted forms) together with the very delicate and intricate strictification problems related to establishing a meaningful shifted Poisson algebra structure on them, immediately made us realize that, outside the derived Deligne-Mumford case, a full-fledged theory of shifted Poisson structures on derived Artin stacks required new ideas and tools. If X𝑋X is a derived Artin stack, locally of finite presentation over the base k𝑘k (always assumed to be a Noetherian commutative \mathbb{Q}-algebra), then its cotangent complex 𝕃Xsubscript𝕃𝑋\mathbb{L}_{X} is perfect, and we may consider the graded commutative differential graded algebra

Pol(X,m):=p0Γ(X,Symp(𝕋X[m])Pol(X,m):=\bigoplus_{p\geq 0}\Gamma(X,Sym^{p}(\mathbb{T}_{X}[-m])

of m𝑚m-shifted polyvectors on X𝑋X. Here, m𝑚m\in\mathbb{Z}, 𝕋Xsubscript𝕋𝑋\mathbb{T}_{X} is the tangent complex of X𝑋X, i.e. the 𝒪Xsubscript𝒪𝑋\mathcal{O}_{X}-dual of 𝕃Xsubscript𝕃𝑋\mathbb{L}_{X}, ΓΓ\Gamma denotes the derived functor of global sections (i.e. hypercohomology), and the external grading, called weight grading, is given by p𝑝p. In order to define shifted Poisson structures on X𝑋X, we have to endow Pol(X,m)𝑃𝑜𝑙𝑋𝑚Pol(X,m) with a degree m𝑚m and weight 11-1 Lie bracket, making it into a graded m+1subscript𝑚1\mathbb{P}_{m+1}-commutative differential graded algebra over k𝑘k. In particular, Pol(X,m)[m]𝑃𝑜𝑙𝑋𝑚delimited-[]𝑚Pol(X,m)[m] will be a graded dg-Lie algebra over k𝑘k, with a weight 11-1 Lie bracket. Then we could adopt the following derived variation of a classical definition of Poisson structure, and put

𝖯𝗈𝗂𝗌𝗌(X,n):=𝖬𝖺𝗉𝖽𝗀𝗅𝗂𝖾kgr(k(2)[1],Pol(X,n+1)[n+1])assign𝖯𝗈𝗂𝗌𝗌𝑋𝑛subscript𝖬𝖺𝗉subscriptsuperscript𝖽𝗀𝗅𝗂𝖾𝑔𝑟𝑘𝑘2delimited-[]1𝑃𝑜𝑙𝑋𝑛1delimited-[]𝑛1\mathsf{Poiss}(X,n):=\mathsf{Map}_{\mathsf{dglie}^{gr}_{k}}(k(2)[-1],Pol(X,n+1)[n+1])

for the space 𝖯𝗈𝗂𝗌𝗌(X,n)𝖯𝗈𝗂𝗌𝗌𝑋𝑛\mathsf{Poiss}(X,n) of n𝑛n-shifted Poisson structures on X𝑋X, for n𝑛n\in\mathbb{Z}. Here k(2)[1]𝑘2delimited-[]1k(2)[-1] is the graded dg-Lie algebra consisting of k𝑘k in pure cohomological degree 1, pure weight 2, and trivial bracket, obviously. If X𝑋X is a smooth underived scheme, n=0𝑛0n=0, and we replace the mapping space 𝖬𝖺𝗉𝖽𝗀𝗅𝗂𝖾kgrsubscript𝖬𝖺𝗉subscriptsuperscript𝖽𝗀𝗅𝗂𝖾𝑔𝑟𝑘\mathsf{Map}_{\mathsf{dglie}^{gr}_{k}} in the model category 𝖽𝗀𝗅𝗂𝖾kgrsubscriptsuperscript𝖽𝗀𝗅𝗂𝖾𝑔𝑟𝑘\mathsf{dglie}^{gr}_{k}, with its Hom set Hom𝖽𝗀𝗅𝗂𝖾kgr𝐻𝑜subscript𝑚subscriptsuperscript𝖽𝗀𝗅𝗂𝖾𝑔𝑟𝑘Hom_{\mathsf{dglie}^{gr}_{k}}, then we obtain the set of bivectors π𝜋\pi on X𝑋X, whose Schouten-Nijenhuis self bracket [π,π]𝜋𝜋[\pi,\pi] is zero, i.e. exactly the set of Poisson bivectors on X𝑋X. The functoriality problems mentioned above prevent any elementary, easy extension of (a shifted version of) the usual Schouten-Nijenhuis bracket when X𝑋X is a general derived Artin stack, locally of finite presentation over k𝑘k. Hence, there is no elementary, easy way of making sense of the above definition of 𝖯𝗈𝗂𝗌𝗌(X,n)𝖯𝗈𝗂𝗌𝗌𝑋𝑛\mathsf{Poiss}(X,n).

Our solution to this problem consists of two steps. First of all, in Section 2.1 we build a very general theory of differential calculus, including de Rham algebras and polyvectors, in an arbitrary symmetric monoidal model \infty-category \mathcal{M} enriched over k𝑘k-dg modules (and satisfying suitable, mild hypotheses). In particular, for any commutative algebra A𝐴A in \mathcal{M}, we are able to make sense of the space 𝖲𝗒𝗆𝗉𝗅(A,n)𝖲𝗒𝗆𝗉𝗅𝐴𝑛\mathsf{Sympl}(A,n) of n𝑛n-shifted symplectic structures on A𝐴A, to define a n+1subscript𝑛1\mathbb{P}_{n+1}-commutative differential graded algebra 𝖯𝗈𝗅(A,n)𝖯𝗈𝗅𝐴𝑛\mathsf{Pol}(A,n) over k𝑘k of n𝑛n-shifted polyvectors on A𝐴A, and hence to make sense of the space 𝖯𝗈𝗂𝗌𝗌(A,n)𝖯𝗈𝗂𝗌𝗌𝐴𝑛\mathsf{Poiss}(A,n) of n𝑛n-shifted Poisson structures on A𝐴A, as explained above. Moreover, we produce a general comparison map from the the space 𝖯𝗈𝗂𝗌𝗌nd(A,n)superscript𝖯𝗈𝗂𝗌𝗌nd𝐴𝑛\mathsf{Poiss}^{\textrm{nd}}(A,n), of suitably defined non-degenerate n𝑛n-shifted Poisson structures on A𝐴A to 𝖲𝗒𝗆𝗉𝗅(A,n)𝖲𝗒𝗆𝗉𝗅𝐴𝑛\mathsf{Sympl}(A,n). The second step is what we call formal localization (Section 2.2), and it concerns essentially the study of the map XXDR𝑋subscript𝑋𝐷𝑅X\to X_{DR}, for X𝑋X derived Artin stack, locally of finite presentation over k𝑘k. Here XDRsubscript𝑋𝐷𝑅X_{DR} is the de Rham stack of X𝑋X (Definition 2.2.3), and the fiber of XXDR𝑋subscript𝑋𝐷𝑅X\to X_{DR} at a closed point x¯:𝐒𝐩𝐞𝐜𝕂XDR:¯𝑥𝐒𝐩𝐞𝐜𝕂subscript𝑋𝐷𝑅\overline{x}:\mathbf{Spec}\,\mathbb{K}\to X_{DR} is the formal completion Xx^^subscript𝑋𝑥\widehat{X_{x}} of X𝑋X at the corresponding point xX𝑥𝑋x\in X 111Note that X𝑋X and XDRsubscript𝑋𝐷𝑅X_{DR} have the same reduced points.; hence XXDR𝑋subscript𝑋𝐷𝑅X\to X_{DR} can be viewed as the family of formal completions of X𝑋X.

The remarkable properties of the map XXDR𝑋subscript𝑋𝐷𝑅X\to X_{DR}, allow us to associate to any derived Artin stack X𝑋X, locally of finite presentation over k𝑘k, a commutative algebra 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty) in a suitable symmetric monoidal model \infty-category Xsubscript𝑋\mathcal{M}_{X} constructed out of X𝑋X, such that:

  • There is an equivalence of spaces 𝖲𝗒𝗆𝗉𝗅(𝒫X(),n)𝖲𝗒𝗆𝗉(X,n)similar-to-or-equals𝖲𝗒𝗆𝗉𝗅subscript𝒫𝑋𝑛𝖲𝗒𝗆𝗉𝑋𝑛\mathsf{Sympl}(\mathcal{P}_{X}(\infty),n)\simeq\mathsf{Symp}(X,n).

  • After forgetting the bracket 𝖯𝗈𝗅(𝒫X(),n)𝖯𝗈𝗅subscript𝒫𝑋𝑛\mathsf{Pol}(\mathcal{P}_{X}(\infty),n) is equivalent to Pol(X,n)𝑃𝑜𝑙𝑋𝑛Pol(X,n), in other words we finally have a way of endowing Pol(X,n)𝑃𝑜𝑙𝑋𝑛Pol(X,n) with the structure of a graded n+1subscript𝑛1\mathbb{P}_{n+1}-commutative differential graded algebra over k𝑘k. In particular, it now makes sense to define 𝖯𝗈𝗂𝗌𝗌(X,n):=𝖬𝖺𝗉𝖽𝗀𝗅𝗂𝖾kgr(k(2)[1],𝖯𝗈𝗅(𝒫X(),n+1)[n+1])assign𝖯𝗈𝗂𝗌𝗌𝑋𝑛subscript𝖬𝖺𝗉subscriptsuperscript𝖽𝗀𝗅𝗂𝖾𝑔𝑟𝑘𝑘2delimited-[]1𝖯𝗈𝗅subscript𝒫𝑋𝑛1delimited-[]𝑛1\mathsf{Poiss}(X,n):=\mathsf{Map}_{\mathsf{dglie}^{gr}_{k}}(k(2)[-1],\mathsf{Pol}(\mathcal{P}_{X}(\infty),n+1)[n+1]).

  • The canonical map 𝖯𝗈𝗂𝗌𝗌nd(X,n)=𝖯𝗈𝗂𝗌𝗌nd(𝒫X(),n)𝖲𝗒𝗆𝗉𝗅(𝒫X(),n)𝖲𝗒𝗆𝗉(X,n)superscript𝖯𝗈𝗂𝗌𝗌nd𝑋𝑛superscript𝖯𝗈𝗂𝗌𝗌ndsubscript𝒫𝑋𝑛𝖲𝗒𝗆𝗉𝗅subscript𝒫𝑋𝑛similar-to-or-equals𝖲𝗒𝗆𝗉𝑋𝑛\mathsf{Poiss}^{\textrm{nd}}(X,n)=\mathsf{Poiss}^{\textrm{nd}}(\mathcal{P}_{X}(\infty),n)\to\mathsf{Sympl}(\mathcal{P}_{X}(\infty),n)\simeq\mathsf{Symp}(X,n) is an equivalence of spaces.

  • The \infty-category 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) of perfect complexes on X𝑋X is equivalent to a suitably defined \infty-category of perfect 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty)-dg-modules.

Let us briefly describe the commutative algebra object 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty) in Xsubscript𝑋\mathcal{M}_{X}. As already observed, the canonical map XXDR𝑋subscript𝑋𝐷𝑅X\longrightarrow X_{DR} realizes X𝑋X as a family of formal derived stacks over XDRsubscript𝑋𝐷𝑅X_{DR}, namely as the family of formal completions at closed points of X𝑋X. By [Lu2] each of these formal completions is determined by a dg-Lie algebra xsubscript𝑥\ell_{x}. The collection of dg-Lie algebras xsubscript𝑥\ell_{x} does not fit together globally in a sheaf of dg-Lie algebras over XDRsubscript𝑋𝐷𝑅X_{DR}, simply because its underlying complex is the shifted tangent complex 𝕋X[1]subscript𝕋𝑋delimited-[]1\mathbb{T}_{X}[-1] of X𝑋X (see [Hen]), which in general does not admit a flat connection and thus does not descend to XDRsubscript𝑋𝐷𝑅X_{DR}. However, a remarkable consequence of derived formal localization is that the Chevalley-Eilenberg complexes of xsubscript𝑥\ell_{x}, viewed as graded mixed commutative dg-algebras, do fit together into a global object over XDRsubscript𝑋𝐷𝑅X_{DR}. Up to a twist (by k()𝑘k(\infty), see Section 2.1), this is exactly 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty). Thus, formal localization tells us how to express global geometric objects on X𝑋X as correspondingly sheafified objects on XDRsubscript𝑋𝐷𝑅X_{DR} related to 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty).

Deformation quantization of derived categories. One of our main original motivations for developing a theory of shifted symplectic and Poisson structures on derived moduli spaces was in fact a prospective application to deformation quantization of derived categories of perfect complexes. We are now able to obtain such applications, and we will briefly describe them here (for more details, see Section 3). We start by defining the deformation quantization problem for n𝑛n-shifted Poisson structures, whenever n0𝑛0n\geq 0. For every such n𝑛n, we consider a 𝔾msubscript𝔾𝑚\mathbb{G}_{m}-equivariant 𝔸k1subscriptsuperscript𝔸1𝑘\mathbb{A}^{1}_{k}-family of k𝑘k-dg-operads 𝔹𝔻n+1𝔹subscript𝔻𝑛1\mathbb{BD}_{n+1} such that its 00-fiber is the Poisson operad n+1subscript𝑛1\mathbb{P}_{n+1} and its generic fiber is the k𝑘k-dg-operad 𝔼n+1subscript𝔼𝑛1\mathbb{E}_{n+1} of chains of the topological operad of little (n+1)𝑛1(n+1)-disks. The general deformation quantization problem can then be loosely stated as follows:

Deformation Quantization Problem. Given a n+1subscript𝑛1\mathbb{P}_{n+1}-algebra stucture on an object in a k𝑘k-linear symmetric monoidal \infty-categorie, does it exist a family of 𝔹𝔻n+1𝔹subscript𝔻𝑛1\mathbb{BD}_{n+1}-algebra structures such that its 00-fiber is the original n+1subscript𝑛1\mathbb{P}_{n+1}-algebra structure ?

To be more precise, let now X𝑋X be a derived Artin stack locally of finite presentation over k𝑘k, and equipped with an n𝑛n-shifted Poisson structure. Using the formality of the 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-operad, and the fact that for n1𝑛1n\geq 1 the homology operad of 𝔼n+1subscript𝔼𝑛1\mathbb{E}_{n+1} is n+1subscript𝑛1\mathbb{P}_{n+1}, we can solve the deformation quantization problem above for the n+1subscript𝑛1\mathbb{P}_{n+1}-algebra structure on 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty). This gives us, in particular, a 𝔾msubscript𝔾𝑚\mathbb{G}_{m}-equivariant 111-parameter family of 𝔼n+1subscript𝔼𝑛1\mathbb{E}_{n+1}-algebra structures on 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty).

One of the main results of formal localization (Section 2.2) tells us that the \infty-category 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) of perfect complexes on X𝑋X is equivalent to the \infty-category of (suitably defined) perfect 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty)-modules (in Xsubscript𝑋\mathcal{M}_{X}). We thus get a 111-parameter deformation of 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) as an 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-monoidal \infty-category, which we call the n𝑛n-quantization of X𝑋X. We also give a version of this result for n<0𝑛0n<0 (where of course 𝔼nsubscript𝔼𝑛\mathbb{E}_{n} will be replaced by 𝔼nsubscript𝔼𝑛\mathbb{E}_{-n}). In contrast, the unshifted n=0𝑛0n=0 case for derived Artin stacks, which was previously addressed for smooth varieties by [Ko1, Ye], is not currently covered by our analysis and seems to require new ideas.

Finally, in Section 3.2, we describe some examples of these n𝑛n-shifted quantizations, especially the quantization on a formal neighborhood of a point, and of various derived moduli stacks of G𝐺G-local systems, for G𝐺G a complex reductive group. Many more examples are awaiting a careful investigation.

Acknowledgements. First of all, we would like to thank our co-authors D. Calaque, B. Toën, and M. Vaquié for the interesting mathematics we did together. We thank V. Melani and M. Porta for their interesting questions that have hopefully led to a clearer text. We are grateful to P. Safronov and N. Rozenblyum for useful exchanges on various topics treated in this review. We are also grateful to the organizers of the 2015 Summer Research Institute on Algebraic Geometry in Salt Lake City, for their invitation to give our talks and to write this review.

Tony Pantev was partially supported by NSF research grant DMS-1302242 and by grant # 347070 from the Simons Foundation. Gabriele Vezzosi is a member of the GNSAGA-INDAM group (Italy) and of PRIN-Geometria delle varietá algebriche (Italy). He would like to point out the serious problems of scientific research in public universities in Italy, due to the substantial lack of acknowledgement from our government of the important cultural and social role of public research in a modern country, and to the subsequent largely insufficient investment of government funds into public research and universities.


Background. We will assume the reader has some familiarity with derived algebraic geometry, for which useful reviews are [To1], and the more recent [To2], while the foundational works are Toën-Vezzosi’s [HAG-II], J. Lurie’s DAG series [Lu3], and also the recent [Lu5], the last two being available at http://www.math.harvard.edu/similar-to\simlurie/ . We will use both the “old” but sometimes still useful language and theory of model categories (see e.g. [Hov, Hir]), and the modern language and theory of \infty-categories ([Lu1, Lu4]).

Notations.

\bullet Throughout this paper k𝑘k will denote a noetherian commutative \mathbb{Q}-algebra.

\bullet We will use (,1)1(\infty,1)-categories [Lu1] as our model for \infty-categories. They will be simply called \infty-categories.

\bullet As a general rule, a model category is written in sans-serif fonts 𝖭𝖭\mathsf{N}, and we denote in bold-face fonts 𝐍:=L(𝖭)assign𝐍𝐿𝖭\mathbf{N}:=L(\mathsf{N}) the \infty-category defined as the homotopy coherent nerve of the Dwyer-Kan localization of fibrant-cofibrant objects in 𝖭𝖭\mathsf{N} along its weak equivalences, with the notable exceptions of the \infty-category of spaces, denoted as 𝐓:=L(𝗌𝖲𝖾𝗍𝗌)assign𝐓𝐿𝗌𝖲𝖾𝗍𝗌\mathbf{T}:=L(\mathsf{sSets}), and of our base \infty-category :=L(𝖬)assign𝐿𝖬\mathcal{M}:=L(\mathsf{M}) (Section 2). The passage from a model category to the associated \infty-category is a localization, and thus very similar to the passage from the category of complexes in an abelian category to the associated derived category. This is a good example to keep in mind.

\bullet All symmetric monoidal categories we use will be symmetric monoidal (bi)closed categories.

\bullet 𝖽𝗀ksubscript𝖽𝗀𝑘\mathsf{dg}_{k} will denote the symmetric monoidal model category of (unbounded) complexes of k𝑘k-modules, with fibrations being degreewise surjective amps, and weak equivalences being quasi-isomorphisms. The associated \infty-category will be denoted by 𝐝𝐠ksubscript𝐝𝐠𝑘\mathbf{dg}_{k}. Note that 𝐝𝐠ksubscript𝐝𝐠𝑘\mathbf{dg}_{k} is then a stable symmetric monoidal \infty-category ([Lu4, Definition 2.0.0.7]).

\bullet 𝐜𝐝𝐠𝐚ksubscript𝐜𝐝𝐠𝐚𝑘\mathbf{cdga}_{k} will denote the \infty-category of non-positively graded differential graded k𝑘k-algebras (with differential increasing the degree). Its objects will be frequently called simply cdga’s. For A𝐜𝐝𝐠𝐚k𝐴subscript𝐜𝐝𝐠𝐚𝑘A\in\mathbf{cdga}_{k}, we will write πiA:=Hi(A)assignsubscript𝜋𝑖𝐴superscript𝐻𝑖𝐴\pi_{i}\,A:=H^{-i}(A) for any i0𝑖0i\geq 0.

\bullet For A𝐜𝐝𝐠𝐚k𝐴subscript𝐜𝐝𝐠𝐚𝑘A\in\mathbf{cdga}_{k}, we will denote either by 𝐋(A)𝐋𝐴\mathbf{L}(A) or 𝐐𝐂𝐨𝐡(A)𝐐𝐂𝐨𝐡𝐴\mathbf{QCoh}(A) the \infty-category of A𝐴A-dg-modules.

\bullet For A𝐜𝐝𝐠𝐚k𝐴subscript𝐜𝐝𝐠𝐚𝑘A\in\mathbf{cdga}_{k}, we will denote by 𝐏𝐞𝐫𝐟(A)𝐏𝐞𝐫𝐟𝐴\mathbf{Perf}(A) the full sub-\infty-category of 𝐐𝐂𝐨𝐡(A)𝐐𝐂𝐨𝐡𝐴\mathbf{QCoh}(A) consisting of perfect A𝐴A-dg-modules.

\bullet If X𝑋X is a derived geometric stack, we will denote by 𝐐𝐂𝐨𝐡(X)𝐐𝐂𝐨𝐡𝑋\mathbf{QCoh}(X) the k𝑘k-linear symmetric monoidal dg-category of quasi-coherent complexes on X𝑋X.

\bullet If X𝑋X is a derived geometric stack, we will denote by 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) the symmetric monoidal sub-dg-category of 𝐐𝐂𝐨𝐡(X)𝐐𝐂𝐨𝐡𝑋\mathbf{QCoh}(X) consisting of dualizable objects, i.e. perfect complexes over X𝑋X.

\bullet If X𝑋X is a derived geometric stack, we will denote by 𝐂𝐨𝐡(X)𝐂𝐨𝐡𝑋\mathbf{Coh}(X) or the full sub-dg category of 𝐐𝐂𝐨𝐡(X)𝐐𝐂𝐨𝐡𝑋\mathbf{QCoh}(X) consisting of complexes whose cohomology sheaves are coherent over the truncation 𝗍0Xsubscript𝗍0𝑋\mathsf{t}_{0}X.

\bullet For a morphism AB𝐴𝐵A\to B of cdga’s, the relative cotangent complex will be denoted 𝕃B/A𝐋(B)subscript𝕃𝐵𝐴𝐋𝐵\mathbb{L}_{B/A}\in\mathbf{L}(B). When A=k𝐴𝑘A=k, we will simply write 𝕃Bsubscript𝕃𝐵\mathbb{L}_{B} instead of 𝕃B/ksubscript𝕃𝐵𝑘\mathbb{L}_{B/k}.

\bullet For derived stacks, we follow the vocabulary of [HAG-II]. In particular derived Artin stacks X𝑋X will be higher derived stacks, unless stated otherwise, and always have a cotangent complex, denoted as 𝕃X𝐐𝐂𝐨𝐡(X)subscript𝕃𝑋𝐐𝐂𝐨𝐡𝑋\mathbb{L}_{X}\in\mathbf{QCoh}(X). The acronym lfp means, as usual, locally finitely presented.

\bullet For a derived stack X𝑋X, Γ(X,)Γ𝑋\Gamma(X,-) will always denote the derived functor of global sections on X𝑋X (i.e. hypercohomology).

1 Shifted symplectic structures

1.1 Definitions

Let ϵ𝖽𝗀kgritalic-ϵsuperscriptsubscript𝖽𝗀𝑘𝑔𝑟\epsilon-\mathsf{dg}_{k}^{gr} be the category of graded mixed complexes of k𝑘k-dg-modules. Its objects are \mathbb{Z}-families of k𝑘k-dg-modules {E(p)}psubscript𝐸𝑝𝑝\{E(p)\}_{p\in\mathbb{Z}}, equipped with dg-module maps ϵ:E(p)E(p+1)[1],:italic-ϵ𝐸𝑝𝐸𝑝1delimited-[]1\epsilon:E(p)\longrightarrow E(p+1)[1], such that ϵ2=0superscriptitalic-ϵ20\epsilon^{2}=0, and the morphisms are \mathbb{Z}-families of morphisms in 𝖽𝗀ksubscript𝖽𝗀𝑘\mathsf{dg}_{k} commuting with ϵitalic-ϵ\epsilon. This is a symmetric monoidal model category: weak equivalences and cofibrations are defined weight-wise (i.e. with respect to the external \mathbb{Z}-grading, that will be called the weight grading), the monoidal structure is defined by (EE)(p):=i+j=pE(i)E(j)assigntensor-product𝐸superscript𝐸𝑝subscriptdirect-sum𝑖𝑗𝑝tensor-product𝐸𝑖superscript𝐸𝑗(E\otimes E^{\prime})(p):=\bigoplus_{i+j=p}E(i)\otimes E^{\prime}(j), and the symmetry constraint does not involve signs, but just swaps the two factors in E(i)E(j)tensor-product𝐸𝑖superscript𝐸𝑗E(i)\otimes E^{\prime}(j). Since our base ring k𝑘k has characteristic zero, the category 𝖢𝗈𝗆𝗆(ϵ𝖽𝗀kgr)=:ϵ𝖼𝖽𝗀𝖺kgr\mathsf{Comm}(\epsilon-\mathsf{dg}_{k}^{gr})=:\epsilon-\mathsf{cdga}_{k}^{gr} of commutative monoid objects in ϵ𝖽𝗀kgritalic-ϵsuperscriptsubscript𝖽𝗀𝑘𝑔𝑟\epsilon-\mathsf{dg}_{k}^{gr} is again model category, with weak equivalences and fibrations inherited via the forgetful functor to ϵ𝖽𝗀kgritalic-ϵsuperscriptsubscript𝖽𝗀𝑘𝑔𝑟\epsilon-\mathsf{dg}_{k}^{gr} (which is then a right Quillen adjoint). According to our general conventions, we will denote by ϵ𝐝𝐠kgritalic-ϵsuperscriptsubscript𝐝𝐠𝑘𝑔𝑟\epsilon-\mathbf{dg}_{k}^{gr} (respectively, ϵ𝐜𝐝𝐠𝐚kgritalic-ϵsuperscriptsubscript𝐜𝐝𝐠𝐚𝑘𝑔𝑟\epsilon-\mathbf{cdga}_{k}^{gr}), the \infty-category associated to ϵ𝖽𝗀kgritalic-ϵsuperscriptsubscript𝖽𝗀𝑘𝑔𝑟\epsilon-\mathsf{dg}_{k}^{gr} (respectively to ϵ𝖼𝖽𝗀𝖺kgritalic-ϵsuperscriptsubscript𝖼𝖽𝗀𝖺𝑘𝑔𝑟\epsilon-\mathsf{cdga}_{k}^{gr}). Informally speaking, ϵ𝐜𝐝𝐠𝐚kgritalic-ϵsuperscriptsubscript𝐜𝐝𝐠𝐚𝑘𝑔𝑟\epsilon-\mathbf{cdga}_{k}^{gr} is therefore the \infty-category of {B(p)𝐝𝐠k}psubscript𝐵𝑝subscript𝐝𝐠𝑘𝑝\{B(p)\in\mathbf{dg}_{k}\}_{p\in\mathbb{Z}} together with mixed differential ϵ:B(p)B(p+1)[1],:italic-ϵ𝐵𝑝𝐵𝑝1delimited-[]1\epsilon:B(p)\to B(p+1)[1], ϵ2=0superscriptitalic-ϵ20\epsilon^{2}=0, and maps B(p)B(q)B(p+q)tensor-product𝐵𝑝𝐵𝑞𝐵𝑝𝑞B(p)\otimes B(q)\to B(p+q) which are unital, associative, commutative, and suitably compatible with ϵitalic-ϵ\epsilon.

The \infty-functor ϵ𝐜𝐝𝐠𝐚kgr𝐜𝐝𝐠𝐚k:{B(p)}B(0):italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘subscript𝐜𝐝𝐠𝐚𝑘maps-to𝐵𝑝𝐵0\epsilon-\mathbf{cdga}^{gr}_{k}\to\mathbf{cdga}_{k}:\{B(p)\}\mapsto B(0) is accessible and preserves limits, thus ([Lu1, Corollary 5.5.2.9]) has a left adjoint 𝐃𝐑:𝐜𝐝𝐠𝐚kϵ𝐜𝐝𝐠𝐚kgr:𝐃𝐑subscript𝐜𝐝𝐠𝐚𝑘italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\mathbf{DR}:\mathbf{cdga}_{k}\to\epsilon-\mathbf{cdga}^{gr}_{k}.

Definition 1.1.1

The functor 𝐃𝐑:𝐜𝐝𝐠𝐚kϵ𝐜𝐝𝐠𝐚kgr:𝐃𝐑subscript𝐜𝐝𝐠𝐚𝑘italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\mathbf{DR}:\mathbf{cdga}_{k}\to\epsilon-\mathbf{cdga}^{gr}_{k} is called the de Rham algebra \infty-functor.

Remark 1.1.2

If A𝐜𝐝𝐠𝐚k𝐴subscript𝐜𝐝𝐠𝐚𝑘A\in\mathbf{cdga}_{k}, we can replace in the previous argument \infty-category 𝐜𝐝𝐠𝐚ksubscript𝐜𝐝𝐠𝐚𝑘\mathbf{cdga}_{k} with A/𝐜𝐝𝐠𝐚k𝐴subscript𝐜𝐝𝐠𝐚𝑘A/\mathbf{cdga}_{k}, and the \infty-category ϵ𝐜𝐝𝐠𝐚kgritalic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\epsilon-\mathbf{cdga}^{gr}_{k} with 𝐃𝐑(A)/ϵ𝐜𝐝𝐠𝐚kgr𝐃𝐑𝐴italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\mathbf{DR}(A)/\epsilon-\mathbf{cdga}^{gr}_{k}, and get a relative de Rham algebra \infty-functor 𝐃𝐑(/A)\mathbf{DR}(-/A).

One can prove that 𝐃𝐑(A)SymA(𝕃A[1])similar-to-or-equals𝐃𝐑𝐴𝑆𝑦subscript𝑚𝐴subscript𝕃𝐴delimited-[]1\mathbf{DR}(A)\simeq Sym_{A}(\mathbb{L}_{A}[-1]) in 𝐜𝐝𝐠𝐚kgrsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\mathbf{cdga}^{gr}_{k} (i.e. as graded cdga’s, by forgetting the mixed differential defined on the rhs). In other words, the construction 𝐃𝐑𝐃𝐑\mathbf{DR} yields the full derived de Rham complex of A𝐴A, including the de Rham differential. We are now able to define shifted closed forms on cdga’s. For m,n𝑚𝑛m,n\in\mathbb{Z}, k(m)[n]𝑘𝑚delimited-[]𝑛k(m)[n] will denote the graded k𝑘k-dg-module sitting in weight degree m𝑚m and in cohomological degree n𝑛-n.

Definition 1.1.3

Let A𝐜𝐝𝐠𝐚k𝐴subscript𝐜𝐝𝐠𝐚𝑘A\in\mathbf{cdga}_{k}

  • The space of closed n𝑛n-shifted p𝑝p-forms on A𝐴A is 𝒜p,cl(A,n):=𝖬𝖺𝗉ϵ𝐝𝐠kgr(k(p)[pn],𝐃𝐑(A))𝕋assignsuperscript𝒜𝑝𝑐𝑙𝐴𝑛subscript𝖬𝖺𝗉italic-ϵsuperscriptsubscript𝐝𝐠𝑘𝑔𝑟𝑘𝑝delimited-[]𝑝𝑛𝐃𝐑𝐴𝕋\mathcal{A}^{p,cl}(A,n):=\mathsf{Map}_{\epsilon-\mathbf{dg}_{k}^{gr}}(k(p)[-p-n],\mathbf{DR}(A))\in\mathbb{T}. An element in π0(𝒜p,cl(A,n))subscript𝜋0superscript𝒜𝑝𝑐𝑙𝐴𝑛\pi_{0}(\mathcal{A}^{p,cl}(A,n)) is called a closed n𝑛n-shifted p𝑝p-form on A𝐴A.

  • The space of n𝑛n-shifted p𝑝p-forms on A𝐴A is 𝒜p(A,n):=𝖬𝖺𝗉𝐝𝐠k(k[n],Ap𝕃A)𝕋assignsuperscript𝒜𝑝𝐴𝑛subscript𝖬𝖺𝗉subscript𝐝𝐠𝑘𝑘delimited-[]𝑛superscriptsubscript𝐴𝑝subscript𝕃𝐴𝕋\mathcal{A}^{p}(A,n):=\mathsf{Map}_{\mathbf{dg}_{k}}(k[-n],\wedge_{A}^{p}\mathbb{L}_{A})\in\mathbb{T}. An element in π0(𝒜p(A,n))subscript𝜋0superscript𝒜𝑝𝐴𝑛\pi_{0}(\mathcal{A}^{p}(A,n)) is called a n𝑛n-shifted p𝑝p-form on A𝐴A.

  • The induced map u:𝒜p,cl(A,n)𝒜p(A,n):𝑢superscript𝒜𝑝𝑐𝑙𝐴𝑛superscript𝒜𝑝𝐴𝑛u:\mathcal{A}^{p,cl}(A,n)\to\mathcal{A}^{p}(A,n) is called the underlying p𝑝p-form map.

Remark 1.1.4

Here is a more concrete description of the space of shifted (closed) forms. If A𝖼𝖽𝗀𝖺k𝐴subscript𝖼𝖽𝗀𝖺𝑘A\in\mathsf{cdga}_{k}, and AAsuperscript𝐴𝐴A^{\prime}\to A is a cofibrant replacement in 𝖼𝖽𝗀𝖺ksubscript𝖼𝖽𝗀𝖺𝑘\mathsf{cdga}_{k}, then p0𝕃A/kp=p0ΩA/kpsubscriptdirect-sum𝑝0superscriptsubscript𝕃𝐴𝑘𝑝subscriptdirect-sum𝑝0superscriptsubscriptΩsuperscript𝐴𝑘𝑝\oplus_{p\geq 0}\mathbb{L}_{A/k}^{p}=\oplus_{p\geq 0}\Omega_{A^{\prime}/k}^{p} is a fourth quadrant bicomplex with vertical differential dvsuperscript𝑑𝑣d^{v} induced by dAsubscript𝑑superscript𝐴d_{A^{\prime}}, and horizontal differential dhsuperscript𝑑d^{h} given by the de Rham differential

dv:ΩA/kp,iΩA/kp,i+1,dh=dDR:ΩA/kp,iΩA/kp+1,i.:superscript𝑑𝑣formulae-sequencesuperscriptsubscriptΩsuperscript𝐴𝑘𝑝𝑖superscriptsubscriptΩsuperscript𝐴𝑘𝑝𝑖1superscript𝑑subscript𝑑𝐷𝑅:superscriptsubscriptΩsuperscript𝐴𝑘𝑝𝑖superscriptsubscriptΩsuperscript𝐴𝑘𝑝1𝑖d^{v}:\Omega_{A^{\prime}/k}^{p,i}\to\Omega_{A^{\prime}/k}^{p,i+1}\,\,,\,\,d^{h}=d_{DR}:\Omega_{A^{\prime}/k}^{p,i}\to\Omega_{A^{\prime}/k}^{p+1,i}.

The Hodge filtration Fsuperscript𝐹F^{\bullet} defined by Fq(A):=pqΩA/kpassignsuperscript𝐹𝑞𝐴subscriptdirect-sum𝑝𝑞superscriptsubscriptΩsuperscript𝐴𝑘𝑝F^{q}(A):=\oplus_{p\geq q}\Omega_{A^{\prime}/k}^{p} is still a fourth quadrant bicomplex, and if we put 𝒜¯p,cl(A,n):=Tot(Fp(A)[n+p]\underline{\mathcal{A}}^{p,cl}(A,n):=\mathrm{Tot}^{\prod}(F^{p}(A)[n+p], we have

𝒜p,cl(A,n)=|𝒜¯p,cl(A,n)|nsuperscript𝒜𝑝𝑐𝑙𝐴𝑛superscript¯𝒜𝑝𝑐𝑙𝐴𝑛𝑛\mathcal{A}^{p,cl}(A,n)=|\underline{\mathcal{A}}^{p,cl}(A,n)|\,\,\,n\in\mathbb{Z}

where |E|𝐸|E| denotes 𝖬𝖺𝗉𝐝𝐠k(k,E)subscript𝖬𝖺𝗉subscript𝐝𝐠𝑘𝑘𝐸\mathsf{Map}_{\mathbf{dg}_{k}}(k,E) i.e. the Dold-Kan construction applied to the 0absent0\leq 0-truncation of the dg-module E𝐸E, and TotsuperscriptTotproduct\mathrm{Tot}^{\prod} is the totalization by products. In particular, we have a corresponding Hodge tower of dg-modules

𝒜¯p,cl(A,0)[p]𝒜¯p1,cl(A,0)[1p]𝒜¯0,cl(A,0),superscript¯𝒜𝑝𝑐𝑙𝐴0delimited-[]𝑝superscript¯𝒜𝑝1𝑐𝑙𝐴0delimited-[]1𝑝superscript¯𝒜0𝑐𝑙𝐴0...\to\underline{\mathcal{A}}^{p,cl}(A,0)[-p]\to\underline{\mathcal{A}}^{p-1,cl}(A,0)[1-p]\to...\to\underline{\mathcal{A}}^{0,cl}(A,0),

where, for any p𝑝p, the cofiber of 𝒜¯p,cl(A,0)[p]𝒜¯p1,cl(A,0)[1p]superscript¯𝒜𝑝𝑐𝑙𝐴0delimited-[]𝑝superscript¯𝒜𝑝1𝑐𝑙𝐴0delimited-[]1𝑝\underline{\mathcal{A}}^{p,cl}(A,0)[-p]\to\underline{\mathcal{A}}^{p-1,cl}(A,0)[1-p] is equivalent to the dg-module 𝒜¯p1(A,0)[1p]:=(Ap1𝕃A)[1p]assignsuperscript¯𝒜𝑝1𝐴0delimited-[]1𝑝superscriptsubscript𝐴𝑝1subscript𝕃𝐴delimited-[]1𝑝\underline{\mathcal{A}}^{p-1}(A,0)[1-p]:=(\wedge_{A}^{p-1}\mathbb{L}_{A})[1-p] of (1p)1𝑝(1-p)-shifted (p1)𝑝1(p-1)-forms on A𝐴A (so that we have an equivalence |𝒜¯p1(A,0)[1p]|𝒜p1(A,1p)similar-to-or-equalssuperscript¯𝒜𝑝1𝐴0delimited-[]1𝑝superscript𝒜𝑝1𝐴1𝑝|\underline{\mathcal{A}}^{p-1}(A,0)[1-p]|\simeq\mathcal{A}^{p-1}(A,1-p) in 𝐓𝐓\mathbf{T}). Finally, let us observe that the rightmost dg-module 𝒜¯0,cl(A,0)superscript¯𝒜0𝑐𝑙𝐴0\underline{\mathcal{A}}^{0,cl}(A,0) in the above Hodge tower, is exactly Illusie’s derived de Rham complex of A𝐴A ([Ill, ch. VIII]).

Remark 1.1.5

Note that the de Rham algebra functor, and hence the notion of (closed) shifted forms, makes sense when 𝐝𝐠ksubscript𝐝𝐠𝑘\mathbf{dg}_{k} is replaced by (essentially) any symmetric monoidal stable k𝑘k-linear \infty-category \mathcal{M}. The intermediate categories of interest will then be ϵgritalic-ϵsuperscript𝑔𝑟\epsilon-\mathcal{M}^{gr} (generalizing ϵ𝐝𝐠kgritalic-ϵsuperscriptsubscript𝐝𝐠𝑘𝑔𝑟\epsilon-\mathbf{dg}_{k}^{gr}), and ϵ𝐂𝐀𝐥𝐠gritalic-ϵsuperscriptsubscript𝐂𝐀𝐥𝐠𝑔𝑟\epsilon-\mathbf{CAlg}_{\mathcal{M}}^{gr} (generalizing ϵ𝖼𝖽𝗀𝖺kgritalic-ϵsuperscriptsubscript𝖼𝖽𝗀𝖺𝑘𝑔𝑟\epsilon-\mathsf{cdga}_{k}^{gr}). For any A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}}, this will yield a cotangent complex 𝕃AA𝐌𝐨𝐝subscriptsuperscript𝕃𝐴𝐴subscript𝐌𝐨𝐝\mathbb{L}^{\mathcal{M}}_{A}\in A-\mathbf{Mod}_{\mathcal{M}}, a de Rham algebra functor 𝐃𝐑:𝐂𝐀𝐥𝐠ϵ𝐂𝐀𝐥𝐠gr:superscript𝐃𝐑subscript𝐂𝐀𝐥𝐠italic-ϵsuperscriptsubscript𝐂𝐀𝐥𝐠𝑔𝑟\mathbf{DR}^{\mathcal{M}}:\mathbf{CAlg}_{\mathcal{M}}\to\epsilon-\mathbf{CAlg}_{\mathcal{M}}^{gr}, and a space of n𝑛n-shifted (closed) p𝑝p-form 𝒜p(A,n)superscriptsubscript𝒜𝑝𝐴𝑛\mathcal{A}_{\mathcal{M}}^{p}(A,n) (𝒜p,cl(A,n)superscriptsubscript𝒜𝑝𝑐𝑙𝐴𝑛\mathcal{A}_{\mathcal{M}}^{p,cl}(A,n)), where the sub/superscript {\mathcal{M}} indicates that all the constructions are performed internally to \mathcal{M}. This level of generality and flexibility in the choice of the context for our differential calculus, will prove extremely useful in the rest of the paper. As relevant cases, the reader should keep in mind the case where =ϵ𝖽𝗀kgritalic-ϵsuperscriptsubscript𝖽𝗀𝑘𝑔𝑟\mathcal{M}=\epsilon-\mathsf{dg}_{k}^{gr} or, more generally, diagrams in ϵ𝖽𝗀kgritalic-ϵsuperscriptsubscript𝖽𝗀𝑘𝑔𝑟\epsilon-\mathsf{dg}_{k}^{gr}. We will come back to this generalization more systematically in Section 2.1, and use it as an essential tool starting from Section 2.3.

We are now ready to globalize the above construction to derived stacks. We start by globalizing the de Rham algebra construction (Definition 1.1.1). The functor A𝐃𝐑(A)𝐴𝐃𝐑𝐴A\to\mathbf{DR}(A), and its relative version (over a fixed base B𝐵B, see Remark 1.1.2), are both derived stacks (for the étale topology) with values in mixed graded dga’s, so we give the following

Definition 1.1.6

(1) Let F𝐒𝐩𝐞𝐜B𝐹𝐒𝐩𝐞𝐜𝐵F\to\mathbf{Spec}\,B be a map in 𝐝𝐒𝐭ksubscript𝐝𝐒𝐭𝑘\mathbf{dSt}_{k}. The relative de Rham algebra of F𝐹F over B𝐵B

𝐃𝐑(F/B):=lim𝐒𝐩𝐞𝐜CF𝐃𝐑(C/B)ϵ𝐜𝐝𝐠𝐚Bgrassign𝐃𝐑𝐹𝐵subscript𝐒𝐩𝐞𝐜𝐶𝐹𝐃𝐑𝐶𝐵italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝐵\mathbf{DR}(F/B):=\lim_{\mathbf{Spec}\,C\to F}\mathbf{DR}(C/B)\in\epsilon-\mathbf{cdga}^{gr}_{B}

(see Remark 1.1.2) where the limit is taken in the \infty-category ϵ𝐜𝐝𝐠𝐚Bgr=B/ϵ𝐜𝐝𝐠𝐚kgritalic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝐵𝐵italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\epsilon-\mathbf{cdga}^{gr}_{B}=B/\epsilon-\mathbf{cdga}^{gr}_{k} of graded mixed B𝐵B-linear cdgas, and over all morphisms 𝐒𝐩𝐞𝐜CF𝐒𝐩𝐞𝐜𝐶𝐹\mathbf{Spec}\,C\rightarrow F of derived stacks over 𝐒𝐩𝐞𝐜B𝐒𝐩𝐞𝐜𝐵\mathbf{Spec}\,B.

(2) For an arbitrary map FG𝐹𝐺F\to G in 𝐝𝐒𝐭ksubscript𝐝𝐒𝐭𝑘\mathbf{dSt}_{k}, we define the relative de Rham algebra of F𝐹F over G𝐺G as

𝐃𝐑(F/G):=lim𝐒𝐩𝐞𝐜AG𝐃𝐑(FA/A)ϵ𝐜𝐝𝐠𝐚kgr,assign𝐃𝐑𝐹𝐺subscript𝐒𝐩𝐞𝐜𝐴𝐺𝐃𝐑subscript𝐹𝐴𝐴italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\mathbf{DR}(F/G):=\lim_{\mathbf{Spec}\,A\to G}\mathbf{DR}(F_{A}/A)\in\epsilon-\mathbf{cdga}^{gr}_{k}\,,

where FAsubscript𝐹𝐴F_{A} denotes the base change of FG𝐹𝐺F\to G along 𝐒𝐩𝐞𝐜AG𝐒𝐩𝐞𝐜𝐴𝐺\mathbf{Spec}\,A\to G, and the limit is taken in the \infty-category ϵ𝐜𝐝𝐠𝐚kgritalic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\epsilon-\mathbf{cdga}^{gr}_{k}.

We now globalize to derived stacks the notion of (closed) shifted forms. From Definition 1.1.3 one deduces \infty-functors 𝒜p,cl(,n):A𝒜p,cl(A,n):superscript𝒜𝑝𝑐𝑙𝑛maps-to𝐴superscript𝒜𝑝𝑐𝑙𝐴𝑛\mathcal{A}^{p,cl}(-,n):A\mapsto\mathcal{A}^{p,cl}(A,n), and 𝒜p(,n)):A𝒜p(A,n)\mathcal{A}^{p}(-,n)):A\mapsto\mathcal{A}^{p}(A,n) from 𝐜𝐝𝐠𝐚ksubscript𝐜𝐝𝐠𝐚𝑘\mathbf{cdga}_{k} to 𝐓𝐓\mathbf{T}. By [PTVV, Proposition 1.11], these functors are derived stacks (for the étale topology). This allows us to globalize Definition 1.1.3 on an arbitrary derived stack.

Definition 1.1.7

Let F𝐹F be a derived stack.

  • The space of closed n𝑛n-shifted p𝑝p-forms on F𝐹F is 𝒜p,cl(F,n):=𝖬𝖺𝗉𝐝𝐒𝐭k(F,𝒜p,cl(,n))assignsuperscript𝒜𝑝𝑐𝑙𝐹𝑛subscript𝖬𝖺𝗉subscript𝐝𝐒𝐭𝑘𝐹superscript𝒜𝑝𝑐𝑙𝑛\mathcal{A}^{p,cl}(F,n):=\mathsf{Map}_{\mathbf{dSt}_{k}}(F,\mathcal{A}^{p,cl}(-,n)).

  • The space of n𝑛n-shifted p𝑝p-forms on F𝐹F is 𝒜p(F,n):=𝖬𝖺𝗉𝐝𝐒𝐭k(F,𝒜p(,n))assignsuperscript𝒜𝑝𝐹𝑛subscript𝖬𝖺𝗉subscript𝐝𝐒𝐭𝑘𝐹superscript𝒜𝑝𝑛\mathcal{A}^{p}(F,n):=\mathsf{Map}_{\mathbf{dSt}_{k}}(F,\mathcal{A}^{p}(-,n)).

  • The resulting induced map u:𝒜p,cl(F,n)𝒜p(F,n):𝑢superscript𝒜𝑝𝑐𝑙𝐹𝑛superscript𝒜𝑝𝐹𝑛u:\mathcal{A}^{p,cl}(F,n)\to\mathcal{A}^{p}(F,n) is called the underlying p𝑝p-form map.

Note that, in general, the homotopy fiber of the underlying p𝑝p-form map u𝑢u can be non-trivial (i.e. not empty nor contractible). Hence being closed is a datum rather than a property, for a general derived stack.

Remark 1.1.8

(1) Equivalently, we have 𝒜p,cl(F,n)lim𝐒𝐩𝐞𝐜A(𝐝𝐀𝐟𝐟/F)op𝒜p,cl(A,n)similar-to-or-equalssuperscript𝒜𝑝𝑐𝑙𝐹𝑛subscript𝐒𝐩𝐞𝐜𝐴superscript𝐝𝐀𝐟𝐟𝐹𝑜𝑝superscript𝒜𝑝𝑐𝑙𝐴𝑛\mathcal{A}^{p,cl}(F,n)\simeq\lim_{\mathbf{Spec}\,A\in(\mathbf{dAff}/F)^{op}}\mathcal{A}^{p,cl}(A,n), and 𝒜p(F,n)lim𝐒𝐩𝐞𝐜A(𝐝𝐀𝐟𝐟/F)op𝒜p,cl(A,n)similar-to-or-equalssuperscript𝒜𝑝𝐹𝑛subscript𝐒𝐩𝐞𝐜𝐴superscript𝐝𝐀𝐟𝐟𝐹𝑜𝑝superscript𝒜𝑝𝑐𝑙𝐴𝑛\mathcal{A}^{p}(F,n)\linebreak\simeq\lim_{\mathbf{Spec}\,A\in(\mathbf{dAff}/F)^{op}}\mathcal{A}^{p,cl}(A,n), where the limits are taken in the \infty-category of \infty-functors from (𝐝𝐀𝐟𝐟/F)opsuperscript𝐝𝐀𝐟𝐟𝐹𝑜𝑝(\mathbf{dAff}/F)^{op} to 𝐓𝐓\mathbf{T}.

(2) Also note that we have an equivalence 𝒜p,cl(F,n)𝖬𝖺𝗉ϵ𝐝𝐠kgr(k(p)[pn],𝐃𝐑(F/k))similar-to-or-equalssuperscript𝒜𝑝𝑐𝑙𝐹𝑛subscript𝖬𝖺𝗉italic-ϵsuperscriptsubscript𝐝𝐠𝑘𝑔𝑟𝑘𝑝delimited-[]𝑝𝑛𝐃𝐑𝐹𝑘\mathcal{A}^{p,cl}(F,n)\simeq\mathsf{Map}_{\epsilon-\mathbf{dg}_{k}^{gr}}(k(p)[-p-n],\mathbf{DR}(F/k)) in 𝐓𝐓\mathbf{T}.

(3) For F=𝐒𝐩𝐞𝐜A𝐹𝐒𝐩𝐞𝐜𝐴F=\mathbf{Spec}\,A, the complex p𝕃Asuperscript𝑝subscript𝕃𝐴\wedge^{p}\mathbb{L}_{A} has non-positive tor-amplitide, hence there are no non-trivial n𝑛n-shifted p𝑝p-forms on F𝐹F, for n>0𝑛0n>0. For F=X𝐹𝑋F=X an underived smooth scheme, a similar argument shows that F𝐹F admits no non-trivial n𝑛n-shifted p𝑝p-forms on F𝐹F, for n<0𝑛0n<0. If moreover X𝑋X is proper over k𝑘k, then the degeneration of the Hodge-to-de Rham spectral sequence implies that the underlying p𝑝p-form map is injective on π0subscript𝜋0\pi_{0}.

For derived (higher) Artin stacks F𝐹F, the space 𝒜p(F,n)superscript𝒜𝑝𝐹𝑛\mathcal{A}^{p}(F,n) has the following equivalent description (smooth descent).

Proposition 1.1.9

[PTVV, Proposition 1.14]  If F𝐝𝐒𝐭𝐹𝐝𝐒𝐭F\in\mathbf{dSt} is Artin, then we have an equivalence

𝒜p(F,n)𝖬𝖺𝗉𝐐𝐂𝐨𝐡(F)(𝒪F,p𝕃F[n]),similar-to-or-equalssuperscript𝒜𝑝𝐹𝑛subscript𝖬𝖺𝗉𝐐𝐂𝐨𝐡𝐹subscript𝒪𝐹superscript𝑝subscript𝕃𝐹delimited-[]𝑛\mathcal{A}^{p}(F,n)\simeq\mathsf{Map}_{\mathbf{QCoh}(F)}(\mathcal{O}_{F},\wedge^{p}\mathbb{L}_{F}[n])\,,

functorial in F𝐹F.

In particular π0(𝒜p(F,n))n(X,p𝕃F)similar-to-or-equalssubscript𝜋0superscript𝒜𝑝𝐹𝑛superscript𝑛𝑋superscript𝑝subscript𝕃𝐹\pi_{0}(\mathcal{A}^{p}(F,n))\simeq\mathbb{H}^{n}(X,\wedge^{p}\mathbb{L}_{F}), for F𝐹F Artin. Thus an n𝑛n-shifted 222-form ω𝜔\omega can be identified with a map ω:𝒪F2𝕃F[n]:𝜔subscript𝒪𝐹superscript2subscript𝕃𝐹delimited-[]𝑛\omega:\mathcal{O}_{F}\to\wedge^{2}\mathbb{L}_{F}[n]. If F𝐹F is moreover locally of finite presentation over k𝑘k (so that its cotangent complex 𝕃Fsubscript𝕃𝐹\mathbb{L}_{F} is perfect, i.e. dualizable in 𝐐𝐂𝐨𝐡(F)𝐐𝐂𝐨𝐡𝐹\mathbf{QCoh}(F)), we may associate to such an ω𝜔\omega an adjoint map ω:𝕋F𝕃F[n]:superscript𝜔subscript𝕋𝐹subscript𝕃𝐹delimited-[]𝑛\omega^{\flat}:\mathbb{T}_{F}\to\mathbb{L}_{F}[n], where 𝕋Fsubscript𝕋𝐹\mathbb{T}_{F} denotes the dual of 𝕃Fsubscript𝕃𝐹\mathbb{L}_{F} in 𝐐𝐂𝐨𝐡(F)𝐐𝐂𝐨𝐡𝐹\mathbf{QCoh}(F), and it is called the tangent complex of F𝐹F. An n𝑛n-shifted 222 form ω𝜔\omega on such an F𝐹F is said to be non-degenerate if the map ωsuperscript𝜔\omega^{\flat} is an equivalence, i.e. an isomorphism in the derived category of quasi-coherent complexes on F𝐹F, and we will denote by 𝒜nd2(F,n)superscriptsubscript𝒜𝑛𝑑2𝐹𝑛\mathcal{A}_{nd}^{2}(F,n) the subspace of 𝒜2(F,n)superscript𝒜2𝐹𝑛\mathcal{A}^{2}(F,n) consisting of connected components of non-degenerate forms.

Definition 1.1.10

Let F𝐹F be an derived Artin stack locally of finite presentation over k𝑘k. The space of n𝑛n-shifted symplectic structures on F𝐹F is defined by the following pullback diagram in 𝕋𝕋\mathbb{T}

𝖲𝗒𝗆𝗉𝗅(F,n)𝒜2,cl(F,n)𝒜nd2(F,n)𝒜2(F,n),𝖲𝗒𝗆𝗉𝗅𝐹𝑛superscript𝒜2𝑐𝑙𝐹𝑛superscriptsubscript𝒜𝑛𝑑2𝐹𝑛superscript𝒜2𝐹𝑛\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 29.21649pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern 0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&\\&\crcr}}}\ignorespaces{\hbox{\kern-29.21649pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathsf{Sympl}(F,n)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 53.21649pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 0.0pt\raise-29.56999pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 53.21649pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathcal{A}^{2,cl}(F,n)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 76.713pt\raise-29.56999pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern-22.35448pt\raise-41.20775pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathcal{A}_{nd}^{2}(F,n)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces{\hbox{\kern 56.09654pt\raise-41.20775pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 56.09654pt\raise-41.20775pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathcal{A}^{2}(F,n)}$}}}}}}}\ignorespaces}}}}\ignorespaces,

and an element in π0(𝖲𝗒𝗆𝗉𝗅(F;n))subscript𝜋0𝖲𝗒𝗆𝗉𝗅𝐹𝑛\pi_{0}(\mathsf{Sympl}(F;n)) is called a n𝑛n-shifted symplectic structure on F𝐹F.

In other words, an n𝑛n-shifted symplectic structure ω𝜔\omega on F𝐹F is a closed n𝑛n-shifted 2-form on F𝐹F whose underlying 2-form is non-degenerate; in particular, 𝐋Fsubscript𝐋𝐹\mathbf{L}_{F} is self dual, up to a shift. We use the word symplectic structure instead of symplectic form because a shifted symplectic structure is a closed- 2-form, and with respect to the underlying form, this consists of additional structure rather than just being a property.

The non-degeneracy condition entails a mixture of the (higher) stacky (i.e. positive degrees in the cotangent complex) and derived (i.e. negative degrees of the cotangent complex) nature of the stack F𝐹F, and in particular it poses severe restrictions on the existence of shifted symplectic structures on a given stack. E.g. it is clear that if 𝕃Fsubscript𝕃𝐹\mathbb{L}_{F} has perfect amplitude in [a,b]𝑎𝑏[a,b], then F𝐹F may only support (a+b)𝑎𝑏(a+b)-shifted symplectic structures. More precisely, it is easy to check that for a smooth underived scheme X𝑋X, not étale over k𝑘k, the space 𝖲𝗒𝗆𝗉𝗅(X,n)𝖲𝗒𝗆𝗉𝗅𝑋𝑛\mathsf{Sympl}(X,n) is empty for n0𝑛0n\neq 0, and either empty or contractible for n=0𝑛0n=0, and moreover, the set of connected components of 𝖲𝗒𝗆𝗉𝗅(X,0)𝖲𝗒𝗆𝗉𝗅𝑋0\mathsf{Sympl}(X,0) is in canonical bijection with the set of usual algebraic symplectic forms on X𝑋X ([PTVV, p. 298]). So, we get nothing new for smooth underived schemes, or more generally, smooth Deligne-Mumford stacks. However, we will see in the following subsection that there are plenty of derived schemes or stacks carrying interesting shifted symplectic forms.

1.2 Existence theorems and examples

We will now review the three basic existence theorems (Theorems 1.2.1, 1.2.5, 1.2.8, below) for shifted symplectic structures established in [PTVV]. In combination they give a long list of non-trivial examples of shifted symplectic stacks.

The first interesting example of a shifted symplectic stack is the classifying stack 𝖡G𝖡𝐺\mathsf{B}G of a smooth affine reductive group scheme over k𝑘k. In this case ,we have ([PTVV, p. 299])

π0(𝖲𝗒𝗆𝗉𝗅(𝖡G,n))={0n2Symk(𝔤)ndGn=2subscript𝜋0𝖲𝗒𝗆𝗉𝗅𝖡𝐺𝑛cases0𝑛2𝑆𝑦subscript𝑚𝑘superscriptsubscriptsuperscript𝔤𝑛𝑑𝐺𝑛2\pi_{0}(\mathsf{Sympl}(\mathsf{B}G,n))=\begin{cases}0&n\neq 2\\ Sym_{k}(\mathfrak{g}^{\vee})_{nd}^{G}&n=2\end{cases}

where 𝔤𝔤\mathfrak{g} is the Lie algebra of G𝐺G, and Symk(𝔤)ndG𝑆𝑦subscript𝑚𝑘superscriptsubscriptsuperscript𝔤𝑛𝑑𝐺Sym_{k}(\mathfrak{g}^{\vee})_{nd}^{G} is the set of non-degenerate G𝐺G-invariant symmetric bilinear forms on 𝔤𝔤\mathfrak{g}. At the level of tangent complex 𝕋𝖡G,e𝔤[1]similar-to-or-equalssubscript𝕋𝖡𝐺𝑒𝔤delimited-[]1\mathbb{T}_{\mathsf{B}G,e}\simeq\mathfrak{g}[1], the underlying 2-form corresponding to a non-degenerate G𝐺G-invariant symmetric bilinear form σ:Symk2(𝔤)k:𝜎𝑆𝑦superscriptsubscript𝑚𝑘2𝔤𝑘\sigma:Sym_{k}^{2}(\mathfrak{g})\to k is given by the composite

𝕋𝖡G,e𝕋𝖡G,esubscript𝕋𝖡𝐺𝑒subscript𝕋𝖡𝐺𝑒\textstyle{\mathbb{T}_{\mathsf{B}G,e}\wedge\mathbb{T}_{\mathsf{B}G,e}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}similar-to\scriptstyle{\sim}𝔤[1]𝔤[1]𝔤delimited-[]1𝔤delimited-[]1\textstyle{\mathfrak{g}[1]\wedge\mathfrak{g}[1]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}similar-to\scriptstyle{\sim}Symk2(𝔤)[2]𝑆𝑦superscriptsubscript𝑚𝑘2𝔤delimited-[]2\textstyle{Sym_{k}^{2}(\mathfrak{g})[2]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}σ[2]𝜎delimited-[]2\scriptstyle{\sigma[2]}k[2],𝑘delimited-[]2\textstyle{k[2],}

where the central quasi-isomorphism is given by décalage. For example, if G=GLn𝐺subscriptGL𝑛G=\mathrm{GL}_{n}, the usual map (A,B)tr(AB),maps-to𝐴𝐵tr𝐴𝐵(A,B)\mapsto\mathrm{tr}(AB), for A,B𝐴𝐵A,B (n×n)𝑛𝑛(n\times n) matrices over k𝑘k, yields a 222-shifted symplectic form on 𝖡GLn𝖡subscriptGL𝑛\mathsf{B}\mathrm{GL}_{n}. This example will be vastly generalized in Theorem 1.2.8 below.

As a second example, for any n𝑛n\in\mathbb{Z}, one has that the n𝑛n-shifted cotangent stack TF[n]:=assignsuperscriptT𝐹delimited-[]𝑛absent\mathrm{T}^{*}F[n]:= 𝐒𝐩𝐞𝐜F(Sym𝒪F(𝕋F[n]))subscript𝐒𝐩𝐞𝐜𝐹𝑆𝑦subscript𝑚subscript𝒪𝐹subscript𝕋𝐹delimited-[]𝑛\mathbf{Spec}_{F}(Sym_{\mathcal{O}_{F}}(\mathbb{T}_{F}[-n])) of a derived Deligne-Mumford stack F𝐹F lfp over k𝑘k, is canonically n𝑛n-shifted symplectic via the de Rham differential of the canonical shifted Liouville form ([PTVV, Proposition 1.21]). Recently, D. Calaque has extended this result to derived Artin stacks lfp over k𝑘k [Cal2].

The first general existence result for shifted symplectic form is an enhanced derived version of the main result in [AKSZ].

Theorem 1.2.1

Let F𝐹F be a derived Artin stack lfp over k𝑘k, equipped with a n𝑛n-shifted symplectic form, and let X𝑋X be an 𝒪𝒪\mathcal{O}-compact derived stack equipped with a d𝑑d-orientation [X]:(X,𝒪X)k[d]:delimited-[]𝑋𝑋subscript𝒪𝑋𝑘delimited-[]𝑑[X]:\mathbb{H}(X,\mathcal{O}_{X})\to k[-d]. If 𝖬𝖠𝖯𝐝𝐒𝐭(X,F)subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝐹\mathsf{MAP}_{\mathbf{dSt}}(X,F) is a derived Artin stack lfp over k𝑘k, then it carries a canonical (nd)𝑛𝑑(n-d)-shifted symplectic structure.

We direct the reader to [PTVV, 2.1] for the definition of 𝒪𝒪\mathcal{O}-compact derived stack, and for the notion of d𝑑d-orientation on a 𝒪𝒪\mathcal{O}-compact derived stack (i.e. for the special properties of the map [X]delimited-[]𝑋[X] in the theorem), and to [PTVV, Theorem 2.5] for a detailed proof. Here we will content ourselves with a few comments.

First of all the class 𝒪𝒪\mathcal{O}-compact derived stacks equipped with a d𝑑d-orientation includes compact smooth oriented topological manifolds M𝑀M of dimension d𝑑d (identified with constant stacks with value M𝑀M, and where capping with the fundamental class gives the d𝑑d-orientation), Calabi-Yau varieties of complex dimension d𝑑d (where the orientation is given by a trivialization of the canonical sheaf followed by the trace map), and De Rham stacks X=YDR𝑋subscript𝑌𝐷𝑅X=Y_{DR} for Y𝑌Y a smooth and proper Deligne-Mumford stack with connected geometric fibers, and relative dimension d/2𝑑2d/2 over 𝐒𝐩𝐞𝐜k𝐒𝐩𝐞𝐜𝑘\mathbf{Spec}\,k 222The notion of de Rham stack will be defined and discussed in Section 2.2 (where the d𝑑d-orientation is induced by the choice of a fundamental class in de Rham cohomology dR(Y/k,𝒪)subscript𝑑𝑅𝑌𝑘𝒪\mathbb{H}_{dR}(Y/k,\mathcal{O})).

As a second comment we outline the proof of Thm 1.2.1. One first uses the evaluation map 𝖾𝗏:X×𝖬𝖠𝖯𝐝𝐒𝐭(X,F)F:𝖾𝗏𝑋subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝐹𝐹\mathsf{ev}:X\times\mathsf{MAP}_{\mathbf{dSt}}(X,F)\to F in order to pull back the n𝑛n-shifted symplectic structure ω𝜔\omega on F𝐹F, to a closed form on X×𝖬𝖠𝖯𝐝𝐒𝐭(X,F)𝑋subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝐹X\times\mathsf{MAP}_{\mathbf{dSt}}(X,F); this pullback is then “integrated along the fiber” of the projection X×𝖬𝖠𝖯𝐝𝐒𝐭(X,F)𝖬𝖠𝖯𝐝𝐒𝐭(X,F)𝑋subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝐹subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝐹X\times\mathsf{MAP}_{\mathbf{dSt}}(X,F)\to\mathsf{MAP}_{\mathbf{dSt}}(X,F), and this integrated form is shown to be (nd)𝑛𝑑(n-d)-shifted symplectic. The possibility of defining an integration along the fiber X𝑋X is a consequence of the definition of d𝑑d-orientation on an 𝒪𝒪\mathcal{O}-compact stack ([PTVV, Definition 2.3]). Finally, we observe that the general question of finding optimal conditions on X𝑋X and F𝐹F ensuring that 𝖬𝖠𝖯𝐝𝐒𝐭(X,F)subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝐹\mathsf{MAP}_{\mathbf{dSt}}(X,F) is a derived Artin stack lfp over k𝑘k is delicate, the Artin-Lurie representability criterion (even in the simplified version of [HAG-II, Appendix]) will give an answer in all the applications we will discuss below.

We come to the second existence theorem for shifted symplectic structures. Before stating it, we need to define the notion of lagrangian structure on a map whose target is a shifted symplectic stack. We start by defining what is an isotropic structure on such a map.

Definition 1.2.2

Let X𝑋X and F𝐹F be derived Artin stacks lfp over k𝑘k, ω𝜔\omega a n𝑛n-shifted symplectic structure on F𝐹F, and f:XF:𝑓𝑋𝐹f:X\to F a map. The space 𝖨𝗌𝗈𝗍𝗋(f;ω)𝖨𝗌𝗈𝗍𝗋𝑓𝜔\mathsf{Isotr}(f;\omega) of isotropic structures on the map f𝑓f relative to ω𝜔\omega is the space of paths Ωfω,0𝒜2,cl(X,n)subscriptΩsuperscript𝑓𝜔0superscript𝒜2𝑐𝑙𝑋𝑛\Omega_{f^{*}\omega,0}\mathcal{A}^{2,cl}(X,n) between fωsuperscript𝑓𝜔f^{*}\omega and 00 in the space of n𝑛n-shifted closed p𝑝p-forms on X𝑋X. An element in γπ0(𝖨𝗌𝗈𝗍𝗋(f;ω))𝛾subscript𝜋0𝖨𝗌𝗈𝗍𝗋𝑓𝜔\gamma\in\pi_{0}(\mathsf{Isotr}(f;\omega)), i.e. a path between fωsuperscript𝑓𝜔f^{*}\omega and 00, is called an isotropic structure on f𝑓f relative to ω𝜔\omega.

The idea, ubiquitous in all of derived geometry and higher category theory, and that we already saw in action in the definition of closed forms (Def 1.1.3), is that it is not enough to say that there exists a path between fωsuperscript𝑓𝜔f^{*}\omega and 00 in 𝒜2,cl(X,n)superscript𝒜2𝑐𝑙𝑋𝑛\mathcal{A}^{2,cl}(X,n) (this would be a property), but one rather has to specify one such path (i.e. a datum).

Once an isotropic structure γ𝛾\gamma is given, it makes sense to say that it has the property of being non-degenerate, as follows. By composition with the underlying 2-form map u:𝒜2,cl(X,n)𝒜2(X,n):𝑢superscript𝒜2𝑐𝑙𝑋𝑛superscript𝒜2𝑋𝑛u:\mathcal{A}^{2,cl}(X,n)\to\mathcal{A}^{2}(X,n), the path γ𝛾\gamma induces a path γusubscript𝛾𝑢\gamma_{u} between u(fω)=f(u(ω))𝑢superscript𝑓𝜔superscript𝑓𝑢𝜔u(f^{*}\omega)=f^{*}(u(\omega)) and 00 in 𝒜2(X,n)superscript𝒜2𝑋𝑛\mathcal{A}^{2}(X,n). Since 𝕃Xsubscript𝕃𝑋\mathbb{L}_{X} is perfect, by adjunction, this yields in turn a path γusuperscriptsubscript𝛾𝑢\gamma_{u}^{\flat} between (f(u(ω)))superscriptsuperscript𝑓𝑢𝜔(f^{*}(u(\omega)))^{\flat} and 00 in 𝖬𝖺𝗉𝐏𝐞𝐫𝐟(X)(𝕋X,𝕃X[n])subscript𝖬𝖺𝗉𝐏𝐞𝐫𝐟𝑋subscript𝕋𝑋subscript𝕃𝑋delimited-[]𝑛\mathsf{Map}_{\mathbf{Perf}(X)}(\mathbb{T}_{X},\mathbb{L}_{X}[n]), where (f(u(ω)))superscriptsuperscript𝑓𝑢𝜔(f^{*}(u(\omega)))^{\flat} is the composite

𝕋Xaf𝕋Ff(u(ω))f𝕃F[n]a[n]𝕃X[n],subscript𝕋𝑋superscript𝑎superscript𝑓subscript𝕋𝐹superscript𝑓𝑢superscript𝜔superscript𝑓subscript𝕃𝐹delimited-[]𝑛𝑎delimited-[]𝑛subscript𝕃𝑋delimited-[]𝑛\lx@xy@svg{\hbox{\raise 0.0pt\hbox{\kern 9.15056pt\hbox{\ignorespaces\ignorespaces\ignorespaces\hbox{\vtop{\kern 0.0pt\offinterlineskip\halign{\entry@#!@&&\entry@@#!@\cr&&&&\crcr}}}\ignorespaces{\hbox{\kern-9.15056pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathbb{T}_{X}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 14.96716pt\raise 5.0625pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-2.0625pt\hbox{$\scriptstyle{a^{\vee}}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 33.15056pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 33.15056pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{f^{*}\mathbb{T}_{F}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 70.02924pt\raise 7.25pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-2.5pt\hbox{$\scriptstyle{f^{*}(u(\omega)^{\flat})}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 113.52394pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 83.52394pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{}$}}}}}}}{\hbox{\kern 113.52394pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{f^{*}\mathbb{L}_{F}[n]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}$}}}}}}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces{}{\hbox{\lx@xy@droprule}}\ignorespaces\ignorespaces\ignorespaces{\hbox{\kern 151.56963pt\raise 6.5pt\hbox{{}\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\hbox{\hbox{\kern 0.0pt\raise-1.75pt\hbox{$\scriptstyle{a[n]}$}}}\kern 3.0pt}}}}}}\ignorespaces{\hbox{\kern 174.48302pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\lx@xy@tip{1}\lx@xy@tip{-1}}}}}}{\hbox{\lx@xy@droprule}}{\hbox{\lx@xy@droprule}}{\hbox{\kern 174.48302pt\raise 0.0pt\hbox{\hbox{\kern 0.0pt\raise 0.0pt\hbox{\hbox{\kern 3.0pt\raise 0.0pt\hbox{$\textstyle{\mathbb{L}_{X}[n]}$}}}}}}}\ignorespaces}}}}\ignorespaces,

a:f𝕃F𝕃X:𝑎superscript𝑓subscript𝕃𝐹subscript𝕃𝑋a:f^{*}\mathbb{L}_{F}\to\mathbb{L}_{X} being the canonical map induced by f𝑓f. If we denote by tf,ωsubscript𝑡𝑓𝜔t_{f,\omega} the composite a[n]f(u(ω))𝑎delimited-[]𝑛superscript𝑓𝑢superscript𝜔a[n]\circ f^{*}(u(\omega)^{\flat}), we thus obtain that γusuperscriptsubscript𝛾𝑢\gamma_{u}^{\flat} is a homotopy commutativity datum for the square

𝕋Xsubscript𝕋𝑋\textstyle{\mathbb{T}_{X}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}asuperscript𝑎\scriptstyle{a^{\vee}}00\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}f𝕋Fsuperscript𝑓subscript𝕋𝐹\textstyle{f^{*}\mathbb{T}_{F}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}tf,ωsubscript𝑡𝑓𝜔\scriptstyle{t_{f,\omega}}𝕃X[n].subscript𝕃𝑋delimited-[]𝑛\textstyle{\mathbb{L}_{X}[n].}

In particular, if we denote by 𝕋f,ωsubscriptsuperscript𝕋perpendicular-to𝑓𝜔\mathbb{T}^{\perp}_{f,\omega} the pullback in the diagram

𝕋f,ωsubscriptsuperscript𝕋perpendicular-to𝑓𝜔\textstyle{\mathbb{T}^{\perp}_{f,\omega}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}00\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}f𝕋Fsuperscript𝑓subscript𝕋𝐹\textstyle{f^{*}\mathbb{T}_{F}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}tf,ωsubscript𝑡𝑓𝜔\scriptstyle{t_{f,\omega}}𝕃X[n]subscript𝕃𝑋delimited-[]𝑛\textstyle{\mathbb{L}_{X}[n]}

(i.e. the kernel of tf,ωsubscript𝑡𝑓𝜔t_{f,\omega}), we get a canonical induced map θγ:𝕋X𝕋f,ω:subscript𝜃𝛾subscript𝕋𝑋subscriptsuperscript𝕋perpendicular-to𝑓𝜔\theta_{\gamma}:\mathbb{T}_{X}\to\mathbb{T}^{\perp}_{f,\omega} in 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X).

Definition 1.2.3

In the setting of Definition 1.2.2, an isotropic structure γπ0(𝖨𝗌𝗈𝗍𝗋(f;ω))𝛾subscript𝜋0𝖨𝗌𝗈𝗍𝗋𝑓𝜔\gamma\in\pi_{0}(\mathsf{Isotr}(f;\omega)) is called non-degenerate or lagrangian if the induced map θγ:𝕋X𝕋f,ω:subscript𝜃𝛾subscript𝕋𝑋subscriptsuperscript𝕋perpendicular-to𝑓𝜔\theta_{\gamma}:\mathbb{T}_{X}\to\mathbb{T}^{\perp}_{f,\omega} is an equivalence in 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) (i.e. an isomorphism in the underlying derived/homotopy category). The space 𝖫𝖺𝗀𝗋(f;ω)𝖫𝖺𝗀𝗋𝑓𝜔\mathsf{Lagr}(f;\omega) of lagrangian structures on f:XF:𝑓𝑋𝐹f:X\to F relative to ω𝜔\omega is the subspace of 𝖨𝗌𝗈𝗍𝗋(f;ω)𝖨𝗌𝗈𝗍𝗋𝑓𝜔\mathsf{Isotr}(f;\omega) consisting of connected components of lagrangian structures.

Remark 1.2.4

(1) It is easy to check that if X𝑋X and F𝐹F are underived smooth schemes, ω𝜔\omega is a usual (i.e. 00-shifted) symplectic structure on F𝐹F, and f𝑓f is a closed immersion, then 𝖫𝖺𝗀𝗋(f;ω)𝖫𝖺𝗀𝗋𝑓𝜔\mathsf{Lagr}(f;\omega) is either empty or contractible, and it is in fact contractible iff X𝑋X is a usual smooth lagrangian subscheme of F𝐹F via f𝑓f. The nondegeneracy condition ensures that the dimension of X𝑋X is then half of the dimension of F𝐹F, and in fact 𝕋f,ωsubscriptsuperscript𝕋perpendicular-to𝑓𝜔\mathbb{T}^{\perp}_{f,\omega} is then quasi-isomorphic to the usual symplectic orthogonal of 𝕋X=TXsubscript𝕋𝑋subscriptT𝑋\mathbb{T}_{X}=\mathrm{T}_{X} in 𝕋F=TXsubscript𝕋𝐹subscriptT𝑋\mathbb{T}_{F}=\mathrm{T}_{X}. Thus, the notion of lagrangian structure reduces to the usual notion of lagrangian subscheme in this case.

(2) By rephrasing Definition 1.2.3, an isotropic structure γπ0(𝖨𝗌𝗈𝗍𝗋(f;ω))𝛾subscript𝜋0𝖨𝗌𝗈𝗍𝗋𝑓𝜔\gamma\in\pi_{0}(\mathsf{Isotr}(f;\omega)) is lagrangian iff the (homotopy) commutative square

𝕋Xsubscript𝕋𝑋\textstyle{\mathbb{T}_{X}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}asuperscript𝑎\scriptstyle{a^{\vee}}00\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}f𝕋Fsuperscript𝑓subscript𝕋𝐹\textstyle{f^{*}\mathbb{T}_{F}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}tf,ωsubscript𝑡𝑓𝜔\scriptstyle{t_{f,\omega}}𝕃X[n]subscript𝕃𝑋delimited-[]𝑛\textstyle{\mathbb{L}_{X}[n]}

is actually a pullback square. But the square

f𝕋Fsuperscript𝑓subscript𝕋𝐹\textstyle{f^{*}\mathbb{T}_{F}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}tf,ωsubscript𝑡𝑓𝜔\scriptstyle{t_{f,\omega}}f(u(ω))superscript𝑓𝑢superscript𝜔\scriptstyle{f^{*}(u(\omega)^{\flat})}𝕃X[n]subscript𝕃𝑋delimited-[]𝑛\textstyle{\mathbb{L}_{X}[n]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}idid\scriptstyle{\mathrm{id}}f𝕃F[n]superscript𝑓subscript𝕃𝐹delimited-[]𝑛\textstyle{f^{*}\mathbb{L}_{F}[n]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}a[n]𝑎delimited-[]𝑛\scriptstyle{a[n]}𝕃X[n]subscript𝕃𝑋delimited-[]𝑛\textstyle{\mathbb{L}_{X}[n]}

is a pullback because f(u(ω))superscript𝑓𝑢superscript𝜔f^{*}(u(\omega)^{\flat}) is an equivalence (u(ω)𝑢𝜔u(\omega) being non-degenerate), hence γ𝛾\gamma is lagrangian iff the outer square in

𝕋Xsubscript𝕋𝑋\textstyle{\mathbb{T}_{X}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}asuperscript𝑎\scriptstyle{a^{\vee}}00\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}f𝕋Fsuperscript𝑓subscript𝕋𝐹\textstyle{f^{*}\mathbb{T}_{F}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}f(u(ω))superscript𝑓𝑢superscript𝜔\scriptstyle{f^{*}(u(\omega)^{\flat})}tf,ωsubscript𝑡𝑓𝜔\scriptstyle{t_{f,\omega}}𝕃X[n]subscript𝕃𝑋delimited-[]𝑛\textstyle{\mathbb{L}_{X}[n]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}idid\scriptstyle{\mathrm{id}}f𝕃F[n]superscript𝑓subscript𝕃𝐹delimited-[]𝑛\textstyle{f^{*}\mathbb{L}_{F}[n]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}a[n]𝑎delimited-[]𝑛\scriptstyle{a[n]}𝕃X[n]subscript𝕃𝑋delimited-[]𝑛\textstyle{\mathbb{L}_{X}[n]}

is cartesian, i.e. iff the induced canonical map ργ:𝕋Xker(a[n])𝕃f[n1]:subscript𝜌𝛾subscript𝕋𝑋kernel𝑎delimited-[]𝑛similar-to-or-equalssubscript𝕃𝑓delimited-[]𝑛1\rho_{\gamma}:\mathbb{T}_{X}\to\ker(a[n])\simeq\mathbb{L}_{f}[n-1] is an equivalence i.e. iff the shifted dual map Θγ:=ργ[n1]:𝕋f𝕃X[n1]:assignsubscriptΘ𝛾superscriptsubscript𝜌𝛾delimited-[]𝑛1subscript𝕋𝑓subscript𝕃𝑋delimited-[]𝑛1\Theta_{\gamma}:=\rho_{\gamma}^{\vee}[n-1]:\mathbb{T}_{f}\to\mathbb{L}_{X}[n-1] is an equivalence. This shows the equivalence between Definition 1.2.3 and [PTVV, Definition 2.8].

(3) As first noticed by D. Calaque, shifted symplectic structures are particular instances of lagrangian structures (a fact that is obviously false inside usual, 00-shifted and underived algebraic geometry). In fact, let n𝑛n\in\mathbb{Z}, X𝑋X be a derived stack lfp over k𝑘k, f:X𝐒𝐩𝐞𝐜k:𝑓𝑋𝐒𝐩𝐞𝐜𝑘f:X\to\mathbf{Spec}\,k be the structure map, and let us endow 𝐒𝐩𝐞𝐜k𝐒𝐩𝐞𝐜𝑘\mathbf{Spec}\,k with the unique (n+1)𝑛1(n+1)-shifted symplectic structure ωn+10superscriptsubscript𝜔𝑛10\omega_{n+1}^{0} (note that 𝕃kΩk/k1[0]0[0]similar-to-or-equalssubscript𝕃𝑘superscriptsubscriptΩ𝑘𝑘1delimited-[]0similar-to-or-equals0delimited-[]0\mathbb{L}_{k}\simeq\Omega_{k/k}^{1}[0]\simeq 0[0], therefore for any m𝑚m\in\mathbb{Z}, there is a unique m𝑚m-shifted symplectic form given by the the shift of the zero form). Now, if γπ0(𝖫𝖺𝗀𝗋(f;ω))𝛾subscript𝜋0𝖫𝖺𝗀𝗋𝑓𝜔\gamma\in\pi_{0}(\mathsf{Lagr}(f;\omega)) as in Definition 1.2.3, then γ𝛾\gamma is actually a loop at 00 inside 𝒜2,cl(X,n+1)superscript𝒜2𝑐𝑙𝑋𝑛1\mathcal{A}^{2,cl}(X,n+1), and its class [γ]π1(Ω0,0(𝒜2,cl(X,n+1)))π0(𝒜2,cl(X,n))delimited-[]𝛾subscript𝜋1subscriptΩ00superscript𝒜2𝑐𝑙𝑋𝑛1similar-to-or-equalssubscript𝜋0superscript𝒜2𝑐𝑙𝑋𝑛[\gamma]\in\pi_{1}(\Omega_{0,0}(\mathcal{A}^{2,cl}(X,n+1)))\simeq\pi_{0}(\mathcal{A}^{2,cl}(X,n))333This isomorphism is perhaps more familar to the reader in the following form: a self-homotopy hh of the zero map of complexes 0:EF:0𝐸𝐹0:E\to F is the same thing as map of complexes EF[1]𝐸𝐹delimited-[]1E\to F[1]. gives an n𝑛n-shifted closed 2-form ωγsubscript𝜔𝛾\omega_{\gamma} on X𝑋X. The non-degeneracy condition on γ𝛾\gamma is equivalent to the fact that the induced map Θγ:𝕋f𝕋X𝕃X[n1]:subscriptΘ𝛾similar-to-or-equalssubscript𝕋𝑓subscript𝕋𝑋subscript𝕃𝑋delimited-[]𝑛1\Theta_{\gamma}:\mathbb{T}_{f}\simeq\mathbb{T}_{X}\to\mathbb{L}_{X}[n-1] of point (2) in this Remark, is an equivalence. But it is easy to check that this map is exactly ωγsubscriptsuperscript𝜔𝛾\omega^{\flat}_{\gamma}, hence ωγsubscript𝜔𝛾\omega_{\gamma} is indeed an n𝑛n-shifted symplectic form on X𝑋X. By using again that π1(Ω0,0(𝒜2,cl(X,n+1)))π0(𝒜2,cl(X,n))similar-to-or-equalssubscript𝜋1subscriptΩ00superscript𝒜2𝑐𝑙𝑋𝑛1subscript𝜋0superscript𝒜2𝑐𝑙𝑋𝑛\pi_{1}(\Omega_{0,0}(\mathcal{A}^{2,cl}(X,n+1)))\simeq\pi_{0}(\mathcal{A}^{2,cl}(X,n)), and running the previous argument backwards, we get an equivalence of spaces 𝖲𝗒𝗆𝗉𝗅(X,n)𝖫𝖺𝗀𝗋(f:X𝐒𝐩𝐞𝐜k,ωn+10)\mathsf{Sympl}(X,n)\simeq\mathsf{Lagr}(f:X\to\mathbf{Spec}\,k,\omega_{n+1}^{0}).

The link between shifted symplectic structures and lagrangian structures expressed in Remark 1.2.4 (3) extends to the case of lagrangian intersections as follows.

Theorem 1.2.5

Let n𝑛n\in\mathbb{Z}, (F,ω)𝐹𝜔(F,\omega) be a n𝑛n-shifted symplectic stack, fi:XiF:subscript𝑓𝑖subscript𝑋𝑖𝐹f_{i}:X_{i}\to F, i=1,2𝑖12i=1,2 maps between derived Artin stacks lfp over k𝑘k, and γisubscript𝛾𝑖\gamma_{i} lagrangian structures on fisubscript𝑓𝑖f_{i} relative to ω𝜔\omega, i=1,2𝑖12i=1,2. Then, there is a canonical induced (n1)𝑛1(n-1)-shifted symplectic structure on the fiber product X1×FX2subscript𝐹subscript𝑋1subscript𝑋2X_{1}\times_{F}X_{2}.

Recall that, according to our conventions, all fiber products of derived stacks are taken in the \infty-category of derived stacks, and are therefore implicitly derived fiber products. We refer the reader to [PTVV, Theorem 2.9] for a proof of Theorem 1.2.5. We will only give a sketch of the argument in the classical case i.e. for n=0𝑛0n=0, F,X1,X2𝐹subscript𝑋1subscript𝑋2F,X_{1},X_{2} underived smooth schemes, and fisubscript𝑓𝑖f_{i} closed immersions, i=1,2𝑖12i=1,2, in order to convey the main idea of why Theorem 1.2.5 is true. Under our hypotheses, X1subscript𝑋1X_{1} and X2subscript𝑋2X_{2} are usual lagrangian smooth subschemes of the smooth symplectic scheme F𝐹F. If X12:=X1×FX2assignsubscript𝑋12subscript𝐹subscript𝑋1subscript𝑋2X_{12}:=X_{1}\times_{F}X_{2} denotes the (derived) intersection of X1subscript𝑋1X_{1} and X2subscript𝑋2X_{2} inside F𝐹F, we may pull back the closed form ω𝜔\omega to X12subscript𝑋12X_{12} in two different ways, i.e. using f1subscript𝑓1f_{1} or f2subscript𝑓2f_{2}, and get two closed forms ω1subscript𝜔1\omega_{1} and ω2subscript𝜔2\omega_{2} on X12subscript𝑋12X_{12}. Now, X12subscript𝑋12X_{12} is a derived fiber product, hence these two pullbacks come equipped with a canonical path γ𝛾\gamma between them inside 𝒜2,cl(X12,0)superscript𝒜2𝑐𝑙subscript𝑋120\mathcal{A}^{2,cl}(X_{12},0). On the other hand, as X1subscript𝑋1X_{1} and X2subscript𝑋2X_{2} are lagrangian subschemes we have ω1=ω2=0subscript𝜔1subscript𝜔20\omega_{1}=\omega_{2}=0, so that γ𝛾\gamma is in fact a loop at 00 in 𝒜2,cl(X12,0)superscript𝒜2𝑐𝑙subscript𝑋120\mathcal{A}^{2,cl}(X_{12},0). Since, π1(𝒜2,cl(X12,0);0)π0(𝒜2,cl(X12,1))similar-to-or-equalssubscript𝜋1superscript𝒜2𝑐𝑙subscript𝑋1200subscript𝜋0superscript𝒜2𝑐𝑙subscript𝑋121\pi_{1}(\mathcal{A}^{2,cl}(X_{12},0);0)\simeq\pi_{0}(\mathcal{A}^{2,cl}(X_{12},-1)), the class [γ]π0(𝒜2,cl(X12,1))delimited-[]𝛾subscript𝜋0superscript𝒜2𝑐𝑙subscript𝑋121[\gamma]\in\pi_{0}(\mathcal{A}^{2,cl}(X_{12},-1)) defines a (1)1(-1)-shifted closed 2-form on X12subscript𝑋12X_{12}, whose non-degeneracy follows easily from the same property for ω𝜔\omega. Thus [γ]delimited-[]𝛾[\gamma] is a (1)1(-1)-shifted symplectic structure on the derived intersection X12subscript𝑋12X_{12}. The appearence of a (1)1(-1)-shift here perhaps explains why this phenomenon, even though arising from a completely classical situation in usual (algebraic) symplectic geometry, was not observed before.

Remark 1.2.6

The following special case of 1.2.5 is particularly relevant for applications to Donaldson-Thomas invariants (see Theorem 1.2.10 below). Let X𝑋X be a smooth underived scheme and g:X𝔸k1:𝑔𝑋superscriptsubscript𝔸𝑘1g:X\to\mathbb{A}_{k}^{1} a smooth function. We may embed X𝑋X inside its cotangent bundle either via the zero section or via the differential dg:XTX:𝑑𝑔𝑋superscriptT𝑋dg:X\to\mathrm{T}^{*}X, and both these immersions are lagrangian with respect to the standard symplectic structure on TXsuperscriptT𝑋\mathrm{T}^{*}X. The derived intersection of these two lagrangians is called the derived critical locus of g𝑔g, and is denoted by 𝖽𝖢𝗋𝗂𝗍(g)𝖽𝖢𝗋𝗂𝗍𝑔\mathsf{dCrit}(g). Note that its truncation is the usual scheme-theoretic critical locus of g𝑔g. Now, Theorem 1.2.5 endow 𝖽𝖢𝗋𝗂𝗍(g)𝖽𝖢𝗋𝗂𝗍𝑔\mathsf{dCrit}(g) with a canonical (1)1(-1)-shifted symplectic structure ωgsubscript𝜔𝑔\omega_{g}. One can rather easily give an explicit description of ωgsubscript𝜔𝑔\omega_{g} using Koszul resolutions (see, e.g. [Vez1]), and observe that a similar result holds by replacing dg𝑑𝑔dg with an arbitrary closed 1-form on X𝑋X. However, since derived critical loci are particularly important as local models of (1)1(-1)-shifted symplectic structures, the generalization to derived zero schemes of closed 1-forms has not yet received much attention. Also notice that the previous construction can be easily generalized to build derived zero loci of sections of arbitrary locally free sheaves on X𝑋X.

Remark 1.2.7

One may combine Theorem 1.2.1 and Definition 1.2.3 together with a notion of relative orientation, in order to get a relative version of Theorem 1.2.1. This is due to D. Calaque ([Cal, 2.3]). Here is an outline of the construction. Given a map of derived stacks b:BX:𝑏𝐵𝑋b:B\to X, and a perfect complex E𝐸E on X𝑋X, we let (XrelB,E)𝑋rel𝐵𝐸\mathbb{H}(X\textrm{rel}B,E) be the fiber of the induced map bE:Γ(X,E)Γ(B,bE):subscript𝑏𝐸Γ𝑋𝐸Γ𝐵superscript𝑏𝐸b_{E}:\Gamma(X,E)\to\Gamma(B,b^{*}E). We define a relative d𝑑d-orientation on b𝑏b is a map ηb:(XrelB,𝒪X)k[d]:subscript𝜂𝑏𝑋rel𝐵subscript𝒪𝑋𝑘delimited-[]𝑑\eta_{b}:\mathbb{H}(X\textrm{rel}B,\mathcal{O}_{X})\to k[-d] satisfying the following two non-degeneracy properties. First of all, we assume that B𝐵B is 𝒪𝒪\mathcal{O}-compact, and we require that the composite map Γ(B,𝒪B)(XrelB,𝒪X)[1]k[d+1]Γ𝐵subscript𝒪𝐵𝑋rel𝐵subscript𝒪𝑋delimited-[]1𝑘delimited-[]𝑑1\Gamma(B,\mathcal{O}_{B})\to\mathbb{H}(X\textrm{rel}B,\mathcal{O}_{X})[1]\to k[-d+1] defines a (d1)𝑑1(d-1)-orientation on B𝐵B. Then, for any E𝐏𝐞𝐫𝐟(X)𝐸𝐏𝐞𝐫𝐟𝑋E\in\mathbf{Perf}(X), we ask that the induced map

Γ(X,E)(XrelB,E)tensor-productΓ𝑋𝐸𝑋rel𝐵superscript𝐸\textstyle{\Gamma(X,E)\otimes\mathbb{H}(X\textrm{rel}B,E^{\vee})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}α𝛼\scriptstyle{\alpha}(XrelB,𝒪X)𝑋rel𝐵subscript𝒪𝑋\textstyle{\mathbb{H}(X\textrm{rel}B,\mathcal{O}_{X})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ηbsubscript𝜂𝑏\scriptstyle{\eta_{b}}k[d]𝑘delimited-[]𝑑\textstyle{k[-d]}

yields, by adjunction, an equivalence Γ(X,E)(XrelB,E)[d]similar-to-or-equalsΓ𝑋𝐸𝑋rel𝐵superscript𝐸delimited-[]𝑑\Gamma(X,E)\simeq\mathbb{H}(X\textrm{rel}B,E^{\vee})[-d]. Here α𝛼\alpha is the map induced on the vertical fibers of

Γ(X,E)Γ(X,E)tensor-productΓ𝑋𝐸Γ𝑋superscript𝐸\textstyle{\Gamma(X,E)\otimes\Gamma(X,E^{\vee})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}tridbEtensor-productidsubscript𝑏superscript𝐸\scriptstyle{\textrm{id}\otimes b_{E^{\vee}}}Γ(X,𝒪X)Γ𝑋subscript𝒪𝑋\textstyle{\Gamma(X,\mathcal{O}_{X})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}b𝒪Xsubscript𝑏subscript𝒪𝑋\scriptstyle{b_{\mathcal{O}_{X}}}Γ(X,E)Γ(B,bE)tensor-productΓ𝑋𝐸Γ𝐵superscript𝑏superscript𝐸\textstyle{\Gamma(X,E)\otimes\Gamma(B,b^{*}E^{\vee})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}tr(bEid)trtensor-productsubscript𝑏𝐸id\scriptstyle{\textrm{tr}\,(b_{E}\otimes\textrm{id})}Γ(B,𝒪B).Γ𝐵subscript𝒪𝐵\textstyle{\Gamma(B,\mathcal{O}_{B}).}

Given a d𝑑d-orientation ηbsubscript𝜂𝑏\eta_{b} on the map b:BX:𝑏𝐵𝑋b:B\to X, a (n+1)𝑛1(n+1)-shifted symplectic stack Ysuperscript𝑌Y^{\prime}, a map :YY:𝑌superscript𝑌\ell:Y\to Y^{\prime}, and a lagrangian structure on f𝑓f, we may consider the derived stack

𝖬𝖠𝖯𝐝𝐒𝐭(b,f):=𝖬𝖠𝖯𝐝𝐒𝐭(B,Y)×𝖬𝖠𝖯𝐝𝐒𝐭(B,Y)𝖬𝖠𝖯𝐝𝐒𝐭(X,Y)assignsubscript𝖬𝖠𝖯𝐝𝐒𝐭𝑏𝑓subscriptsubscript𝖬𝖠𝖯𝐝𝐒𝐭𝐵superscript𝑌subscript𝖬𝖠𝖯𝐝𝐒𝐭𝐵𝑌subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋superscript𝑌\mathsf{MAP}_{\mathbf{dSt}}(b,f):=\mathsf{MAP}_{\mathbf{dSt}}(B,Y)\times_{\mathsf{MAP}_{\mathbf{dSt}}(B,Y^{\prime})}\mathsf{MAP}_{\mathbf{dSt}}(X,Y^{\prime})

of arrows from b𝑏b to f𝑓f. The generalization of Theorem 1.2.1 to this relative situation says that if 𝖬𝖠𝖯𝐝𝐒𝐭(b,)subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑏\mathsf{MAP}_{\mathbf{dSt}}(b,\ell) and 𝖬𝖠𝖯𝐝𝐒𝐭(X,Y)subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝑌\mathsf{MAP}_{\mathbf{dSt}}(X,Y) are derived Artin stacks lfp over k𝑘k, then 𝖬𝖠𝖯𝐝𝐒𝐭(b,)subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑏\mathsf{MAP}_{\mathbf{dSt}}(b,\ell) has a canonical (nd+1)𝑛𝑑1(n-d+1)-shifted symplectic form, and there is a canonical lagrangian structure on the natural map 𝖬𝖠𝖯𝐝𝐒𝐭(X,Y)𝖬𝖠𝖯𝐝𝐒𝐭(b,)subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝑌subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑏\mathsf{MAP}_{\mathbf{dSt}}(X,Y)\to\mathsf{MAP}_{\mathbf{dSt}}(b,\ell). If we take B𝐵B to be empty (so that ηbsubscript𝜂𝑏\eta_{b} is just a d𝑑d-orientation on X𝑋X), and Y=(𝐒𝐩𝐞𝐜k,ωn+10)𝑌𝐒𝐩𝐞𝐜𝑘superscriptsubscript𝜔𝑛10Y=(\mathbf{Spec}\,k,\omega_{n+1}^{0}), we have 𝖬𝖠𝖯𝐝𝐒𝐭(b,)(𝐒𝐩𝐞𝐜k,ωnd+10)similar-to-or-equalssubscript𝖬𝖠𝖯𝐝𝐒𝐭𝑏𝐒𝐩𝐞𝐜𝑘superscriptsubscript𝜔𝑛𝑑10\mathsf{MAP}_{\mathbf{dSt}}(b,\ell)\simeq(\mathbf{Spec}\,k,\omega_{n-d+1}^{0}), and by Remark 1.2.4 (3), we get back Theorem 1.2.1. But we may also take Y=(𝐒𝐩𝐞𝐜k,ωn+10)𝑌𝐒𝐩𝐞𝐜𝑘superscriptsubscript𝜔𝑛10Y=(\mathbf{Spec}\,k,\omega_{n+1}^{0}), and B𝐵B arbitrary (non-empty), and we get a lagrangian structure on the restriction map 𝖬𝖠𝖯𝐝𝐒𝐭(X,Y)𝖬𝖠𝖯𝐝𝐒𝐭(B,Y)𝖬𝖠𝖯𝐝𝐒𝐭(b,)subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝑌subscript𝖬𝖠𝖯𝐝𝐒𝐭𝐵𝑌similar-to-or-equalssubscript𝖬𝖠𝖯𝐝𝐒𝐭𝑏\mathsf{MAP}_{\mathbf{dSt}}(X,Y)\to\mathsf{MAP}_{\mathbf{dSt}}(B,Y)\simeq\mathsf{MAP}_{\mathbf{dSt}}(b,\ell), where 𝖬𝖠𝖯𝐝𝐒𝐭(B,Y)subscript𝖬𝖠𝖯𝐝𝐒𝐭𝐵𝑌\mathsf{MAP}_{\mathbf{dSt}}(B,Y) is (nd+1)𝑛𝑑1(n-d+1)-symplectic (consistently with Theorem 1.2.1, since B𝐵B is (d1)𝑑1(d-1)-oriented by hypothesis, and Y𝑌Y is n𝑛n-shifted symplectic by Remark 1.2.4 (3)). In other words, what we gain in this case, is that restriction to the boundary (for maps to a fixed shifted symplectic target) is endowed with a lagrangian structure. Examples of relative orientations includes topological examples (where b𝑏b is the inclusion of the boundary in a compact oriented topological d𝑑d-manifold with boundary), and algebro-geometric ones where b𝑏b is the inclusion of the derived zero locus (Remark 1.2.6) of a section of the anti-canonical bundle of a smooth projective variety of dimension d𝑑d. When B𝐵B is a K3𝐾3K3 surface inside a Fano 333-fold, this might be compared with [Ty, Proposition 2.2]. For more details, we address the reader to [Cal, 3.2.2], and [To2, p. 227].

The last general existence theorem for shifted symplectic structures is a generalization of the 222-shifted symplectic structure on 𝖡GLn𝖡subscriptGL𝑛\mathsf{B}\mathrm{GL}_{n} described at the beginning of this subsection.
Let 𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿\mathsf{Perf} be the derived stack classifying perfect complexes. It can be defined as the functor sending a cdga A𝐴A to the nerve of the category of cofibrant perfect A𝐴A-dg-modules with morphisms only the quasi-isomorphisms (as an \infty-functor it sends A𝐴A to the coherent nerve of the Dwyer-Kan localization of the previous category). The truncation of 𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿\mathsf{Perf} is the (higher) stack first introduced and studied in [HS]. Though 𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿\mathsf{Perf} is not strictly speaking a derived Artin stack lfp over k𝑘k, it is quite close to it: it is locally geometric, i.e. it is a union of open derived Artin substacks lfp over k𝑘k, e.g. 𝖯𝖾𝗋𝖿n0𝖯𝖾𝗋𝖿[n,n]similar-to-or-equals𝖯𝖾𝗋𝖿subscript𝑛0superscript𝖯𝖾𝗋𝖿𝑛𝑛\mathsf{Perf}\simeq\cup_{n\geq 0}\mathsf{Perf}^{[-n,n]}, where 𝖯𝖾𝗋𝖿[n,n]superscript𝖯𝖾𝗋𝖿𝑛𝑛\mathsf{Perf}^{[-n,n]} is the derived stack classifying perfect complexes of Tor-amplitude contained in [n,n]𝑛𝑛[-n,n] ([To-Va, Proposition 3.7]). In particular, it makes sense to ask wether 𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿\mathsf{Perf} carries a shifted symplectic structure, as we know its substack 𝖯𝖾𝗋𝖿[0,0]=𝖡GLn0𝖡GLnsuperscript𝖯𝖾𝗋𝖿00𝖡GLsimilar-to-or-equalssubscriptcoproduct𝑛0𝖡subscriptGL𝑛\mathsf{Perf}^{[-0,0]}=\mathsf{B}\mathrm{GL}\simeq\coprod_{n\geq 0}\mathsf{B}\mathrm{GL}_{n} does. As shown in [PTVV, Theorem 2.12], the answer is affirmative.

Theorem 1.2.8

The derived stack 𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿\mathsf{Perf} has a canonical 222-shifted symplectic structure.

We will briefly give an idea of the proof, and address the readers to [PTVV, Theorem 2.12] for all details. First of all, by definition, 𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿\mathsf{Perf} carries a universal perfect complex \mathcal{E}, and we consider its perfect 𝒪𝖯𝖾𝗋𝖿subscript𝒪𝖯𝖾𝗋𝖿\mathcal{O}_{\mathsf{Perf}}-Algebra of endomorphisms :=End¯𝒪𝖯𝖾𝗋𝖿()𝒪𝖯𝖾𝗋𝖿assignsubscript¯𝐸𝑛𝑑subscript𝒪𝖯𝖾𝗋𝖿similar-to-or-equalssubscripttensor-productsubscript𝒪𝖯𝖾𝗋𝖿superscript\mathcal{B}:=\mathbb{R}\underline{End}_{\mathcal{O}_{\mathsf{Perf}}}(\mathcal{E})\simeq\mathcal{E}^{\vee}\otimes_{\mathcal{O}_{\mathsf{Perf}}}\mathcal{E}. One checks that 𝕋𝖯𝖾𝗋𝖿[1]similar-to-or-equalssubscript𝕋𝖯𝖾𝗋𝖿delimited-[]1\mathbb{T}_{\mathsf{Perf}}\simeq\mathcal{B}[1], and thus gets a well defined map

ω0:𝕋𝖯𝖾𝗋𝖿𝒪𝖯𝖾𝗋𝖿𝕋𝖯𝖾𝗋𝖿:superscript𝜔0subscriptsubscript𝒪𝖯𝖾𝗋𝖿subscript𝕋𝖯𝖾𝗋𝖿subscript𝕋𝖯𝖾𝗋𝖿\textstyle{\omega^{0}:\mathbb{T}_{\mathsf{Perf}}\wedge_{\mathcal{O}_{\mathsf{Perf}}}\mathbb{T}_{\mathsf{Perf}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}similar-to\scriptstyle{\sim}Sym𝒪𝖯𝖾𝗋𝖿()[2]𝑆𝑦subscript𝑚subscript𝒪𝖯𝖾𝗋𝖿delimited-[]2\textstyle{Sym_{\mathcal{O}_{\mathsf{Perf}}}(\mathcal{B})[2]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}mult[2]𝒪𝖯𝖾𝗋𝖿[2]similar-to-or-equalsdelimited-[]2subscripttensor-productsubscript𝒪𝖯𝖾𝗋𝖿superscriptdelimited-[]2\textstyle{\mathcal{B}[2]\simeq\mathcal{E}^{\vee}\otimes_{\mathcal{O}_{\mathsf{Perf}}}\mathcal{E}[2]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}tr[2]trdelimited-[]2\scriptstyle{\mathrm{tr[2]}}𝒪𝖯𝖾𝗋𝖿[2]subscript𝒪𝖯𝖾𝗋𝖿delimited-[]2\textstyle{\mathcal{O}_{\mathsf{Perf}}[2]}

where tr𝑡𝑟tr denotes the trace (or evaluation) map for the perfect complex \mathcal{E}. So, ω0subscript𝜔0\omega_{0} is a well defined, and non-degenerate 222-shifted 2-form on 𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿\mathsf{Perf}. So, it only remains to show that there exists a 222-shifted closed 2-form ω𝜔\omega on 𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿\mathsf{Perf}, whose underlying 2-form is ω0superscript𝜔0\omega^{0}. By [To-Ve-2], or [Hoy, Theorem 2.1], one can prove that the weight 222 component of the refined Chern character 𝖢𝗁()𝖢𝗁\mathsf{Ch}(\mathcal{E}) (with values in negative cyclic homology) as defined in [To-Ve-1] is indeed a 222-shifted closed 2-form on 𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿\mathsf{Perf} whose underlying 2-form is 12ω012superscript𝜔0\frac{1}{2}\omega^{0}, thus ω:=2𝖢𝗁2()assign𝜔2subscript𝖢𝗁2\omega:=2\mathsf{Ch}_{2}(\mathcal{E}) is indeed a 222-shifted symplectic form on 𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿\mathsf{Perf} whose underlying 2-form is ω0superscript𝜔0\omega^{0}.

By combining Theorem 1.2.1, 1.2.5, and 1.2.8, we get the following (non-exhaustive) list of geometrically interesting, classes of examples of n𝑛n-shifted symplectic derived stacks:

\bullet the derived stack 𝖯𝖾𝗋𝖿(X):=𝖬𝖠𝖯𝐝𝐒𝐭(X,𝖯𝖾𝗋𝖿)assign𝖯𝖾𝗋𝖿𝑋subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝖯𝖾𝗋𝖿\mathsf{Perf}(X):=\mathsf{MAP}_{\mathbf{dSt}}(X,\mathsf{Perf}) of perfect complexes on a Calabi-Yau variety of dimension d𝑑d (n=2d𝑛2𝑑n=2-d).

\bullet the derived stack 𝖯𝖾𝗋𝖿(M):=𝖬𝖠𝖯𝐝𝐒𝐭(const(M),𝖯𝖾𝗋𝖿)assign𝖯𝖾𝗋𝖿𝑀subscript𝖬𝖠𝖯𝐝𝐒𝐭const𝑀𝖯𝖾𝗋𝖿\mathsf{Perf}(M):=\mathsf{MAP}_{\mathbf{dSt}}(\mathrm{const}(M),\mathsf{Perf}) of perfect complexes on a compact oriented topological manifold M𝑀M of dimension d𝑑d (n=2d𝑛2𝑑n=2-d).

\bullet the derived stack 𝖵𝖾𝖼𝗍n(X)subscript𝖵𝖾𝖼𝗍𝑛𝑋\mathbb{R}\mathsf{Vect}_{n}(X) of rank n𝑛\ n vector bundles on a Calabi-Yau variety of dimension d𝑑d (n=2d𝑛2𝑑n=2-d).

\bullet the derived stack 𝖵𝖾𝖼𝗍n(M)subscript𝖵𝖾𝖼𝗍𝑛𝑀\mathbb{R}\mathsf{Vect}_{n}(M) of rank n𝑛n vector bundles on a compact oriented topological manifold M𝑀M of dimension d𝑑d (n=2d𝑛2𝑑n=2-d).

\bullet the derived stack 𝖡𝗎𝗇G(X)subscript𝖡𝗎𝗇𝐺𝑋\mathbb{R}\mathsf{Bun}_{G}(X) of G𝐺G-torsors on a Calabi-Yau variety of dimension d𝑑d, for G𝐺G reductive (n=2d𝑛2𝑑n=2-d).

\bullet the derived stack 𝖫𝗈𝖼G(M)subscript𝖫𝗈𝖼𝐺𝑀\mathbb{R}\mathsf{Loc}_{G}(M) of G𝐺G-local systems on a compact oriented topological manifold M𝑀M of dimension d𝑑d (n=2d𝑛2𝑑n=2-d)

\bullet 𝖬𝖠𝖯𝐝𝐒𝐭(X,Y)subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝑌\mathsf{MAP}_{\mathbf{dSt}}(X,Y), for X𝑋X a smooth and proper d𝑑d-dimensional Calabi-Yau scheme, and Y𝑌Y a smooth symplectic scheme (n=d𝑛𝑑n=-d).

\bullet 𝖬𝖠𝖯𝐝𝐒𝐭(X,TY[m])subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋superscriptT𝑌delimited-[]𝑚\mathsf{MAP}_{\mathbf{dSt}}(X,\mathrm{T}^{*}Y[m]), for X𝑋X a smooth and proper d𝑑d-dimensional Calabi-Yau scheme, and Y𝑌Y a derived Artin stack lfp over k𝑘k (n=md𝑛𝑚𝑑n=m-d).

\bullet iterated derived free loop spaces 𝖬𝖠𝖯𝐝𝐒𝐭(const((S1)d),Y)subscript𝖬𝖠𝖯𝐝𝐒𝐭constsuperscriptsuperscript𝑆1𝑑𝑌\mathsf{MAP}_{\mathbf{dSt}}(\mathrm{const}((S^{1})^{d}),Y) of a smooth symplectic scheme Y𝑌Y, and more generally, 𝖬𝖠𝖯𝐝𝐒𝐭(const(M),Y)subscript𝖬𝖠𝖯𝐝𝐒𝐭const𝑀𝑌\mathsf{MAP}_{\mathbf{dSt}}(\mathrm{const}(M),Y), for M𝑀M a compact oriented topological d𝑑d-manifold, and Y𝑌Y a smooth symplectic scheme (n=d𝑛𝑑n=-d).

\bullet the derived stack LocGDR(X)𝐿𝑜subscriptsuperscript𝑐𝐷𝑅𝐺𝑋\mathbb{R}Loc^{DR}_{G}(X) of flat G𝐺G-bundles on X𝑋X, a d𝑑d-dimensional smooth and proper Deligne-Mumford stack (n=22d𝑛22𝑑n=2-2d).

\bullet the derived moduli stack 𝖬𝖠𝖯𝐝𝐒𝐭(XDol,𝖯𝖾𝗋𝖿)subscript𝖬𝖠𝖯𝐝𝐒𝐭subscript𝑋Dol𝖯𝖾𝗋𝖿\mathsf{MAP}_{\mathbf{dSt}}(X_{\textrm{Dol}},\mathsf{Perf}) of Higgs fields on a proper and smooth Deligne-Mumford stack of dimension d𝑑d (n=22d𝑛22𝑑n=2-2d).

\bullet the derived moduli stack Tsubscript𝑇\mathcal{M}_{T} of compact objects in a d𝑑d-Calabi-Yau dg-category T𝑇T, e.g. the so-called non-commutative K3𝐾3K3 sub-dg-category of the derived dg-category of a cubic 444-fold (n=2d𝑛2𝑑n=2-d, and d=2𝑑2d=2 for a non-commutative K3𝐾3K3).444See e.g. [KS, 3.3] for a definition of a d𝑑d-Calabi–Yau dg-category, and [To2, 5.3] for a sketch of a proof of this result.

Remark 1.2.9

If X𝑋X is a proper and smooth scheme over k𝑘k, and D𝐷D is a smooth Calabi-Yau divisor in the anti-canonical class, then Remark 1.2.7 together with Thms. 1.2.1 and 1.2.8 give us a lagrangian structure on the restriction map 𝖯𝖾𝗋𝖿(X)=𝖬𝖠𝖯𝐝𝐒𝐭(X,𝖯𝖾𝗋𝖿)𝖬𝖠𝖯𝐝𝐒𝐭(D,𝖯𝖾𝗋𝖿)=𝖯𝖾𝗋𝖿(D)𝖯𝖾𝗋𝖿𝑋subscript𝖬𝖠𝖯𝐝𝐒𝐭𝑋𝖯𝖾𝗋𝖿subscript𝖬𝖠𝖯𝐝𝐒𝐭𝐷𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿𝐷\mathsf{Perf}(X)=\mathsf{MAP}_{\mathbf{dSt}}(X,\mathsf{Perf})\to\mathsf{MAP}_{\mathbf{dSt}}(D,\mathsf{Perf})=\mathsf{Perf}(D), The same is true for the restriction map 𝖡𝗎𝗇G(X)𝖡𝗎𝗇G(D)subscript𝖡𝗎𝗇𝐺𝑋subscript𝖡𝗎𝗇𝐺𝐷\mathbb{R}\mathsf{Bun}_{G}(X)\to\mathbb{R}\mathsf{Bun}_{G}(D) between the derived stack of torsors under a smooth reductive group scheme G𝐺G over k𝑘k.

As a sample consequence of Theorem 1.2.1, 1.2.5, and 1.2.8, we recall the following important result ([BBJ, Corollary 5.19]) by Brav-Bussi-Joyce, establishing the existence of a local algebraic potential for Donaldson-Thomas moduli spaces attached to Calabi-Yau 333-folds.

Theorem 1.2.10

(Brav-Bussi-Joyce) The moduli space DTsubscript𝐷𝑇\mathcal{M}_{DT} of simple coherent sheaves or of complexes of coherent sheaves on a Calabi-Yau 333-fold is Zariski-locally isomorphic, as a (1)1(-1)-shifted symplectic derived scheme, to the derived critical locus (as in Remark 1.2.6) of a regular function f:U𝔸k1:𝑓𝑈superscriptsubscript𝔸𝑘1f:U\to\mathbb{A}_{k}^{1} over a smooth k𝑘k-scheme U𝑈U.

The function in the statement is called the Donaldson-Thomas potential. Our existence theorems combine to give DTsubscript𝐷𝑇\mathcal{M}_{DT} a (1)1(-1)-shifted symplectic structure, and the authors achieve Theorem 1.2.10 by proving a local structure theorem (Darboux style) for derived schemes X𝑋X endowed with a (1)1(-1)-shifted symplectic structure: any such X𝑋X is Zariski-locally a derived critical locus of a regular function on a smooth scheme. A similar statement in the \mathbb{C}-analytic category was proved before by Joyce-Song [JS, Theorem 5.4] using Gauge Theory, and a version valid formally locally at any point of DTsubscript𝐷𝑇\mathcal{M}_{DT}, by Kontsevich-Soibelman [KS, Section 3.3]. Obviously, Theorem 1.2.10 is a considerably stronger and more precise result.

2 Shifted Poisson structures

Setting up a notion of shifted Poisson structure for sufficiently general derived Artin stacks turns out to be much more complicated than the case of shifted symplectic structures, described in the previous section. On the other hand, a shifted Poisson structure on a derived stack F𝐹F is exactly the right structure that controls the deformation quantization of the \infty-category of perfect complexes on F𝐹F. Therefore, in order to establish deformation quantizations for all the shifted symplectic derived moduli spaces listed at the end of Section 1, one also needs a comparison between shifted symplectic structures and (non-degenerate) shifted Poisson structures. Unfortunately, this comparison, which classically takes no more than two lines, is rather tricky in the derived setting, due to the fact that all the structures involved in the comparison are weak ones i.e. defined up-to-homotopy. The general theory of shifted Poisson structures, a comparison with shifted symplectic structures, and applications to deformation quantization of derived moduli spaces, have all been developed recently in [CPTVV], which is more than 100 pages long. In this Section we give a summary of the main constructions and results from [CPTVV], and a guide to its reading.

2.1 Differential calculus

In order to be able to define and study shifted Poisson structures on derived Artin stacks, we will need to have at our disposal a machinery of derived differential calculus (de Rham complex, shifted polyvectors etc.) working in sufficiently general setting. Thus, let \mathcal{M} be a stable presentable symmetric monoidal \infty-category which is k𝑘k-linear i.e. enriched over the \infty-category 𝐝𝐠ksubscript𝐝𝐠𝑘\mathbf{dg}_{k} ([Lu4, Definition 2.0.0.7]). In this paper, \mathcal{M} will always be obtained as the coherent nerve of the Dwyer-Kan localization of a 𝖽𝗀ksubscript𝖽𝗀𝑘\mathsf{dg}_{k}-enriched symmetric monoidal model category 𝖬𝖬\mathsf{M} satisfying some additional technical properties for which we address the reader to [CPTVV, 1.1 and 1.2] 555By [Ni-Sa], it is likely that any k𝑘k-linear stable presentable symmetric monoidal \infty-category can be obtained this way.. And we will suppose that the enrichment is induced by a symmetric monoidal functor 𝖽𝗀k𝖬subscript𝖽𝗀𝑘𝖬\mathsf{dg}_{k}\to\mathsf{M}.

For our present purposes, it will be enough to keep in mind some of the examples we will be most frequently interested in: \mathcal{M} could be 𝐝𝐠ksubscript𝐝𝐠𝑘\mathbf{dg}_{k} itself, the \infty-category 𝐝𝐠kgrsubscriptsuperscript𝐝𝐠𝑔𝑟𝑘\mathbf{dg}^{gr}_{k} of \mathbb{Z}-graded dg-modules over k𝑘k, the \infty-category ϵ𝐝𝐠kitalic-ϵsubscript𝐝𝐠𝑘\epsilon-\mathbf{dg}_{k} of mixed dg-modules over k𝑘k, the \infty-category ϵ𝐝𝐠kgritalic-ϵsubscriptsuperscript𝐝𝐠𝑔𝑟𝑘\epsilon-\mathbf{dg}^{gr}_{k} of \mathbb{Z}-graded mixed dg-modules over k𝑘k, or more generally, any category of diagrams in the previous examples.

We denote by 𝖢𝖠𝗅𝗀𝖬subscript𝖢𝖠𝗅𝗀𝖬\mathsf{CAlg}_{\mathsf{M}} the model category of commutative algebras in 𝖬𝖬\mathsf{M}, and by 𝐂𝐀𝐥𝐠=L(𝖢𝖠𝗅𝗀𝖬)subscript𝐂𝐀𝐥𝐠𝐿subscript𝖢𝖠𝗅𝗀𝖬\mathbf{CAlg}_{\mathcal{M}}=L(\mathsf{CAlg}_{\mathsf{M}}), the corresponding \infty-category.

The symmetric monoidal model category ϵ𝖬gritalic-ϵsuperscript𝖬𝑔𝑟\epsilon-\mathsf{M}^{gr} of mixed graded objects in 𝖬𝖬\mathsf{M} is defined by replacing 𝖽𝗀ksubscript𝖽𝗀𝑘\mathsf{dg}_{k} with 𝖬𝖬\mathsf{M} in the definition at the beginning of Section 1.1 (and with the cohomological shift defined here as P[1]:=Pk[1]assign𝑃delimited-[]1superscript𝑃𝑘delimited-[]1P[1]:=P^{k[-1]}, for P𝖬gr𝑃superscript𝖬𝑔𝑟P\in\mathsf{M}^{gr}, using the co-tensor enrichment of 𝖬𝖬\mathsf{M} over 𝖽𝗀ksubscript𝖽𝗀𝑘\mathsf{dg}_{k}). The model category of commutative monoids in ϵ𝖬gritalic-ϵsuperscript𝖬𝑔𝑟\epsilon-\mathsf{M}^{gr} is denoted by ϵ𝖢𝖠𝗅𝗀𝖬gritalic-ϵsubscriptsuperscript𝖢𝖠𝗅𝗀𝑔𝑟𝖬\epsilon-\mathsf{CAlg}^{gr}_{\mathsf{M}}, and called the model category of graded mixed algebras in 𝖬𝖬\mathsf{M}. The corresponding \infty-categories will be denoted by ϵgritalic-ϵsuperscript𝑔𝑟\epsilon-\mathcal{M}^{gr}, and ϵ𝐂𝐀𝐥𝐠gritalic-ϵsubscriptsuperscript𝐂𝐀𝐥𝐠𝑔𝑟\epsilon-\mathbf{CAlg}^{gr}_{\mathcal{M}}. Note that we have a canonical equivalence of \infty-categories 𝐂𝐀𝐥𝐠(ϵgr)ϵ𝐂𝐀𝐥𝐠grsimilar-to-or-equals𝐂𝐀𝐥𝐠italic-ϵsuperscript𝑔𝑟italic-ϵsubscriptsuperscript𝐂𝐀𝐥𝐠𝑔𝑟\mathbf{CAlg}(\epsilon-\mathcal{M}^{gr})\simeq\epsilon-\mathbf{CAlg}^{gr}_{\mathcal{M}} ([Lu4, Theorem 4.5.4.7]).

2.1.1 De Rham theory

As explained in Remark 1.1.5, for any A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}}, we have the associated de Rham algebra 𝐃𝐑(A)superscript𝐃𝐑𝐴\mathbf{DR}^{\mathcal{M}}(A), a mixed graded algebra in \mathcal{M}, where 𝐃𝐑:𝐂𝐀𝐥𝐠ϵ𝐂𝐀𝐥𝐠gr:superscript𝐃𝐑subscript𝐂𝐀𝐥𝐠italic-ϵsuperscriptsubscript𝐂𝐀𝐥𝐠𝑔𝑟\mathbf{DR}^{\mathcal{M}}:\mathbf{CAlg}_{\mathcal{M}}\to\epsilon-\mathbf{CAlg}_{\mathcal{M}}^{gr} is the left adjoint to the \infty-functor sending Bϵ𝐂𝐀𝐥𝐠gr𝐵italic-ϵsuperscriptsubscript𝐂𝐀𝐥𝐠𝑔𝑟B\in\epsilon-\mathbf{CAlg}_{\mathcal{M}}^{gr} to its weight 00 part B(0)𝐵0B(0). Note that if A𝐜𝐝𝐠𝐚k𝐴subscript𝐜𝐝𝐠𝐚𝑘A\in\mathbf{cdga}_{k}, then 𝐃𝐑(A)𝐃𝐑𝐴\mathbf{DR}(A) of Definition 1.1.1 is exactly 𝐃𝐑(A)superscript𝐃𝐑𝐴\mathbf{DR}^{\mathcal{M}}(A), with =𝐝𝐠ksubscript𝐝𝐠𝑘\mathcal{M}=\mathbf{dg}_{k}. There is an \infty-functor

A𝐌𝐨𝐝𝐓,P𝖣𝖾𝗋(A,P):=𝖬𝖺𝗉𝐂𝐀𝐥𝐠/A(A,AP)formulae-sequence𝐴subscript𝐌𝐨𝐝𝐓𝑃𝖣𝖾𝗋𝐴𝑃assignsubscript𝖬𝖺𝗉subscript𝐂𝐀𝐥𝐠𝐴𝐴direct-sum𝐴𝑃A-\mathbf{Mod}_{\mathcal{M}}\longrightarrow\mathbf{T}\,,\qquad P\longmapsto\mathsf{Der}(A,P):=\mathsf{Map}_{\mathbf{CAlg}_{\mathcal{M}}/A}(A,A\oplus P)

which is co-represented by an A𝐴A-module 𝕃Asubscriptsuperscript𝕃𝐴\mathbb{L}^{\mathcal{M}}_{A} in \mathcal{M}, called the cotangent complex or cotangent object of A𝐴A. As in Section 1.1, one can prove ([CPTVV, Proposition 1.3.12]) that the canonical map

SymA(𝕃A[1])𝐃𝐑(A)𝑆𝑦subscript𝑚𝐴subscriptsuperscript𝕃𝐴delimited-[]1superscript𝐃𝐑𝐴Sym_{A}(\mathbb{L}^{\mathcal{M}}_{A}[-1])\longrightarrow\mathbf{DR}^{\mathcal{M}}(A)

is functorial in A𝐴A, and an equivalence in 𝐂𝐀𝐥𝐠grsuperscriptsubscript𝐂𝐀𝐥𝐠𝑔𝑟\mathbf{CAlg}_{\mathcal{M}}^{gr} (i.e. forgetting the mixed structure in the target). In other words, the construction 𝐃𝐑superscript𝐃𝐑\mathbf{DR}^{\mathcal{M}} endows the graded algebra SymA(𝕃A[1])𝑆𝑦subscript𝑚𝐴subscriptsuperscript𝕃𝐴delimited-[]1Sym_{A}(\mathbb{L}^{\mathcal{M}}_{A}[-1]) with a canonical (weak) mixed differential, the (weak) de Rham differential.

For the sake of brevity, we will omit the completely analogous definitions and results in the relative setting, i.e. for morphisms in 𝐂𝐀𝐥𝐠subscript𝐂𝐀𝐥𝐠\mathbf{CAlg}_{\mathcal{M}}. We will instead say a few words about strict models of the above constructions, i.e. inside the model category 𝖬𝖬\mathsf{M}. First of all, if A𝐂𝐀𝐥𝐠𝖬superscript𝐴subscript𝐂𝐀𝐥𝐠𝖬A^{\prime}\in\mathbf{CAlg}_{\mathsf{M}}, then the functor

A𝐌𝐨𝐝𝖬𝐓,PDer(A,P):=Hom𝐂𝐀𝐥𝐠𝖬/A(A,AP)formulae-sequencesuperscript𝐴subscript𝐌𝐨𝐝𝖬𝐓𝑃𝐷𝑒𝑟superscript𝐴𝑃assign𝐻𝑜subscript𝑚subscript𝐂𝐀𝐥𝐠𝖬superscript𝐴superscript𝐴direct-sumsuperscript𝐴𝑃A^{\prime}-\mathbf{Mod}_{\mathsf{M}}\longrightarrow\mathbf{T}\,,\qquad P\longmapsto Der(A^{\prime},P):=Hom_{\mathbf{CAlg}_{\mathsf{M}}/A^{\prime}}(A^{\prime},A^{\prime}\oplus P)

is co-represented by an Asuperscript𝐴A^{\prime}-module Ω𝖬,A1superscriptsubscriptΩ𝖬superscript𝐴1\Omega_{\mathsf{M},A^{\prime}}^{1} in 𝖬𝖬\mathsf{M} called the module of Kähler differentials of Asuperscript𝐴A^{\prime}. If A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}} and QA𝑄𝐴QA is a cofibrant model for A𝐴A in 𝐂𝐀𝐥𝐠𝖬subscript𝐂𝐀𝐥𝐠𝖬\mathbf{CAlg}_{\mathsf{M}}, then we have a canonical equivalence Ω𝖬,QA1𝕃Asimilar-to-or-equalssuperscriptsubscriptΩ𝖬𝑄𝐴1subscriptsuperscript𝕃𝐴\Omega_{\mathsf{M},QA}^{1}\simeq\mathbb{L}^{\mathcal{M}}_{A} in A𝐌𝐨𝐝QA𝐌𝐨𝐝similar-to-or-equals𝐴subscript𝐌𝐨𝐝𝑄𝐴subscript𝐌𝐨𝐝A-\mathbf{Mod}_{\mathcal{M}}\simeq QA-\mathbf{Mod}_{\mathcal{M}}. Furthermore, the functor ϵ𝐂𝐀𝐥𝐠𝖬gr𝐂𝐀𝐥𝐠𝖬italic-ϵsuperscriptsubscript𝐂𝐀𝐥𝐠𝖬𝑔𝑟subscript𝐂𝐀𝐥𝐠𝖬\epsilon-\mathbf{CAlg}_{\mathsf{M}}^{gr}\to\mathbf{CAlg}_{\mathsf{M}} selecting the weight 00 component, has a left adjoint 𝐃𝐑𝖬superscript𝐃𝐑𝖬\mathbf{DR}^{\mathsf{M}}, thus for any A𝐂𝐀𝐥𝐠𝖬superscript𝐴subscript𝐂𝐀𝐥𝐠𝖬A^{\prime}\in\mathbf{CAlg}_{\mathsf{M}}, 𝐃𝐑𝖬(A)superscript𝐃𝐑𝖬superscript𝐴\mathbf{DR}^{\mathsf{M}}(A^{\prime}) is a strict mixed graded algebra in 𝖬𝖬\mathsf{M}. Still for an arbitrary A𝐂𝐀𝐥𝐠𝖬superscript𝐴subscript𝐂𝐀𝐥𝐠𝖬A^{\prime}\in\mathbf{CAlg}_{\mathsf{M}}, we also have a functorial isomorphism

SymA(Ω𝖬,A1[1])𝐃𝐑𝖬(A)𝑆𝑦subscript𝑚superscript𝐴superscriptsubscriptΩ𝖬superscript𝐴1delimited-[]1superscript𝐃𝐑𝖬superscript𝐴Sym_{A^{\prime}}(\Omega_{\mathsf{M},A^{\prime}}^{1}[-1])\longrightarrow\mathbf{DR}^{\mathsf{M}}(A^{\prime})

in the category 𝐂𝐀𝐥𝐠𝖬grsubscriptsuperscript𝐂𝐀𝐥𝐠𝑔𝑟𝖬\mathbf{CAlg}^{gr}_{\mathsf{M}} of strict graded algebras in 𝖬𝖬\mathsf{M}. This precisely says that SymA(Ω𝖬,A1[1])𝑆𝑦subscript𝑚superscript𝐴superscriptsubscriptΩ𝖬superscript𝐴1delimited-[]1Sym_{A^{\prime}}(\Omega_{\mathsf{M},A^{\prime}}^{1}[-1]) has a strict mixed structure, the de Rham differential. If A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}} and QA𝑄𝐴QA is a cofibrant model for A𝐴A in 𝐂𝐀𝐥𝐠𝖬subscript𝐂𝐀𝐥𝐠𝖬\mathbf{CAlg}_{\mathsf{M}}, then there is a an equivalence 𝐃𝐑𝖬(QA)𝐃𝐑(A)similar-to-or-equalssuperscript𝐃𝐑𝖬𝑄𝐴superscript𝐃𝐑𝐴\mathbf{DR}^{\mathsf{M}}(QA)\simeq\mathbf{DR}^{\mathcal{M}}(A) in ϵ𝐂𝐀𝐥𝐠gritalic-ϵsuperscriptsubscript𝐂𝐀𝐥𝐠𝑔𝑟\epsilon-\mathbf{CAlg}_{\mathcal{M}}^{gr}, i.e. 𝐃𝐑𝖬(QA)superscript𝐃𝐑𝖬𝑄𝐴\mathbf{DR}^{\mathsf{M}}(QA), or equivalently SymA(Ω𝖬,QA1[1])𝑆𝑦subscript𝑚superscript𝐴superscriptsubscriptΩ𝖬𝑄𝐴1delimited-[]1Sym_{A^{\prime}}(\Omega_{\mathsf{M},QA}^{1}[-1]) with the induced mixed structure, is a strict model for 𝐃𝐑(A)superscript𝐃𝐑𝐴\mathbf{DR}^{\mathcal{M}}(A) in ϵ𝐂𝐀𝐥𝐠gritalic-ϵsuperscriptsubscript𝐂𝐀𝐥𝐠𝑔𝑟\epsilon-\mathbf{CAlg}_{\mathcal{M}}^{gr}.

Differential forms. Again as in Section 1.1, as suggested in Remark 1.1.5, we may define (closed) shifted differential forms for commutative algebras in \mathcal{M}.

Definition 2.1.1

Let A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}}.

  • The space of closed n𝑛n-shifted p𝑝p-forms on A𝐴A is

    𝒜p,cl(A,n):=𝖬𝖺𝗉ϵgr(𝟏(p)[pn],𝐃𝐑(A))𝕋.assignsuperscriptsubscript𝒜𝑝𝑐𝑙𝐴𝑛subscript𝖬𝖺𝗉italic-ϵsuperscript𝑔𝑟subscript1𝑝delimited-[]𝑝𝑛superscript𝐃𝐑𝐴𝕋\mathcal{A}_{\mathcal{M}}^{p,cl}(A,n):=\mathsf{Map}_{\epsilon-\mathcal{M}^{gr}}(\mathbf{1}_{\mathcal{M}}(p)[-p-n],\mathbf{DR}^{\mathcal{M}}(A))\in\mathbb{T}.

    An element in π0(𝒜p,cl(A,n))subscript𝜋0superscriptsubscript𝒜𝑝𝑐𝑙𝐴𝑛\pi_{0}(\mathcal{A}_{\mathcal{M}}^{p,cl}(A,n)) is called a closed n𝑛n-shifted p𝑝p-form on A𝐴A.

  • The space of n𝑛n-shifted p𝑝p-forms on A𝐴A is 𝒜p(A,n):=𝖬𝖺𝗉(𝟏[n],Ap𝕃A)𝕋assignsuperscriptsubscript𝒜𝑝𝐴𝑛subscript𝖬𝖺𝗉subscript1delimited-[]𝑛superscriptsubscript𝐴𝑝subscriptsuperscript𝕃𝐴𝕋\mathcal{A}_{\mathcal{M}}^{p}(A,n):=\mathsf{Map}_{\mathcal{M}}(\mathbf{1}_{\mathcal{M}}[-n],\wedge_{A}^{p}\mathbb{L}^{\mathcal{M}}_{A})\in\mathbb{T}. An element in π0(𝒜p(A,n))subscript𝜋0superscriptsubscript𝒜𝑝𝐴𝑛\pi_{0}(\mathcal{A}_{\mathcal{M}}^{p}(A,n)) is called a n𝑛n-shifted p𝑝p-form on A𝐴A.

  • Since SymA(𝕃A[1])𝐃𝐑(A)similar-to-or-equals𝑆𝑦subscript𝑚𝐴subscriptsuperscript𝕃𝐴delimited-[]1superscript𝐃𝐑𝐴Sym_{A}(\mathbb{L}^{\mathcal{M}}_{A}[-1])\simeq\mathbf{DR}^{\mathcal{M}}(A), there is an induced map u:𝒜p,cl(A,n)𝒜p(A,n):𝑢superscript𝒜𝑝𝑐𝑙𝐴𝑛superscript𝒜𝑝𝐴𝑛u:\mathcal{A}^{p,cl}(A,n)\to\mathcal{A}^{p}(A,n), called the underlying p𝑝p-form map.

If ω𝜔\omega is a n𝑛n-shifted 222-form on A𝐴A, and we assume that 𝕃Asubscriptsuperscript𝕃𝐴\mathbb{L}^{\mathcal{M}}_{A} is dualizable in A𝐌𝐨𝐝𝐴subscript𝐌𝐨𝐝A-\mathbf{Mod}_{\mathcal{M}}, then u(ω)𝑢𝜔u(\omega) induces a map u(ω):𝕋A𝕃A[n]:𝑢superscript𝜔subscriptsuperscript𝕋𝐴subscriptsuperscript𝕃𝐴delimited-[]𝑛u(\omega)^{\flat}:\mathbb{T}^{\mathcal{M}}_{A}\to\mathbb{L}^{\mathcal{M}}_{A}[n] in A𝐌𝐨𝐝𝐴subscript𝐌𝐨𝐝A-\mathbf{Mod}_{\mathcal{M}}, where 𝕋Asubscriptsuperscript𝕋𝐴\mathbb{T}^{\mathcal{M}}_{A} denotes the dual of 𝕃Asubscriptsuperscript𝕃𝐴\mathbb{L}^{\mathcal{M}}_{A}. We say that ω𝜔\omega is non-degenerate if u(ω)𝑢superscript𝜔u(\omega)^{\flat} is an equivalence.

Definition 2.1.2

Let A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}}. The space of n𝑛n-shifted symplectic structures on A𝐴A is the subspace 𝖲𝗒𝗆𝗉𝗅(A,n)𝖲𝗒𝗆𝗉𝗅𝐴𝑛\mathsf{Sympl}(A,n) of 𝒜p,cl(A,n)superscriptsubscript𝒜𝑝𝑐𝑙𝐴𝑛\mathcal{A}_{\mathcal{M}}^{p,cl}(A,n) whose connected components consist of non degenerate forms.

Remark 2.1.3

Note that, even if the notation 𝖲𝗒𝗆𝗉𝗅(A,n)𝖲𝗒𝗆𝗉𝗅𝐴𝑛\mathsf{Sympl}(A,n) does not record \mathcal{M}, this space obviously depends on the category \mathcal{M} inside which we are working. In Section 2.2, we will explain how the abstract Definition 2.1.2 gives back the definition of a shifted symplectic structure on a derived Artin stack (Definition 1.1.10).

2.1.2 Polyvectors

Let 𝕃iekgr𝕃𝑖subscriptsuperscript𝑒𝑔𝑟𝑘\mathbb{L}ie^{gr}_{k} be the graded Lie operad in 𝖽𝗀ksubscript𝖽𝗀𝑘\mathsf{dg}_{k}, where the bracket operation has degree 11-1. We let 𝖽𝗀𝖫𝗂𝖾kgrsuperscriptsubscript𝖽𝗀𝖫𝗂𝖾𝑘𝑔𝑟\mathsf{dgLie}_{k}^{gr} be the model category of 𝕃iekgr𝕃𝑖subscriptsuperscript𝑒𝑔𝑟𝑘\mathbb{L}ie^{gr}_{k}-algebras, i.e. graded Lie-algebras in 𝖽𝗀ksubscript𝖽𝗀𝑘\mathsf{dg}_{k}, where the (strict) Lie bracket has weight 11-1; we denote by 𝐝𝐠𝐋𝐢𝐞kgrsuperscriptsubscript𝐝𝐠𝐋𝐢𝐞𝑘𝑔𝑟\mathbf{dgLie}_{k}^{gr}, the corresponding \infty-category L(𝖽𝗀𝖫𝗂𝖾kgr)𝐿superscriptsubscript𝖽𝗀𝖫𝗂𝖾𝑘𝑔𝑟L(\mathsf{dgLie}_{k}^{gr}). By using our enriching functor 𝖽𝗀k𝖬subscript𝖽𝗀𝑘𝖬\mathsf{dg}_{k}\to\mathsf{M}, we get an operad 𝕃ie𝖬gr𝕃𝑖subscriptsuperscript𝑒𝑔𝑟𝖬\mathbb{L}ie^{gr}_{\mathsf{M}} in 𝖬𝖬\mathsf{M}. Taking algebras with respect to this operad, we have 𝖫𝗂𝖾𝖬grsuperscriptsubscript𝖫𝗂𝖾𝖬𝑔𝑟\mathsf{Lie}_{\mathsf{M}}^{gr}, the model category of graded Lie-algebras in 𝖬𝖬\mathsf{M}, and 𝐋𝐢𝐞grsuperscriptsubscript𝐋𝐢𝐞𝑔𝑟\mathbf{Lie}_{\mathcal{M}}^{gr}, the corresponding \infty-category L(𝖫𝗂𝖾𝖬gr)𝐿superscriptsubscript𝖫𝗂𝖾𝖬𝑔𝑟L(\mathsf{Lie}_{\mathsf{M}}^{gr}).

Similarly, for n𝑛n\in\mathbb{Z}, we will denote by nsubscript𝑛\mathbb{P}_{n} the dg-operad (i.e. the operad in 𝖽𝗀ksubscript𝖽𝗀𝑘\mathsf{dg}_{k}) whose algebras are Poisson cdga’s with a bracket of degree (1n)1𝑛(1-n). Recall that, for n>1𝑛1n>1, nsubscript𝑛\mathbb{P}_{n} can be identified with the operad of chains of the topological n𝑛n-little disks operad 𝔼nsubscript𝔼𝑛\mathbb{E}_{n} (see [Coh]). We will also be interested in a graded version of nsubscript𝑛\mathbb{P}_{n}, denoted by ngrsubscriptsuperscript𝑔𝑟𝑛\mathbb{P}^{gr}_{n}: this is an operad in 𝖽𝗀kgrsubscriptsuperscript𝖽𝗀𝑔𝑟𝑘\mathsf{dg}^{gr}_{k}, whose algebras have multiplication of weight (= external grading) 00, and bracket of weight 11-1. The corresponding model categories of algebras will be denoted by n𝖠𝗅𝗀𝖽𝗀k=n𝖼𝖽𝗀𝖺ksubscript𝑛subscript𝖠𝗅𝗀subscript𝖽𝗀𝑘subscript𝑛subscript𝖼𝖽𝗀𝖺𝑘\mathbb{P}_{n}-\mathsf{Alg}_{\mathsf{dg}_{k}}=\mathbb{P}_{n}-\mathsf{cdga}_{k}, and ngr𝖠𝗅𝗀𝖽𝗀k=n𝖼𝖽𝗀𝖺kgrsubscriptsuperscript𝑔𝑟𝑛subscript𝖠𝗅𝗀subscript𝖽𝗀𝑘subscript𝑛subscriptsuperscript𝖼𝖽𝗀𝖺𝑔𝑟𝑘\mathbb{P}^{gr}_{n}-\mathsf{Alg}_{\mathsf{dg}_{k}}=\mathbb{P}_{n}-\mathsf{cdga}^{gr}_{k}; the associated \infty-categories by n𝐀𝐥𝐠𝖽𝗀k=n𝐜𝐝𝐠𝐚ksubscript𝑛subscript𝐀𝐥𝐠subscript𝖽𝗀𝑘subscript𝑛subscript𝐜𝐝𝐠𝐚𝑘\mathbb{P}_{n}-\mathbf{Alg}_{\mathsf{dg}_{k}}=\mathbb{P}_{n}-\mathbf{cdga}_{k}, and ngr𝐀𝐥𝐠𝖽𝗀k=n𝐜𝐝𝐠𝐚kgrsubscriptsuperscript𝑔𝑟𝑛subscript𝐀𝐥𝐠subscript𝖽𝗀𝑘subscript𝑛subscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\mathbb{P}^{gr}_{n}-\mathbf{Alg}_{\mathsf{dg}_{k}}=\mathbb{P}_{n}-\mathbf{cdga}^{gr}_{k}.

By using our enriching functor 𝖽𝗀k𝖬subscript𝖽𝗀𝑘𝖬\mathsf{dg}_{k}\to\mathsf{M}, we thus get operads 𝖬,nsubscript𝖬𝑛\mathbb{P}_{\mathsf{M},n} in 𝖬𝖬\mathsf{M}, and 𝖬,ngrsubscriptsuperscript𝑔𝑟𝖬𝑛\mathbb{P}^{gr}_{\mathsf{M},n} in 𝖬grsuperscript𝖬𝑔𝑟\mathsf{M}^{gr}. Note that 𝖬,ngrsubscriptsuperscript𝑔𝑟𝖬𝑛\mathbb{P}^{gr}_{\mathsf{M},n}-algebras are commutative algebras B𝐵B in 𝖬grsuperscript𝖬𝑔𝑟\mathsf{M}^{gr} endowed with a bracket operation [,]p,q:B(p)𝖬B(q)B(b+q1)[1n]:subscript𝑝𝑞subscripttensor-product𝖬𝐵𝑝𝐵𝑞𝐵𝑏𝑞1delimited-[]1𝑛[-,-]_{p,q}:B(p)\otimes_{\mathsf{M}}B(q)\to B(b+q-1)[1-n], which is a graded bi-derivation, and endows B[n1]𝐵delimited-[]𝑛1B[n-1] with a Lie algebra structure inside 𝖬grsuperscript𝖬𝑔𝑟\mathsf{M}^{gr}. The corresponding \infty-categories L(𝖬,n𝖠𝗅𝗀)𝐿subscript𝖬𝑛𝖠𝗅𝗀L(\mathbb{P}_{\mathsf{M},n}-\mathsf{Alg}) and L(𝖬,ngr𝖠𝗅𝗀)𝐿subscriptsuperscript𝑔𝑟𝖬𝑛𝖠𝗅𝗀L(\mathbb{P}^{gr}_{\mathsf{M},n}-\mathsf{Alg}) of algebras over these operads, will be denoted by ,n𝐀𝐥𝐠subscript𝑛𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},n}-\mathbf{Alg} and ,ngr𝐀𝐥𝐠subscriptsuperscript𝑔𝑟𝑛𝐀𝐥𝐠\mathbb{P}^{gr}_{\mathcal{M},n}-\mathbf{Alg}, respectively.

For A𝖢𝖠𝗅𝗀𝖬superscript𝐴subscript𝖢𝖠𝗅𝗀𝖬A^{\prime}\in\mathsf{CAlg}_{\mathsf{M}}, p0𝑝0p\geq 0, and m𝑚m\in\mathbb{Z}, we define the object (in 𝖬𝖬\mathsf{M}) of m𝑚m-shifted, degree p𝑝p polyvectors 𝖯𝗈𝗅𝖬(A,n)(p)superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛𝑝\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n)(p) on Asuperscript𝐴A^{\prime}, as follows. We consider 𝖳(p)(A,m)𝖬superscript𝖳𝑝superscript𝐴𝑚𝖬\mathsf{T}^{(p)}(A^{\prime},m)\in\mathsf{M} the sub-object of Hom¯𝖬(Ap,A[mp])subscript¯𝐻𝑜𝑚𝖬superscript𝐴superscripttensor-product𝑝superscript𝐴delimited-[]𝑚𝑝\underline{Hom}_{\mathsf{M}}(A^{\prime\otimes^{p}},A^{\prime}[mp]) (where Hom¯𝖬subscript¯𝐻𝑜𝑚𝖬\underline{Hom}_{\mathsf{M}} is the internal Hom-object in 𝖬𝖬\mathsf{M}) consisting of (shifted) multiderivations ([CPTVV, Section 1.4.2.]). There is a natural action of the symmetric group ΣpsubscriptΣ𝑝\Sigma_{p} on Hom¯𝖬(Ap,A[mp])subscript¯𝐻𝑜𝑚𝖬superscript𝐴superscripttensor-product𝑝superscript𝐴delimited-[]𝑚𝑝\underline{Hom}_{\mathsf{M}}(A^{\prime\otimes^{p}},A^{\prime}[mp]) induced by its standard action on Apsuperscript𝐴superscripttensor-product𝑝A^{\prime\otimes^{p}} and by the (1)m×Signsuperscript1𝑚Sign(-1)^{m}\times\mathrm{Sign} action on A[mp]superscript𝐴delimited-[]𝑚𝑝A^{\prime}[mp]. this action restricts to an action on 𝖳(p)(A,m)superscript𝖳𝑝superscript𝐴𝑚\mathsf{T}^{(p)}(A^{\prime},m), and we denote by 𝖳(p)(A,m)Σpsuperscript𝖳𝑝superscriptsuperscript𝐴𝑚subscriptΣ𝑝\mathsf{T}^{(p)}(A^{\prime},m)^{\Sigma_{p}} the 𝖬𝖬\mathsf{M}-object of ΣpsubscriptΣ𝑝\Sigma_{p}-invariants. By standard conventions, we put 𝖳(0)(A,m)=Asuperscript𝖳0superscript𝐴𝑚superscript𝐴\mathsf{T}^{(0)}(A^{\prime},m)=A^{\prime}, and notice that 𝖳(1)(A,m)=Hom¯𝖬(Ω𝖬,A1,A[m])superscript𝖳1superscript𝐴𝑚subscript¯𝐻𝑜𝑚𝖬superscriptsubscriptΩ𝖬superscript𝐴1superscript𝐴delimited-[]𝑚\mathsf{T}^{(1)}(A^{\prime},m)=\underline{Hom}_{\mathsf{M}}(\Omega_{\mathsf{M},A^{\prime}}^{1},A^{\prime}[m]).

Definition 2.1.4

For A𝖢𝖠𝗅𝗀𝖬superscript𝐴subscript𝖢𝖠𝗅𝗀𝖬A^{\prime}\in\mathsf{CAlg}_{\mathsf{M}}, and n𝑛n\in\mathbb{Z}, we define the 𝖬grsuperscript𝖬𝑔𝑟\mathsf{M}^{gr}-object of n𝑛n-shifted polyvectors on Asuperscript𝐴A^{\prime} to be

𝖯𝗈𝗅𝖬(A,n):=p𝖳(p)(A,n)Σp.assignsuperscript𝖯𝗈𝗅𝖬superscript𝐴𝑛subscriptdirect-sum𝑝superscript𝖳𝑝superscriptsuperscript𝐴𝑛subscriptΣ𝑝\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n):=\bigoplus_{p}\mathsf{T}^{(p)}(A^{\prime},-n)^{\Sigma_{p}}.

As in the classical case, there is a “composition” (by insertion) of shifted multiderivations and there is a Schouten-Nijenhuis-like bracket of shifted multiderivations, so that 𝖯𝗈𝗅𝖬(A,n)superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n) becomes a graded n+1subscript𝑛1\mathbb{P}_{n+1}-algebra in 𝖬𝖬\mathsf{M}, i.e. an object in 𝖬,n+1𝖠𝗅𝗀subscript𝖬𝑛1𝖠𝗅𝗀\mathbb{P}_{\mathsf{M},n+1}-\mathsf{Alg}.

Remark 2.1.5

If Asuperscript𝐴A^{\prime} is such that Ω𝖬,A1superscriptsubscriptΩ𝖬superscript𝐴1\Omega_{\mathsf{M},A^{\prime}}^{1} is (strictly) dualizable in A𝖬𝗈𝖽𝖬superscript𝐴subscript𝖬𝗈𝖽𝖬A^{\prime}-\mathsf{Mod}_{\mathsf{M}}, then it is easy to prove that there is an isomorphism 𝖳(p)(A,m)ΣpSymp(T𝖬,A[m])similar-to-or-equalssuperscript𝖳𝑝superscriptsuperscript𝐴𝑚subscriptΣ𝑝𝑆𝑦superscript𝑚𝑝subscriptT𝖬superscript𝐴delimited-[]𝑚\mathsf{T}^{(p)}(A^{\prime},m)^{\Sigma_{p}}\simeq Sym^{p}(\mathrm{T}_{\mathsf{M},A^{\prime}}[-m]) in 𝖬𝖬\mathsf{M}, and more generally 𝖯𝗈𝗅𝖬(A,n)p0Symp(T𝖬,A[m])similar-to-or-equalssuperscript𝖯𝗈𝗅𝖬superscript𝐴𝑛subscriptdirect-sum𝑝0𝑆𝑦superscript𝑚𝑝subscriptT𝖬superscript𝐴delimited-[]𝑚\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n)\simeq\oplus_{p\geq 0}Sym^{p}(\mathrm{T}_{\mathsf{M},A^{\prime}}[-m]) in 𝖬,n+1𝖠𝗅𝗀,subscript𝖬𝑛1𝖠𝗅𝗀\mathbb{P}_{\mathsf{M},n+1}-\mathsf{Alg}, where T𝖬,AsubscriptT𝖬superscript𝐴\mathrm{T}_{\mathsf{M},A^{\prime}} is the Asuperscript𝐴A^{\prime}-dual of Ω𝖬,A1superscriptsubscriptΩ𝖬superscript𝐴1\Omega_{\mathsf{M},A^{\prime}}^{1} (and, on the rhs the multiplication is the one induced by Sym𝑆𝑦𝑚Sym, while the bracket is the one canonically induced by the Lie bracket in T𝖬,AsubscriptT𝖬superscript𝐴\mathrm{T}_{\mathsf{M},A^{\prime}}). Note that, however, we have not used T𝖬,AsubscriptT𝖬superscript𝐴\mathrm{T}_{\mathsf{M},A^{\prime}} directly, in order to define 𝖯𝗈𝗅𝖬(A,n)superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n). We used multiderivations instead, and the the possible definitions agrees only under the hypotheses that Ω𝖬,A1superscriptsubscriptΩ𝖬superscript𝐴1\Omega_{\mathsf{M},A^{\prime}}^{1} is (strictly) dualizable.

The construction A𝖯𝗈𝗅𝖬(A,n)maps-tosuperscript𝐴superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛A^{\prime}\mapsto\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n) is not fully functorial, since we can neither pullback nor pushforward multiderivations in general. However it is possible to define a restricted functoriality ([CPTVV, Lemma 1.4.13]) at the level of \infty-categories, enabling us to give the following

Definition 2.1.6

Let n𝑛n\in\mathbb{Z}, and 𝐂𝐀𝐥𝐠fétsubscriptsuperscript𝐂𝐀𝐥𝐠fét\mathbf{CAlg}^{\textrm{f\'{e}t}}_{\mathcal{M}} the sub \infty-category of 𝐂𝐀𝐥𝐠subscript𝐂𝐀𝐥𝐠\mathbf{CAlg}_{\mathcal{M}} whose maps are only the formally étale ones (i.e. maps AB𝐴𝐵A\to B such that 𝕃B/A0similar-to-or-equalssubscript𝕃𝐵𝐴0\mathbb{L}_{B/A}\simeq 0): Then there is a well-defined \infty-functor

𝐏𝐨𝐥(,n):𝐂𝐀𝐥𝐠fét,n+1𝐀𝐥𝐠,:superscript𝐏𝐨𝐥𝑛subscriptsuperscript𝐂𝐀𝐥𝐠fétsubscript𝑛1𝐀𝐥𝐠\mathbf{Pol}^{\mathcal{M}}(-,n):\mathbf{CAlg}^{\textrm{f\'{e}t}}_{\mathcal{M}}\longrightarrow\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg},

such that if A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}} and Asuperscript𝐴A^{\prime} is a fibrant-cofibrant replacement of A𝐴A inside 𝖢𝖠𝗅𝗀𝖬subscript𝖢𝖠𝗅𝗀𝖬\mathsf{CAlg}_{\mathsf{M}}, then we have an equivalence 𝐏𝐨𝐥(A,n)𝖯𝗈𝗅𝖬(A,n)similar-to-or-equalssuperscript𝐏𝐨𝐥𝐴𝑛superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛\mathbf{Pol}^{\mathcal{M}}(A,n)\simeq\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n) in ,n+1𝐀𝐥𝐠subscript𝑛1𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg}.

Remark 2.1.7

If A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}} is such that 𝕃Asubscriptsuperscript𝕃𝐴\mathbb{L}^{\mathcal{M}}_{A} is dualizable in A𝖬𝗈𝖽𝐴subscript𝖬𝗈𝖽A-\mathsf{Mod}_{\mathcal{M}}, then one deduces from Remark 2.1.5 an equivalence 𝐏𝐨𝐥(A,m)p0Symp(𝕋A[m])similar-to-or-equalssuperscript𝐏𝐨𝐥𝐴𝑚subscriptdirect-sum𝑝0𝑆𝑦superscript𝑚𝑝subscriptsuperscript𝕋𝐴delimited-[]𝑚\mathbf{Pol}^{\mathcal{M}}(A,m)\simeq\oplus_{p\geq 0}Sym^{p}(\mathbb{T}^{\mathcal{M}}_{A}[-m]).

We are now in a position to give the definition of a shifted Poisson structure. Recall that if A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}}, then 𝐏𝐨𝐥(A,n),n+1𝐀𝐥𝐠superscript𝐏𝐨𝐥𝐴𝑛subscript𝑛1𝐀𝐥𝐠\mathbf{Pol}^{\mathcal{M}}(A,n)\in\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg}, so that, in particular, 𝐏𝐨𝐥(A,n)[n]𝐋𝐢𝐞grsuperscript𝐏𝐨𝐥𝐴𝑛delimited-[]𝑛superscriptsubscript𝐋𝐢𝐞𝑔𝑟\mathbf{Pol}^{\mathcal{M}}(A,n)[n]\in\mathbf{Lie}_{\mathcal{M}}^{gr}.

Definition 2.1.8

Let n𝑛n\in\mathbb{Z}, and A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}}. The space of n𝑛n-shifted Poisson structures on A𝐴A is

𝐏𝐨𝐢𝐬𝐬(A,n):=𝖬𝖺𝗉𝐋𝐢𝐞gr(𝟏[1](2),𝐏𝐨𝐥(A,n+1)[n+1]),assign𝐏𝐨𝐢𝐬𝐬𝐴𝑛subscript𝖬𝖺𝗉superscriptsubscript𝐋𝐢𝐞𝑔𝑟subscript1delimited-[]12superscript𝐏𝐨𝐥𝐴𝑛1delimited-[]𝑛1\mathbf{Poiss}(A,n):=\mathsf{Map}_{\mathbf{Lie}_{\mathcal{M}}^{gr}}(\mathbf{1}_{\mathcal{M}}[-1](2),\mathbf{Pol}^{\mathcal{M}}(A,n+1)[n+1]),

where 𝟏[1](2)subscript1delimited-[]12\mathbf{1}_{\mathcal{M}}[-1](2) is the Lie algebra object in \mathcal{M} given by 𝟏subscript1\mathbf{1}_{\mathcal{M}} sitting in pure cohomological degree 111, and pure weight 222, with (strictly) trivial bracket.

This definition mimicks the classical one: if 𝖬=𝖽𝗀k𝖬subscript𝖽𝗀𝑘\mathsf{M}=\mathsf{dg}_{k}, and we replace 𝖬𝖺𝗉𝖬𝖺𝗉\mathsf{Map} by strict Hom𝐻𝑜𝑚Hom-set in 𝖽𝗀𝖫𝗂𝖾kgrsuperscriptsubscript𝖽𝗀𝖫𝗂𝖾𝑘𝑔𝑟\mathsf{dgLie}_{k}^{gr}, in the above definition, we get exactly the set of Poisson bivectors on A𝐴A (the bracket being trivial on 𝟏[1](2)=k[1](2)1delimited-[]12𝑘delimited-[]12\mathbf{1}[-1](2)=k[-1](2), the image of 1k1𝑘1\in k gives a biderivation on π𝜋\pi on A𝐴A, such that [π,π]=0𝜋𝜋0[\pi,\pi]=0).

2.1.3 Standard realizations over k𝑘k

For the theory of shifted Poisson structures on derived stacks that we develop in Section 2.3 we will systematically need to pass to global sections. This procedure can be implemented already at the level of generality of the present section, by considering what we call standard realizations over k𝑘k. A more technical notion of realization over k𝑘k (the Tate realization) will be discussed later on in this Section.

One of our standing hypotheses on the base symmetric monoidal 𝖽𝗀ksubscript𝖽𝗀𝑘\mathsf{dg}_{k} model category 𝖬𝖬\mathsf{M} is that the unit 𝟏𝖬subscript1𝖬\mathbf{1}_{\mathsf{M}} is cofibrant. Hence we get a Quillen adjunction, with 𝟏𝖬\mathbf{1}_{\mathsf{M}}\otimes- left adjoint,

𝟏𝖬:𝖽𝗀k\textstyle{\mathbf{1}_{\mathsf{M}}\otimes-:\mathsf{dg}_{k}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝖬:Hom¯k(𝟏𝖬,):𝖬subscript¯𝐻𝑜𝑚𝑘subscript1𝖬\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\mathsf{M}:\underline{Hom}_{k}(\mathbf{1}_{\mathsf{M}},-)}

where tensor-product\otimes denotes the tensor 𝖽𝗀ksubscript𝖽𝗀𝑘\mathsf{dg}_{k}-enrichment of 𝖬𝖬\mathsf{M}, and Hom¯ksubscript¯𝐻𝑜𝑚𝑘\underline{Hom}_{k} is the Hom-enrichment of 𝖬𝖬\mathsf{M} in 𝖽𝗀ksubscript𝖽𝗀𝑘\mathsf{dg}_{k}. This induces a derived \infty-adjunction on the associated \infty-categories

𝟏𝖬:𝐝𝐠k\textstyle{\mathbf{1}_{\mathsf{M}}\otimes-:\mathbf{dg}_{k}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}:Hom¯k(𝟏𝖬,),:subscript¯𝐻𝑜𝑚𝑘subscript1𝖬\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\mathcal{M}:\mathbb{R}\underline{Hom}_{k}(\mathbf{1}_{\mathsf{M}},-)\,,}

and we define the standard realization functor as the right adjoint ||:=Hom¯k(𝟏𝖬,)|-|:=\mathbb{R}\underline{Hom}_{k}(\mathbf{1}_{\mathsf{M}},-). Actually this is the first of a series of realization functors induced on various categories of algebras in 𝖬𝖬\mathsf{M}. In fact, since 𝟏𝖬subscript1𝖬\mathbf{1}_{\mathsf{M}} is a comonoid object in 𝖬𝖬\mathsf{M} (hence in \mathcal{M}), |||-| is actually a lax symmetric monoidal \infty-functor, hence it is a right adjoint on various functor on categories of “algebras” in \mathcal{M}. Our notation will always be |||-| for each of these induced realization functors. For example, we have

||:𝐂𝐀𝐥𝐠𝐜𝐝𝐠𝐚k,||:𝐂𝐀𝐥𝐠gr𝐜𝐝𝐠𝐚kgr,||:ϵ𝐂𝐀𝐥𝐠ϵ𝐜𝐝𝐠𝐚k,||:ϵ𝐂𝐀𝐥𝐠grϵ𝐜𝐝𝐠𝐚kgr,|-|:\mathbf{CAlg}_{\mathcal{M}}\to\mathbf{cdga}_{k}\,,\,\,\,\,\,|-|:\mathbf{CAlg}^{gr}_{\mathcal{M}}\to\mathbf{cdga}^{gr}_{k}\,,\,\,\,\,\,|-|:\epsilon-\mathbf{CAlg}_{\mathcal{M}}\to\epsilon-\mathbf{cdga}_{k}\,,\,\,\,\,\,|-|:\epsilon-\mathbf{CAlg}^{gr}_{\mathcal{M}}\to\epsilon-\mathbf{cdga}^{gr}_{k}\,,
||:𝐋𝐢𝐞𝐝𝐠𝐋𝐢𝐞k,||:𝐋𝐢𝐞gr𝐝𝐠𝐋𝐢𝐞kgr,||:,n𝐀𝐥𝐠n𝐜𝐝𝐠𝐚k,||:,ngr𝐀𝐥𝐠n𝐜𝐝𝐠𝐚kgr.|-|:\mathbf{Lie}_{\mathcal{M}}\to\mathbf{dgLie}_{k}\,,\,\,\,\,|-|:\mathbf{Lie}^{gr}_{\mathcal{M}}\to\mathbf{dgLie}^{gr}_{k}\,,\,\,\,\,|-|:\mathbb{P}_{\mathcal{M},n}-\mathbf{Alg}\to\mathbb{P}_{n}-\mathbf{cdga}_{k}\,,\,\,\,\,|-|:\mathbb{P}^{gr}_{\mathcal{M},n}-\mathbf{Alg}\to\mathbb{P}_{n}-\mathbf{cdga}^{gr}_{k}.

In particular we get realizations

|𝕃B/A|=:𝕃B/A𝐝𝐠k,|𝐃𝐑(B/A)|=:𝐃𝐑(B/A)ϵ𝐜𝐝𝐠𝐚kgr,|𝐏𝐨𝐥(A,n)|=:𝐏𝐨𝐥(A,n)n+1𝐜𝐝𝐠𝐚kgr.|\mathbb{L}^{\mathcal{M}}_{B/A}|=:\mathbb{L}_{B/A}\in\mathbf{dg}_{k}\,,\,\,\,\,|\mathbf{DR}^{\mathcal{M}}(B/A)|=:\mathbf{DR}(B/A)\in\epsilon-\mathbf{cdga}^{gr}_{k}\,,\,\,\,\,|\mathbf{Pol}^{\mathcal{M}}(A,n)|=:\mathbf{Pol}(A,n)\in\mathbb{P}_{n+1}-\mathbf{cdga}_{k}^{gr}.

Note that when =𝐝𝐠ksubscript𝐝𝐠𝑘\mathcal{M}=\mathbf{dg}_{k}, all this realization functor are (equivalent to) the identity functors.

Later on, we will consider realization functors on category of diagrams in \mathcal{M}, and it will be useful to recall the following fact. The 𝐝𝐠ksubscript𝐝𝐠𝑘\mathbf{dg}_{k}-enriched Hom in a category of diagrams Iopsuperscript𝐼𝑜𝑝I^{op}\to\mathcal{M}, satisfies

Hom¯Iop(𝟏Iop,F)limxIopHom¯(𝟏,F(x)),similar-to-or-equalssubscript¯𝐻𝑜𝑚superscriptsuperscript𝐼𝑜𝑝subscript1superscriptsuperscript𝐼𝑜𝑝𝐹subscript𝑥superscript𝐼𝑜𝑝subscript¯𝐻𝑜𝑚subscript1𝐹𝑥\underline{Hom}_{\mathcal{M}^{I^{op}}}(\mathbf{1}_{\mathcal{M}^{I^{op}}},F)\simeq\lim_{x\in I^{op}}\underline{Hom}_{\mathcal{M}}(\mathbf{1}_{\mathcal{M}},F(x)),

since the monoidal unit 𝟏Iopsubscript1superscriptsuperscript𝐼𝑜𝑝\mathbf{1}_{\mathcal{M}^{I^{op}}} is given by the constant Iopsuperscript𝐼𝑜𝑝I^{op}-diagram at 𝟏subscript1\mathbf{1}_{\mathcal{M}}. This observation is the reason for the appearance, in Thm 2.2.12, of (derived) global sections of various algebras over the derived affine site of a derived stack.

Example 2.1.9

As examples of the use of standard realization, by definition of |||-| as a right adjoint, we have, for any A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}}, the following equivalences in 𝐓𝐓\mathbf{T}

𝐏𝐨𝐢𝐬𝐬(A,n)𝖬𝖺𝗉𝐝𝐠𝐋𝐢𝐞kgr(k[1](2),𝐏𝐨𝐥(A,n+1)[n+1]),similar-to-or-equals𝐏𝐨𝐢𝐬𝐬𝐴𝑛subscript𝖬𝖺𝗉superscriptsubscript𝐝𝐠𝐋𝐢𝐞𝑘𝑔𝑟𝑘delimited-[]12𝐏𝐨𝐥𝐴𝑛1delimited-[]𝑛1\mathbf{Poiss}(A,n)\simeq\mathsf{Map}_{\mathbf{dgLie}_{k}^{gr}}(k[-1](2),\mathbf{Pol}(A,n+1)[n+1]),
𝒜p,cl(A,n)𝖬𝖺𝗉ϵ𝐝𝐠kgr(k(p)[pn],𝐃𝐑(A)),𝒜p(A,n)𝖬𝖺𝗉𝐝𝐠k(k[n],|A|p𝕃A).formulae-sequencesimilar-to-or-equalssuperscriptsubscript𝒜𝑝𝑐𝑙𝐴𝑛subscript𝖬𝖺𝗉italic-ϵsuperscriptsubscript𝐝𝐠𝑘𝑔𝑟𝑘𝑝delimited-[]𝑝𝑛𝐃𝐑𝐴similar-to-or-equalssuperscriptsubscript𝒜𝑝𝐴𝑛subscript𝖬𝖺𝗉subscript𝐝𝐠𝑘𝑘delimited-[]𝑛superscriptsubscript𝐴𝑝subscript𝕃𝐴\mathcal{A}_{\mathcal{M}}^{p,cl}(A,n)\simeq\mathsf{Map}_{\epsilon-\mathbf{dg}_{k}^{gr}}(k(p)[-p-n],\mathbf{DR}(A))\,,\,\,\,\,\,\mathcal{A}_{\mathcal{M}}^{p}(A,n)\simeq\mathsf{Map}_{\mathbf{dg}_{k}}(k[-n],\wedge_{|A|}^{p}\mathbb{L}_{A}).

2.1.4 Comparison between Poisson algebras and shifted Poisson pairs

We need a few notations. For an arbitrary \infty-category 𝐂𝐂\mathbf{C}, (𝐂)𝐂\mathcal{I}(\mathbf{C}) will denote its moduli space or maximal \infty-subgroupoid of equivalences (so that (𝐂)𝐓𝐂𝐓\mathcal{I}(\mathbf{C})\in\mathbf{T}), 𝐂Δ[1]superscript𝐂Δdelimited-[]1\mathbf{C}^{\Delta[1]} the \infty-category of morphisms in 𝐂𝐂\mathbf{C}, and ev0,ev1:𝐂Δ[1]𝐂:subscriptev0subscriptev1superscript𝐂Δdelimited-[]1𝐂\textrm{ev}_{0},\textrm{ev}_{1}:\mathbf{C}^{\Delta[1]}\to\mathbf{C} the source and target \infty-functors. For a model category 𝖢𝖢\mathsf{C}, we denote by 𝖢cfWsubscriptsuperscript𝖢𝑊𝑐𝑓\mathsf{C}^{W}_{cf} the category of cofibrant-fibrant objects in 𝖢𝖢\mathsf{C} with morphisms given by weak equivalences.

The nerve of 𝖢cfWsubscriptsuperscript𝖢𝑊𝑐𝑓\mathsf{C}^{W}_{cf} is called the moduli space or classifying space (of objects and equivalences) of 𝖢𝖢\mathsf{C}. It is a general fact (going back essentially to Dwyer and Kan) that we have an equivalence of spaces between the moduli space of 𝖢𝖢\mathsf{C} and the space (𝐂)𝐂\mathcal{I}(\mathbf{C}). In other words, we may (and will) identify the moduli space of 𝖢𝖢\mathsf{C} with the moduli space of the associated \infty-category 𝐂𝐂\mathbf{C}. We will be interested, below, in the moduli spaces of the \infty-categories ,n𝐀𝐥𝐠subscript𝑛𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},n}-\mathbf{Alg}, 𝐋𝐢𝐞grsuperscriptsubscript𝐋𝐢𝐞𝑔𝑟\mathbf{Lie}_{\mathcal{M}}^{gr}, and 𝐂𝐀𝐥𝐠subscript𝐂𝐀𝐥𝐠\mathbf{CAlg}_{\mathcal{M}}, associated to the model categories 𝖬,n+1𝖠𝗅𝗀subscript𝖬𝑛1𝖠𝗅𝗀\mathbb{P}_{\mathsf{M},n+1}-\mathsf{Alg}, 𝖫𝗂𝖾𝖬grsuperscriptsubscript𝖫𝗂𝖾𝖬𝑔𝑟\mathsf{Lie}_{\mathsf{M}}^{gr}, and 𝖢𝖠𝗅𝗀𝖬subscript𝖢𝖠𝗅𝗀𝖬\mathsf{CAlg}_{\mathsf{M}}, respectively.

For A,n𝐀𝐥𝐠𝐴subscript𝑛𝐀𝐥𝐠A\in\mathbb{P}_{\mathcal{M},n}-\mathbf{Alg}, we will write ,n(A)subscript𝑛𝐴\mathbb{P}_{\mathcal{M},n}(A) for the space of nsubscript𝑛\mathbb{P}_{n} algebra structures on A𝐴A having the given underlying commutative algebra structure. More precisely, there is a forgetful \infty-functor U:,n𝐀𝐥𝐠𝐂𝐀𝐥𝐠:𝑈subscript𝑛𝐀𝐥𝐠subscript𝐂𝐀𝐥𝐠U:\mathbb{P}_{\mathcal{M},n}-\mathbf{Alg}\to\mathbf{CAlg}_{\mathcal{M}} (forgetting the bracket structure), and the fiber of U𝑈U at a given A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}} is an \infty-groupoid, i.e. a space, that we denote by ,n(A)subscript𝑛𝐴\mathbb{P}_{\mathcal{M},n}(A).

We have a canonical \infty-functor

v:(𝐂𝐀𝐥𝐠)A(𝟏[1](2),𝐏𝐨𝐥(A,n)[n])𝐋𝐢𝐞𝐀𝐥𝐠gr×𝐋𝐢𝐞𝐀𝐥𝐠gr:𝑣containssubscript𝐂𝐀𝐥𝐠𝐴1delimited-[]12superscript𝐏𝐨𝐥𝐴𝑛delimited-[]𝑛superscriptsubscript𝐋𝐢𝐞𝐀𝐥𝐠𝑔𝑟superscriptsubscript𝐋𝐢𝐞𝐀𝐥𝐠𝑔𝑟v:\mathcal{I}(\mathbf{CAlg}_{\mathcal{M}})\ni A\longmapsto(\mathbf{1}[-1](2),\mathbf{Pol}^{\mathcal{M}}(A,n)[n])\in\mathbf{LieAlg}_{\mathcal{M}}^{gr}\times\mathbf{LieAlg}_{\mathcal{M}}^{gr}

(note that A𝐏𝐨𝐥(A,n)maps-to𝐴superscript𝐏𝐨𝐥𝐴𝑛A\mapsto\mathbf{Pol}^{\mathcal{M}}(A,n) is functorial with respect to equivalences in 𝐂𝐀𝐥𝐠subscript𝐂𝐀𝐥𝐠\mathbf{CAlg}_{\mathcal{M}}, since obviously equivalences are formally étale). Borrowing ideas from [Vez2, 3.1], we define the space 𝐏𝐨𝐢𝐬𝐬,neqsubscriptsuperscript𝐏𝐨𝐢𝐬𝐬eq𝑛\mathbf{Poiss}^{\mathrm{eq}}_{\mathcal{M},n} as the following pullback of in 𝐓𝐓\mathbf{T}

𝐏𝐨𝐢𝐬𝐬,neqsubscriptsuperscript𝐏𝐨𝐢𝐬𝐬eq𝑛\textstyle{\mathbf{Poiss}^{\mathrm{eq}}_{\mathcal{M},n}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}q𝑞\scriptstyle{q}((𝐋𝐢𝐞𝐀𝐥𝐠gr)Δ[1])superscriptsuperscriptsubscript𝐋𝐢𝐞𝐀𝐥𝐠𝑔𝑟Δdelimited-[]1\textstyle{\mathcal{I}((\mathbf{LieAlg}_{\mathcal{M}}^{gr})^{\Delta[1]})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}((ev0),(ev1))subscriptev0subscriptev1\scriptstyle{(\mathcal{I}(\mathrm{ev}_{0}),\mathcal{I}(\mathrm{ev}_{1}))}(𝐂𝐀𝐥𝐠)subscript𝐂𝐀𝐥𝐠\textstyle{\mathcal{I}(\mathbf{CAlg}_{\mathcal{M}})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}(v)𝑣\scriptstyle{\mathcal{I}(v)}(𝐋𝐢𝐞𝐀𝐥𝐠gr×𝐋𝐢𝐞𝐀𝐥𝐠gr).superscriptsubscript𝐋𝐢𝐞𝐀𝐥𝐠𝑔𝑟superscriptsubscript𝐋𝐢𝐞𝐀𝐥𝐠𝑔𝑟\textstyle{\mathcal{I}(\mathbf{LieAlg}_{\mathcal{M}}^{gr}\times\mathbf{LieAlg}_{\mathcal{M}}^{gr})\,.}

Thus, informally speaking, 𝐏𝐨𝐢𝐬𝐬,neqsubscriptsuperscript𝐏𝐨𝐢𝐬𝐬eq𝑛\mathbf{Poiss}^{\mathrm{eq}}_{\mathcal{M},n} is the moduli space of n𝑛n-shifted Poisson pairs (A,π)𝐴𝜋(A,\pi), consisting of a commutative algebra A𝐴A in \mathcal{M} together with a n𝑛n-shifted Poisson structure on A𝐴A (see Definition 2.1.8). More precisely, let 𝖯𝗈𝗂𝗌𝗌𝖬,neqsubscriptsuperscript𝖯𝗈𝗂𝗌𝗌eq𝖬𝑛\mathsf{Poiss}^{\textrm{eq}}_{\mathsf{M},n} be the category whose objects are pairs (A,π)𝐴𝜋(A,\pi) where A𝐴A is a fibrant-cofibrant object in 𝖢𝖠𝗅𝗀𝖬subscript𝖢𝖠𝗅𝗀𝖬\mathsf{CAlg}_{\mathsf{M}}, and π𝜋\pi is a n𝑛n-shifted Poisson structure on A𝐴A, i.e a map 𝟏𝖬[1](2)𝖯𝗈𝗅𝖬(A,n+1)[n+1]subscript1𝖬delimited-[]12superscript𝖯𝗈𝗅𝖬𝐴𝑛1delimited-[]𝑛1\mathbf{1}_{\mathsf{M}}[-1](2)\to\mathsf{Pol}^{\mathsf{M}}(A,n+1)[n+1] in the homotopy category of 𝐋𝐢𝐞𝐀𝐥𝐠𝖬grsuperscriptsubscript𝐋𝐢𝐞𝐀𝐥𝐠𝖬𝑔𝑟\mathbf{LieAlg}_{\mathsf{M}}^{gr}, and whose morphisms (A,π)(A,π)𝐴𝜋superscript𝐴superscript𝜋(A,\pi)\to(A^{\prime},\pi^{\prime}) are weak equivalences u:AA:𝑢𝐴superscript𝐴u:A\to A^{\prime} in 𝖢𝖠𝗅𝗀𝖬subscript𝖢𝖠𝗅𝗀𝖬\mathsf{CAlg}_{\mathsf{M}} such that the diagram

𝖯𝗈𝗅𝖬(A,n+1)[n+1]superscript𝖯𝗈𝗅𝖬𝐴𝑛1delimited-[]𝑛1\textstyle{\mathsf{Pol}^{\mathsf{M}}(A,n+1)[n+1]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}𝖯𝗈𝗅𝖬(u,n+1)[n+1]superscript𝖯𝗈𝗅𝖬𝑢𝑛1delimited-[]𝑛1\scriptstyle{\mathsf{Pol}^{\mathsf{M}}(u,n+1)[n+1]}𝟏𝖬[1](2)subscript1𝖬delimited-[]12\textstyle{\mathbf{1}_{\mathsf{M}}[-1](2)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}π𝜋\scriptstyle{\pi}πsuperscript𝜋\scriptstyle{\pi^{\prime}}𝖯𝗈𝗅𝖬(A,n+1)[n+1]superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛1delimited-[]𝑛1\textstyle{\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n+1)[n+1]}

is commutative in the homotopy category of 𝖫𝗂𝖾𝖠𝗅𝗀𝖬grsuperscriptsubscript𝖫𝗂𝖾𝖠𝗅𝗀𝖬𝑔𝑟\mathsf{LieAlg}_{\mathsf{M}}^{gr}. Then, the nerve of 𝖯𝗈𝗂𝗌𝗌𝖬,neqsubscriptsuperscript𝖯𝗈𝗂𝗌𝗌eq𝖬𝑛\mathsf{Poiss}^{\mathrm{eq}}_{\mathsf{M},n} is equivalent to 𝐏𝐨𝐢𝐬𝐬,neqsubscriptsuperscript𝐏𝐨𝐢𝐬𝐬eq𝑛\mathbf{Poiss}^{\mathrm{eq}}_{\mathcal{M},n}.

There is a (strict) functor w:𝖬,n+1𝐀𝐥𝐠cfW𝖯𝗈𝗂𝗌𝗌𝖬,neq:𝑤subscript𝖬𝑛1superscriptsubscript𝐀𝐥𝐠𝑐𝑓𝑊subscriptsuperscript𝖯𝗈𝗂𝗌𝗌eq𝖬𝑛w:\mathbb{P}_{\mathsf{M},n+1}-\mathbf{Alg}_{cf}^{W}\to\mathsf{Poiss}^{\mathrm{eq}}_{\mathsf{M},n}, sending a strict n+1subscript𝑛1\mathbb{P}_{n+1}-algebra B𝐵B in 𝖬𝖬\mathsf{M} to the pair (B,π)𝐵𝜋(B,\pi), where π𝜋\pi is induced, in the standard way, by the (strict) Lie bracket on B𝐵B (since the bracket is strict, it is a strict biderivation on B𝐵B). Restriction to weak equivalences (between cofibrant-fibrant objects) in 𝖬,n+1𝐀𝐥𝐠subscript𝖬𝑛1𝐀𝐥𝐠\mathbb{P}_{\mathsf{M},n+1}-\mathbf{Alg} ensures this is a functor, and note that objects in the image of w𝑤w are, by definition, strict pairs (B,π)𝐵𝜋(B,\pi), i.e. the shifted Poisson structure π:𝟏𝖬[1](2)𝖯𝗈𝗅𝖬(A,n+1)[n+1]:𝜋subscript1𝖬delimited-[]12superscript𝖯𝗈𝗅𝖬𝐴𝑛1delimited-[]𝑛1\pi:\mathbf{1}_{\mathsf{M}}[-1](2)\to\mathsf{Pol}^{\mathsf{M}}(A,n+1)[n+1] is an actual morphism in 𝐋𝐢𝐞𝐀𝐥𝐠𝖬grsuperscriptsubscript𝐋𝐢𝐞𝐀𝐥𝐠𝖬𝑔𝑟\mathbf{LieAlg}_{\mathsf{M}}^{gr} (rather than a map in its homotopy category). We have a commutative diagram of functors

𝖬,n+1𝐀𝐥𝐠cfWsubscript𝖬𝑛1superscriptsubscript𝐀𝐥𝐠𝑐𝑓𝑊\textstyle{\mathbb{P}_{\mathsf{M},n+1}-\mathbf{Alg}_{cf}^{W}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}u𝑢\scriptstyle{u}w𝑤\scriptstyle{w}𝖯𝗈𝗂𝗌𝗌𝖬,neqsubscriptsuperscript𝖯𝗈𝗂𝗌𝗌eq𝖬𝑛\textstyle{\mathsf{Poiss}^{\mathrm{eq}}_{\mathsf{M},n}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}q𝑞\scriptstyle{q}(𝖢𝖠𝗅𝗀𝖬)cfWsuperscriptsubscriptsubscript𝖢𝖠𝗅𝗀𝖬𝑐𝑓𝑊\textstyle{(\mathsf{CAlg}_{\mathsf{M}})_{cf}^{W}}

where u𝑢u forgets the bracket structure, and q𝑞q is the functor (A,π)Amaps-to𝐴𝜋𝐴(A,\pi)\mapsto A. Taking the nerves of the previous diagram, we get a commutative diagram in 𝐓𝐓\mathbf{T} (where we have kept the same name for the maps)

(,n+1𝐀𝐥𝐠)subscript𝑛1𝐀𝐥𝐠\textstyle{\mathcal{I}(\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}u𝑢\scriptstyle{u}w𝑤\scriptstyle{w}𝐏𝐨𝐢𝐬𝐬,neqsubscriptsuperscript𝐏𝐨𝐢𝐬𝐬eq𝑛\textstyle{\mathbf{Poiss}^{\mathrm{eq}}_{\mathcal{M},n}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}q𝑞\scriptstyle{q}(𝐂𝐀𝐥𝐠)subscript𝐂𝐀𝐥𝐠\textstyle{\mathcal{I}(\mathbf{CAlg}_{\mathcal{M}})}

Note that u𝑢u, and q𝑞q are both surjective, since they both have a section given by the trivial bracket or strict shifted Poisson structure. Moreover, the fiber of q𝑞q at A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}} is obviously equivalent to the space 𝖯𝗈𝗂𝗌𝗌(A,n)subscript𝖯𝗈𝗂𝗌𝗌𝐴𝑛\mathsf{Poiss}_{\mathcal{M}}(A,n) of n𝑛n-shifted Poisson structures on A𝐴A (Definition 2.1.8). The following is a slight enhancement of [Me, Theorem 3.2].

Theorem 2.1.10

The map of spaces w:(,n+1𝐀𝐥𝐠)𝐏𝐨𝐢𝐬𝐬,neq:𝑤subscript𝑛1𝐀𝐥𝐠subscriptsuperscript𝐏𝐨𝐢𝐬𝐬eq𝑛w:\mathcal{I}(\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg})\to\mathbf{Poiss}^{\mathrm{eq}}_{\mathcal{M},n} is an equivalence.

Proof. It is enough to prove that for any fibrant-cofibrant A𝖢𝖠𝗅𝗀𝖬𝐴subscript𝖢𝖠𝗅𝗀𝖬A\in\mathsf{CAlg}_{\mathsf{M}}, the map induced by w𝑤w between the u𝑢u and q𝑞q fibers over A𝐴A is an equivalence. But, since these fibers are, by definition, ,n+1(A)subscript𝑛1𝐴\mathbb{P}_{\mathcal{M},n+1}(A) (the moduli space of nsubscript𝑛\mathbb{P}_{n}-algebra structures on A𝐴A having the given underlying commutative algebra structure), and 𝖯𝗈𝗂𝗌𝗌(A,n)subscript𝖯𝗈𝗂𝗌𝗌𝐴𝑛\mathsf{Poiss}_{\mathcal{M}}(A,n) (the space of n𝑛n-shifted Poisson structure on A)A), this is exactly [Me, Theorem 3.2]. \Box

For future reference, we also state the following immediate consequence.

Corollary 2.1.11

(Melani) For any A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}}, the map w𝑤w of Theorem 2.1.10 induces an equivalence wA:,n+1(A)𝖯𝗈𝗂𝗌𝗌(A,n):subscript𝑤𝐴similar-to-or-equalssubscript𝑛1𝐴subscript𝖯𝗈𝗂𝗌𝗌𝐴𝑛w_{A}:\mathbb{P}_{\mathcal{M},n+1}(A)\simeq\mathsf{Poiss}_{\mathcal{M}}(A,n) between the moduli space of nsubscript𝑛\mathbb{P}_{n}-algebra structures on A𝐴A (having the given underlying commutative algebra structure), and the moduli space of n𝑛n-shifted Poisson structures on A𝐴A.

Proof. As mentioned above, this is exactly [Me, Theorem 3.2]. \Box

Remark 2.1.12

(1) Theorem 3.2 in [Me] is stated for the model category 𝖬𝖬\mathsf{M} of bounded above cochain complexes of k𝑘k-modules, but the proof is general and it extends immediately to our general 𝖬𝖬\mathsf{M}. The original statement seems moreover to require a restriction to those cdga’s having a dualizable cotangent complex. This is due to the fact that the author uses the tangent complex (i.e. the dual of the cotangent complex) in order to identify derivations. However, the actual proof produces an equivalence between (weak, shifted) Lie brackets and (weak) biderivations. Therefore if one identifies derivations using the linear dual of the symmetric algebra of the cotangent complex, the need to pass to the tangent complex disappears, and the result holds, with the same proof and without the assumption of the cotangent complex being dualizable (see Remark 2.1.7). This is the main reason we adopted Definition 2.1.4 and 2.1.6 as our definition of polyvectors.

(2) Melani’s proof of [Me, Theorem 3.2] also shows that the natural map from the classifying space (,n+1𝐀𝐥𝐠)subscript𝑛1𝐀𝐥𝐠\mathcal{I}(\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg}) of strict n+1subscript𝑛1\mathbb{P}_{n+1}-algebras in 𝖬𝖬\mathsf{M} to the classifying space (,n+1,𝐀𝐥𝐠)subscript𝑛1𝐀𝐥𝐠\mathcal{I}(\mathbb{P}_{\mathcal{M},n+1,\infty}-\mathbf{Alg}) of weak n+1subscript𝑛1\mathbb{P}_{n+1}-algebras in 𝖬𝖬\mathsf{M} (where the operad ,n+1,subscript𝑛1\mathbb{P}_{\mathcal{M},n+1,\infty} is any cofibrant resolution of ,n+1subscript𝑛1\mathbb{P}_{\mathcal{M},n+1} in the model category of operads in \mathcal{M}) is an equivalence.

We also have the following, easier, strict analog of Corollary 2.1.11.

Proposition 2.1.13

[CPTVV, Proposition 1.4.8] For any A𝖢𝖠𝗅𝗀𝖬𝐴subscript𝖢𝖠𝗅𝗀𝖬A\in\mathsf{CAlg}_{\mathsf{M}}, there is a natural bijection

𝖬,n(A)HomLie𝖬gr(𝟏𝖬(2)[1],𝖯𝗈𝗅𝖬(A,n)[n])similar-to-or-equalssubscript𝖬𝑛𝐴𝐻𝑜subscript𝑚𝐿𝑖superscriptsubscript𝑒𝖬𝑔𝑟subscript1𝖬2delimited-[]1superscript𝖯𝗈𝗅𝖬𝐴𝑛delimited-[]𝑛\mathbb{P}_{\mathsf{M},n}(A)\simeq Hom_{Lie_{\mathsf{M}}^{gr}}(\mathbf{1}_{\mathsf{M}}(2)[-1],\mathsf{Pol}^{\mathsf{M}}(A,n)[n])

between the set of (strict) nsubscript𝑛\mathbb{P}_{n}-algebra structures on A𝐴A in 𝖬𝖬\mathsf{M}, and the set of morphisms 𝟏(2)[1]𝖯𝗈𝗅𝖬(A,n)[n]12delimited-[]1superscript𝖯𝗈𝗅𝖬𝐴𝑛delimited-[]𝑛\mathbf{1}(2)[-1]\to\mathsf{Pol}^{\mathsf{M}}(A,n)[n] of Lie algebra objects in 𝖬grsuperscript𝖬𝑔𝑟\mathsf{M}^{gr}.

As an immediate consequence of Theorem 2.1.10, we get the following, useful, strictification result

Corollary 2.1.14

Any weak shifted Poisson pair in 𝖯𝗈𝗂𝗌𝗌𝖬,neqsubscriptsuperscript𝖯𝗈𝗂𝗌𝗌eq𝖬𝑛\mathsf{Poiss}^{\textrm{eq}}_{\mathsf{M},n} is equivalent to a strict pair.

Proof. By Theorem 2.1.10, an object (A,π)𝖯𝗈𝗂𝗌𝗌𝖬,neq𝐴𝜋subscriptsuperscript𝖯𝗈𝗂𝗌𝗌eq𝖬𝑛(A,\pi)\in\mathsf{Poiss}^{\textrm{eq}}_{\mathsf{M},n} (i.e. an a priori weak pair), is equivalent to a pair of the form w(B)𝑤𝐵w(B), where B𝖬,n+1𝐀𝐥𝐠𝐵subscript𝖬𝑛1𝐀𝐥𝐠B\in\mathbb{P}_{\mathsf{M},n+1}-\mathbf{Alg} (i.e. is a strict n+1subscript𝑛1\mathbb{P}_{n+1}-algebra in 𝖬𝖬\mathsf{M}), whose underlying commutative algebra is weakly equivalent to A𝐴A in 𝖢𝖠𝗅𝗀𝖬subscript𝖢𝖠𝗅𝗀𝖬\mathsf{CAlg}_{\mathsf{M}}. We conclude by observing that objects in the image of w𝑤w are always strict pairs. \Box

The 𝐃𝐑𝐃𝐑\mathbf{DR}-to-𝐏𝐨𝐥𝐏𝐨𝐥\mathbf{Pol} construction. Let A𝖬,n𝐀𝐥𝐠.superscript𝐴subscript𝖬𝑛𝐀𝐥𝐠A^{\prime}\in\mathbb{P}_{\mathsf{M},n}-\mathbf{Alg}. By Proposition 2.1.13, the nsubscript𝑛\mathbb{P}_{n} algebra structure on the underlying commutative algebra of Asuperscript𝐴A^{\prime} is encoded by a strict map π:𝟏𝖬[1](2)𝖯𝗈𝗅𝖬(A,n)[n]:𝜋subscript1𝖬delimited-[]12superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛delimited-[]𝑛\pi:\mathbf{1}_{\mathsf{M}}[-1](2)\to\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n)[n] of graded Lie algebras in 𝖬𝖬\mathsf{M}. Since the weight q𝑞q part of 𝖯𝗈𝗅𝖬(A,n)superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n) is, by definition, 𝖳(q)(A,n)Σqsuperscript𝖳𝑞superscriptsuperscript𝐴𝑛subscriptΣ𝑞\mathsf{T}^{(q)}(A^{\prime},-n)^{\Sigma_{q}}, π𝜋\pi induces a map 𝟏M𝖳(2)(A,n)Σ2[n+1]subscript1𝑀superscript𝖳2superscriptsuperscript𝐴𝑛subscriptΣ2delimited-[]𝑛1\mathbf{1}_{M}\to\mathsf{T}^{(2)}(A^{\prime},-n)^{\Sigma_{2}}[n+1] that we still denote by π𝜋\pi. Write

[,]p,q:𝖯𝗈𝗅𝖬(A,n)(p)𝖬𝖯𝗈𝗅𝖬(A,n)(q)𝖯𝗈𝗅𝖬(A,n)(p+q1)[n]:subscript𝑝𝑞subscripttensor-product𝖬superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛𝑝superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛𝑞superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛𝑝𝑞1delimited-[]𝑛[-,-]_{p,q}:\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n)(p)\otimes_{\mathsf{M}}\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n)(q)\to\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n)(p+q-1)[-n]

for the Lie bracket part of the graded n+1subscript𝑛1\mathbb{P}_{n+1}-algebra structure on 𝖯𝗈𝗅𝖬(A,n)superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n), the family of composite maps in 𝖬𝖬\mathsf{M}

𝖳(q)(A,n)superscript𝖳𝑞superscript𝐴𝑛\textstyle{\mathsf{T}^{(q)}(A^{\prime},-n)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϵqsubscriptitalic-ϵ𝑞\scriptstyle{\epsilon_{q}}similar-to-or-equals\textstyle{\simeq}𝟏M𝖬𝖳(q)(A,n)subscripttensor-product𝖬subscript1𝑀superscript𝖳𝑞superscript𝐴𝑛\textstyle{\mathbf{1}_{M}\otimes_{\mathsf{M}}\mathsf{T}^{(q)}(A^{\prime},-n)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}πidtensor-product𝜋id\scriptstyle{\pi\otimes\mathrm{id}}𝖳(2)(A,n)Σ2[n+1]𝖬𝖳(q)(A,n)Σqsubscripttensor-product𝖬superscript𝖳2superscriptsuperscript𝐴𝑛subscriptΣ2delimited-[]𝑛1superscript𝖳𝑞superscriptsuperscript𝐴𝑛subscriptΣ𝑞\textstyle{\mathsf{T}^{(2)}(A^{\prime},-n)^{\Sigma_{2}}[n+1]\otimes_{\mathsf{M}}\mathsf{T}^{(q)}(A^{\prime},-n)^{\Sigma_{q}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}[,]2,q[n+1]subscript2𝑞delimited-[]𝑛1\scriptstyle{[-,-]_{2,q}[n+1]}𝖳(q+1)(A,n)Σq+1[1],superscript𝖳𝑞1superscriptsuperscript𝐴𝑛subscriptΣ𝑞1delimited-[]1\textstyle{\mathsf{T}^{(q+1)}(A^{\prime},-n)^{\Sigma_{q+1}}[1],}

for q𝑞q\in\mathbb{N}, is easily verified to be the mixed differential of a mixed graded algebra structure on 𝖯𝗈𝗅𝖬(A,n)superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n) inside 𝖬𝖬\mathsf{M}.
Moreover, since 𝖯𝗈𝗅𝖬(A,n)(0)=Asuperscript𝖯𝗈𝗅𝖬superscript𝐴𝑛0superscript𝐴\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n)(0)=A^{\prime}, by the universal property of 𝖣𝖱𝖬(A)superscript𝖣𝖱𝖬superscript𝐴\mathsf{DR}^{\mathsf{M}}(A^{\prime}) the identity AAsuperscript𝐴superscript𝐴A^{\prime}\to A^{\prime} induces a map

ϕA,π:𝖣𝖱𝖬(A)𝖯𝗈𝗅𝖬(A,n):subscriptitalic-ϕsuperscript𝐴𝜋superscript𝖣𝖱𝖬superscript𝐴superscript𝖯𝗈𝗅𝖬superscript𝐴𝑛\phi_{A^{\prime},\pi}:\mathsf{DR}^{\mathsf{M}}(A^{\prime})\longrightarrow\mathsf{Pol}^{\mathsf{M}}(A^{\prime},n)

of mixed graded algebras in 𝖬𝖬\mathsf{M}.

Remark 2.1.15

The above construction of the mixed differential ϵ=(ϵq)qitalic-ϵsubscriptsubscriptitalic-ϵ𝑞𝑞\epsilon=(\epsilon_{q})_{q} is a generalization of the classical construction associating to a classical Poisson bivector π𝜋\pi on a smooth manifold, the differential dπ:=[π,]assignsubscript𝑑𝜋𝜋d_{\pi}:=[\pi,-] on polyvectors, where [,][-,-] is the Schouten-Nijenhuis bracket, and dπ2=0subscriptsuperscript𝑑2𝜋0d^{2}_{\pi}=0 is equivalent to the bivector π𝜋\pi being Poisson.

A slight elaboration of this construction (by choosing strict models, see [CPTVV, 1.4.3] for details), yields the following derived version

  • functors 𝐃𝐑,𝐏𝐨𝐥(,n):(,n+1𝐀𝐥𝐠)eqϵ𝐂𝐀𝐥𝐠gr:superscript𝐃𝐑superscript𝐏𝐨𝐥𝑛superscriptsubscript𝑛1𝐀𝐥𝐠𝑒𝑞italic-ϵsuperscriptsubscript𝐂𝐀𝐥𝐠𝑔𝑟\mathbf{DR}^{\mathcal{M}},\mathbf{Pol}^{\mathcal{M}}(-,n):(\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg})^{eq}\to\epsilon-\mathbf{CAlg}_{\mathcal{M}}^{gr}

  • a morphism Φ:𝐃𝐑𝐏𝐨𝐥(,n):Φsuperscript𝐃𝐑superscript𝐏𝐨𝐥𝑛\Phi:\mathbf{DR}^{\mathcal{M}}\to\mathbf{Pol}^{\mathcal{M}}(-,n) betwen the above functors.

Here (,n+1𝐀𝐥𝐠)eqsuperscriptsubscript𝑛1𝐀𝐥𝐠𝑒𝑞(\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg})^{eq} is the \infty-category of n+1subscript𝑛1\mathbb{P}_{n+1}-algebras in \mathcal{M} with only equivalences as morphisms (this ensures that 𝐏𝐨𝐥(,n)superscript𝐏𝐨𝐥𝑛\mathbf{Pol}^{\mathcal{M}}(-,n) is indeed a functor), and we slightly abuse notation by writing 𝐃𝐑superscript𝐃𝐑\mathbf{DR}^{\mathcal{M}} for the composition of the usual 𝐃𝐑superscript𝐃𝐑\mathbf{DR}^{\mathcal{M}} with the forgetful functor (,n+1𝐀𝐥𝐠)eq𝐂𝐀𝐥𝐠superscriptsubscript𝑛1𝐀𝐥𝐠𝑒𝑞subscript𝐂𝐀𝐥𝐠(\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg})^{eq}\to\mathbf{CAlg}_{\mathcal{M}}.

2.1.5 From non-degenerate Poisson algebra structures to symplectic structures

Classically, one way of stating that a Poisson structure (on a smooth scheme or manifold X𝑋X) is non degenerate is by declaring that the analog of the above map ΦΦ\Phi establishes an isomorphism of mixed graded algebras between the de Rham algebra and the algebra of polyvectors. Analogously, we say that A,n+1𝐀𝐥𝐠𝐴subscript𝑛1𝐀𝐥𝐠A\in\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg} is non degenerate if ΦA:𝐃𝐑(A)𝐏𝐨𝐥(A,n):subscriptΦ𝐴superscript𝐃𝐑𝐴superscript𝐏𝐨𝐥𝐴𝑛\Phi_{A}:\mathbf{DR}^{\mathcal{M}}(A)\to\mathbf{Pol}^{\mathcal{M}}(A,n) is an equivalence of mixed graded algebras in \mathcal{M}. For such an A𝐴A, by Corollary 2.1.11, we also get a map αA:𝟏(2)𝐏𝐨𝐥(A,n)[n+1]:subscript𝛼𝐴subscript12superscript𝐏𝐨𝐥𝐴𝑛delimited-[]𝑛1\alpha_{A}:\mathbf{1}_{\mathcal{M}}(2)\to\mathbf{Pol}^{\mathcal{M}}(A,n)[n+1] in ϵ𝐌gritalic-ϵsuperscript𝐌𝑔𝑟\epsilon-\mathbf{M}^{gr}. By putting these together, we get a diagram

𝐃𝐑(A)[n+1]superscript𝐃𝐑𝐴delimited-[]𝑛1\textstyle{\mathbf{DR}^{\mathcal{M}}(A)[n+1]\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ΦAsubscriptΦ𝐴\scriptstyle{\Phi_{A}}𝐏𝐨𝐥(A,n)[n+1]superscript𝐏𝐨𝐥𝐴𝑛delimited-[]𝑛1\textstyle{\mathbf{Pol}^{\mathcal{M}}(A,n)[n+1]}𝟏(2)subscript12\textstyle{\mathbf{1}_{\mathcal{M}}(2)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}αAsubscript𝛼𝐴\scriptstyle{\alpha_{A}}

in ϵ𝐌gritalic-ϵsuperscript𝐌𝑔𝑟\epsilon-\mathbf{M}^{gr}, exhibiting both 𝐃𝐑(A)[n+1]superscript𝐃𝐑𝐴delimited-[]𝑛1\mathbf{DR}^{\mathcal{M}}(A)[n+1] and 𝟏(2)subscript12\mathbf{1}_{\mathcal{M}}(2) as objects in the overcategory ϵ𝐌gr/𝐏𝐨𝐥(A,n)[n+1]italic-ϵsuperscript𝐌𝑔𝑟superscript𝐏𝐨𝐥𝐴𝑛delimited-[]𝑛1\epsilon-\mathbf{M}^{gr}/\mathbf{Pol}^{\mathcal{M}}(A,n)[n+1]. We can then give the following

Definition 2.1.16

Let A𝖬,n+1𝐀𝐥𝐠𝐴subscript𝖬𝑛1𝐀𝐥𝐠A\in\mathbb{P}_{\mathsf{M},n+1}-\mathbf{Alg}. The space of closed 222-shifted forms compatible with the given nsubscript𝑛\mathbb{P}_{n}-structure on A𝐴A is the space of lifts of αAsubscript𝛼𝐴\alpha_{A} along ΦAsubscriptΦ𝐴\Phi_{A}, i.e. the space

𝖢𝗈𝗆𝗉2,cl(A,n):=𝖬𝖺𝗉ϵ𝐌gr/𝐏𝐨𝐥(A,n)[n+1](𝟏(2),𝐃𝐑(A)[n+1]).assignsuperscriptsubscript𝖢𝗈𝗆𝗉2𝑐𝑙𝐴𝑛subscript𝖬𝖺𝗉italic-ϵsuperscript𝐌𝑔𝑟superscript𝐏𝐨𝐥𝐴𝑛delimited-[]𝑛1subscript12superscript𝐃𝐑𝐴delimited-[]𝑛1\mathsf{Comp}_{\mathcal{M}}^{2,cl}(A,n):=\mathsf{Map}_{\epsilon-\mathbf{M}^{gr}/\mathbf{Pol}^{\mathcal{M}}(A,n)[n+1]}(\mathbf{1}_{\mathcal{M}}(2),\mathbf{DR}^{\mathcal{M}}(A)[n+1]).

The reasons for using the words “closed forms” in the previous definition are the following. First of all there is a “forgetful” map

𝖢𝗈𝗆𝗉2,cl(A,n)𝖬𝖺𝗉ϵ𝐌gr(𝟏(2),𝐃𝐑(A)[n+1])=𝒜2,cl(A,n1)superscriptsubscript𝖢𝗈𝗆𝗉2𝑐𝑙𝐴𝑛subscript𝖬𝖺𝗉italic-ϵsuperscript𝐌𝑔𝑟subscript12superscript𝐃𝐑𝐴delimited-[]𝑛1superscriptsubscript𝒜2𝑐𝑙𝐴𝑛1\mathsf{Comp}_{\mathcal{M}}^{2,cl}(A,n)\longrightarrow\mathsf{Map}_{\epsilon-\mathbf{M}^{gr}}(\mathbf{1}_{\mathcal{M}}(2),\mathbf{DR}^{\mathcal{M}}(A)[n+1])=\mathcal{A}_{\mathcal{M}}^{2,cl}(A,n-1)

to the actual space of closed 222-shifted forms on A𝐴A. Moreover, if A𝐴A happens to be non degenerate, then 𝖢𝗈𝗆𝗉2,cl(A,n1)superscriptsubscript𝖢𝗈𝗆𝗉2𝑐𝑙𝐴𝑛1\mathsf{Comp}_{\mathcal{M}}^{2,cl}(A,n-1) is contractible (since ΦAsubscriptΦ𝐴\Phi_{A} is an equivalence), hence there is a unique closed (n1)𝑛1(n-1)-shifted 222-form ωAπ0(𝒜2,cl(A,n1))subscript𝜔𝐴subscript𝜋0superscriptsubscript𝒜2𝑐𝑙𝐴𝑛1\omega_{A}\in\pi_{0}(\mathcal{A}_{\mathcal{M}}^{2,cl}(A,n-1)) on A𝐴A, via the above “forgetful” map. Moreover, by definition, such an ωAsubscript𝜔𝐴\omega_{A} is non degenerate since A𝐴A is: ω𝜔\omega is thus a symplectic structure on A𝐴A. If we denote by ,nnd(A)subscriptsuperscript𝑛𝑑𝑛𝐴\mathbb{P}^{nd}_{\mathcal{M},n}(A) the subspace of ,n(A)subscript𝑛𝐴\mathbb{P}_{\mathcal{M},n}(A) whose connected components consist of non degenerate elements, we get the following

Proposition 2.1.17

For A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}} the above construction yields a well-defined map of spaces    WA:,nnd(A)𝖲𝗒𝗆𝗉𝗅(A,n1):subscript𝑊𝐴subscriptsuperscript𝑛𝑑𝑛𝐴subscript𝖲𝗒𝗆𝗉𝗅𝐴𝑛1W_{A}:\mathbb{P}^{nd}_{\mathcal{M},n}(A)\to\mathsf{Sympl}_{\mathcal{M}}(A,n-1).

There is a parallel (and in fact equivalent) notion of non-degenerate shifted Poisson structure in \mathcal{M}. Let B𝐂𝐀𝐥𝐠𝐵subscript𝐂𝐀𝐥𝐠B\in\mathbf{CAlg}_{\mathcal{M}} such that 𝕃Bsubscriptsuperscript𝕃𝐵\mathbb{L}^{\mathcal{M}}_{B} is a dualizable in B𝐌𝐨𝐝𝐵subscript𝐌𝐨𝐝B-\mathbf{Mod}_{\mathcal{M}}. An n𝑛n-shifted Poisson structure ππ0(𝖯𝗈𝗂𝗌𝗌(B,n))𝜋subscript𝜋0subscript𝖯𝗈𝗂𝗌𝗌𝐵𝑛\pi\in\pi_{0}(\mathsf{Poiss}_{\mathcal{M}}(B,n)) defines a map 𝟏SymB2(𝕋B[n1])[n+2]subscript1𝑆𝑦superscriptsubscript𝑚𝐵2subscriptsuperscript𝕋𝐵delimited-[]𝑛1delimited-[]𝑛2\mathbf{1}_{\mathcal{M}}\to Sym_{B}^{2}(\mathbb{T}^{\mathcal{M}}_{B}[-n-1])[n+2] in \mathcal{M}, and thus induces, by adjunction, a map 𝕃B𝕋B[n]subscriptsuperscript𝕃𝐵subscriptsuperscript𝕋𝐵delimited-[]𝑛\mathbb{L}^{\mathcal{M}}_{B}\to\mathbb{T}^{\mathcal{M}}_{B}[-n] in B𝐌𝐨𝐝𝐵subscript𝐌𝐨𝐝B-\mathbf{Mod}_{\mathcal{M}}: we say that π𝜋\pi is non degenerate if this map is an equivalence. The subspace of 𝖯𝗈𝗂𝗌𝗌(B,n)subscript𝖯𝗈𝗂𝗌𝗌𝐵𝑛\mathsf{Poiss}_{\mathcal{M}}(B,n) whose connected components consist of non degenerate elements will be denoted by 𝖯𝗈𝗂𝗌𝗌nd(B,n)subscriptsuperscript𝖯𝗈𝗂𝗌𝗌𝑛𝑑𝐵𝑛\mathsf{Poiss}^{nd}_{\mathcal{M}}(B,n). The notions of non degeneracy for Poisson algebras and for Poisson structures coincide in the following sense

Proposition 2.1.18

Let B𝐂𝐀𝐥𝐠𝐵subscript𝐂𝐀𝐥𝐠B\in\mathbf{CAlg}_{\mathcal{M}} such that 𝕃Bsubscriptsuperscript𝕃𝐵\mathbb{L}^{\mathcal{M}}_{B} is a dualizable in B𝐌𝐨𝐝𝐵subscript𝐌𝐨𝐝B-\mathbf{Mod}_{\mathcal{M}}. The equivalence wBsubscript𝑤𝐵w_{B} of Corollary 2.1.11, restricts to an equivalence ,nnd(B)𝖯𝗈𝗂𝗌𝗌nd(B,n1)similar-to-or-equalssubscriptsuperscript𝑛𝑑𝑛𝐵subscriptsuperscript𝖯𝗈𝗂𝗌𝗌𝑛𝑑𝐵𝑛1\mathbb{P}^{nd}_{\mathcal{M},n}(B)\simeq\mathsf{Poiss}^{nd}_{\mathcal{M}}(B,n-1). As a consequence of Proposition 2.1.17, we thus get a map

WB:𝖯𝗈𝗂𝗌𝗌nd(B,n1)𝖲𝗒𝗆𝗉𝗅(B,n1):superscriptsubscript𝑊𝐵subscriptsuperscript𝖯𝗈𝗂𝗌𝗌𝑛𝑑𝐵𝑛1subscript𝖲𝗒𝗆𝗉𝗅𝐵𝑛1W_{B}^{\prime}:\mathsf{Poiss}^{nd}_{\mathcal{M}}(B,n-1)\longrightarrow\mathsf{Sympl}_{\mathcal{M}}(B,n-1)

The map WBsubscriptsuperscript𝑊𝐵W^{\prime}_{B} in Proposition 2.1.18 is called the comparison map between non degenerate shifted Poisson structures and shifted symplectic structures. We will study it for derived Artin stacks in Section 2.4.

Remark 2.1.19

It is possible that the map WBsubscriptsuperscript𝑊𝐵W^{\prime}_{B} in Proposition 2.1.18 will turn out to be an equivalence, for very general \mathcal{M}. We have proven this in our geometric case of interest (see Theorem 2.4.2). The difficulty in general stems from the fact that while for the source of WBsuperscriptsubscript𝑊𝐵W_{B}^{\prime} we have Theorem 2.1.11 and Proposition 2.1.13, and we thus are able to perform the 𝐃𝐑𝐃𝐑\mathbf{DR}-to-𝐏𝐨𝐥𝐏𝐨𝐥\mathbf{Pol} construction and produce the map WBsubscriptsuperscript𝑊𝐵W^{\prime}_{B}, we don’t have anything similar for the target of WBsubscriptsuperscript𝑊𝐵W^{\prime}_{B}. One runs into serious difficulties even just trying to construct an inverse equivalence to WBsubscriptsuperscript𝑊𝐵W^{\prime}_{B} at the level of connected components. The first obstacle is that a shifted symplectic structure is, by definition, a weak map (i.e. a map in the relevant homotopy category). Even if we could strictify this map (getting a strictly closed shifted 222-form), we are still left with the problem that non degeneracy is a weak property, i.e. the property of a map being a quasi-isomorphism, and therefore cannot be readily used to build a strict Lie bracket on the de Rham algebra. One might be able to overcome these difficulties in general but we do not know how to do this at the moment.

2.1.6 Tate realizations over k𝑘k

In this section, we will concentrate on the special case 𝖬=ϵ𝖽𝗀kgr𝖬italic-ϵsubscriptsuperscript𝖽𝗀𝑔𝑟𝑘\mathsf{M}=\epsilon-\mathsf{dg}^{gr}_{k}, with its associated \infty-category =ϵ𝐝𝐠gritalic-ϵsuperscript𝐝𝐠𝑔𝑟\mathcal{M}=\epsilon-\mathbf{dg}^{gr}, and in the next Section we will apply the definitions and results obtained here to categories of diagrams in \mathcal{M}.

The unit of the symmetric monoidal category 𝖬𝖬\mathsf{M} is k(0)𝑘0k(0), i.e. the complex k[0]𝑘delimited-[]0k[0] sitting in pure weight 00, with the trivial mixed differential. The enriching symmetric monoidal functor is given by 𝖽𝗀k𝖬:VV(0):subscript𝖽𝗀𝑘𝖬maps-to𝑉𝑉0\mathsf{dg}_{k}\to\mathsf{M}\,:V\mapsto V(0), the tensor enrichment is then given by VE:=V(0)𝖬Eassigntensor-product𝑉𝐸subscripttensor-product𝖬𝑉0𝐸V\otimes E:=V(0)\otimes_{\mathsf{M}}E, for E𝖬𝐸𝖬E\in\mathsf{M}. The enriched hom object is thus

Hom¯k(E,E):=Zϵ(Hom¯(E,E)(0))𝖽𝗀k,assignsubscript¯𝐻𝑜𝑚𝑘𝐸superscript𝐸subscriptZitalic-ϵsubscript¯𝐻𝑜𝑚𝐸superscript𝐸0subscript𝖽𝗀𝑘\underline{Hom}_{k}(E,E^{\prime}):=\mathrm{Z}_{\epsilon}(\underline{Hom}_{\mathcal{M}}(E,E^{\prime})(0))\in\mathsf{dg}_{k}\,,

where Hom¯𝖬subscript¯𝐻𝑜𝑚𝖬\underline{Hom}_{\mathsf{M}} denotes the internal Hom in 𝖬𝖬\mathsf{M}, and, for F𝖬𝐹𝖬F\in\mathsf{M}, we write Zϵ(F(0))𝐝𝐠ksubscriptZitalic-ϵ𝐹0subscript𝐝𝐠𝑘\mathrm{Z}_{\epsilon}(F(0))\in\mathbf{dg}_{k} for the kernel of the map of k𝑘k-dg-modules ϵ:F(0)F(1)[1]:italic-ϵ𝐹0𝐹1delimited-[]1\epsilon:F(0)\to F(1)[1].

The standard realization functor ||:𝐝𝐠k|-|:\mathcal{M}\to\mathbf{dg}_{k}, for =ϵ𝐝𝐠gritalic-ϵsuperscript𝐝𝐠𝑔𝑟\mathcal{M}=\epsilon-\mathbf{dg}^{gr}, is in some sense unsatisfactory since an easy computation ([CPTVV, Proposition 1.5.1]) shows that |E|p0E(p)similar-to-or-equals𝐸subscriptproduct𝑝0𝐸𝑝|E|\simeq\prod_{p\geq 0}E(p), for E𝐸E\in\mathcal{M}, so that all negative weights are lost under standard realization. An obvious way to modify |||-| and remedy this flaw is to consider |E|t:=colimi0piE(p)assignsuperscript𝐸𝑡subscriptcolim𝑖0subscriptproduct𝑝𝑖𝐸𝑝|E|^{t}:=\textrm{colim}_{i\geq 0}\prod_{p\geq-i}E(p), instead. This new functor ||t:𝐝𝐠k|-|^{t}:\mathcal{M}\to\mathbf{dg}_{k} will be called the Tate realization functor for \mathcal{M}. By definition, there is a canonical morphism ||||t|-|\to|-|^{t} of \infty-functors. One can show, exactly as for |||-|, that ||t|-|^{t} is lax symmetric monoidal as well, so that it is inherited by categories of algebras. In particular, we also get Tate realization functors

||t:ϵ𝐂𝐀𝐥𝐠grϵ𝐜𝐝𝐠𝐚kgr,||t:,ngr𝐀𝐥𝐠n𝐜𝐝𝐠𝐚kgr.|-|^{t}:\epsilon-\mathbf{CAlg}^{gr}_{\mathcal{M}}\to\epsilon-\mathbf{cdga}^{gr}_{k}\,,\,\,\,\,|-|^{t}:\mathbb{P}^{gr}_{\mathcal{M},n}-\mathbf{Alg}\to\mathbb{P}_{n}-\mathbf{cdga}^{gr}_{k}\,.

As in the linear case, there are canonical morphisms ||||t|-|\to|-|^{t} of \infty-functors between realizations on the level of these algebra structures.

Let us put this into a broader perspective, and relate the Tate realization to a standard realization (on a different category). Let us start by the fact that there is an equivalence in 𝐝𝐠ksubscript𝐝𝐠𝑘\mathbf{dg}_{k}

om¯k(k(i),k(i+1))=Hom¯k(k(i),k(i+1))k[0]subscript¯𝑜𝑚𝑘𝑘𝑖𝑘𝑖1subscript¯𝐻𝑜𝑚𝑘𝑘𝑖𝑘𝑖1similar-to-or-equals𝑘delimited-[]0\underline{\mathcal{H}om}_{k}(k(i),k(i+1))=\mathbb{R}\underline{Hom}_{k}(k(i),k(i+1))\simeq k[0]\,

(where om¯ksubscript¯𝑜𝑚𝑘\underline{\mathcal{H}om}_{k} denotes the 𝐝𝐠ksubscript𝐝𝐠𝑘\mathbf{dg}_{k}-enriched Hom in \mathcal{M}), giving rise to the following canonical ind-object in \mathcal{M}

k():={k(0)k(1)k(i)k(i+1)}𝐈𝐧𝐝().assign𝑘𝑘0𝑘1𝑘𝑖𝑘𝑖1𝐈𝐧𝐝k(\infty):=\{k(0)\to k(1)\to\dots\to k(i)\to k(i+1)\to\cdots\}\,\in\mathbf{Ind}(\mathcal{M}).

One can then show that the Tate realization functor for \mathcal{M} is related to the standard realization ||𝐈𝐧𝐝|-|_{\mathbf{Ind}} for 𝐈𝐧𝐝()𝐈𝐧𝐝\mathbf{Ind}(\mathcal{M}), by

||t|k()|𝐈𝐧𝐝:𝐝𝐠k|-|^{t}\simeq|\,-\otimes k(\infty)\,|_{\mathbf{Ind}}\,:\mathcal{M}\longrightarrow\mathbf{dg}_{k}\,

where we have implicitly used the canonical functor 𝐈𝐧𝐝()𝐈𝐧𝐝\mathcal{M}\to\mathbf{Ind}(\mathcal{M}), sending an object E𝐸E to the constant ind-diagram in \mathcal{M} with value E𝐸E.

Moreover, since k(i)k(j)k(i+j)similar-to-or-equalstensor-product𝑘𝑖𝑘𝑗𝑘𝑖𝑗k(i)\otimes k(j)\simeq k(i+j), k()𝑘k(\infty) is a commutative monoid object in 𝐈𝐧𝐝()𝐈𝐧𝐝\mathbf{Ind}(\mathcal{M}), hence

A𝐂𝐀𝐥𝐠A():=Ak()k()/𝐂𝐀𝐥𝐠𝐈𝐧𝐝().𝐴subscript𝐂𝐀𝐥𝐠𝐴assigntensor-product𝐴𝑘𝑘subscript𝐂𝐀𝐥𝐠𝐈𝐧𝐝A\in\mathbf{CAlg}_{\mathcal{M}}\,\,\Rightarrow\,\,A(\infty):=A\otimes k(\infty)\in k(\infty)/\mathbf{CAlg}_{\mathbf{Ind}(\mathcal{M})}.

Therefore it make sense to consider the relative objects

𝐃𝐑𝐈𝐧𝐝()(A()/k())ϵ𝐀𝐥𝐠𝐈𝐧𝐝()gr,𝐏𝐨𝐥𝐈𝐧𝐝()(A()/k(),n)𝐈𝐧𝐝(),n+1𝐀𝐥𝐠gr,formulae-sequencesuperscript𝐃𝐑𝐈𝐧𝐝𝐴𝑘italic-ϵsubscriptsuperscript𝐀𝐥𝐠𝑔𝑟𝐈𝐧𝐝superscript𝐏𝐨𝐥𝐈𝐧𝐝𝐴𝑘𝑛subscript𝐈𝐧𝐝𝑛1superscript𝐀𝐥𝐠𝑔𝑟\mathbf{DR}^{\mathbf{Ind}(\mathcal{M})}(A(\infty)/k(\infty))\in\epsilon-\mathbf{Alg}^{gr}_{\mathbf{Ind}(\mathcal{M})}\,,\,\,\,\,\mathbf{Pol}^{\mathbf{Ind}(\mathcal{M})}(A(\infty)/k(\infty),n)\in\mathbb{P}_{\mathbf{Ind}(\mathcal{M}),n+1}-\mathbf{Alg}^{gr}\,,

and their standard realizations (on algebras in 𝐈𝐧𝐝()𝐈𝐧𝐝\mathbf{Ind}(\mathcal{M}))

𝐃𝐑(A()/k())ϵ𝐜𝐝𝐠𝐚kgr,𝐏𝐨𝐥(A()/k(),n)n+1𝐜𝐝𝐠𝐚kgr.formulae-sequence𝐃𝐑𝐴𝑘italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘𝐏𝐨𝐥𝐴𝑘𝑛subscript𝑛1superscriptsubscript𝐜𝐝𝐠𝐚𝑘𝑔𝑟\mathbf{DR}(A(\infty)/k(\infty))\in\epsilon-\mathbf{cdga}^{gr}_{k}\,,\,\,\,\,\mathbf{Pol}(A(\infty)/k(\infty),n)\in\mathbb{P}_{n+1}-\mathbf{cdga}_{k}^{gr}\,.

For de Rham algebras and polyvectors, we have the following comparison result

Proposition 2.1.20

If A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}}, then we have a canonical equivalence

𝐃𝐑t(A):=|𝐃𝐑(A)|t𝐃𝐑(A()/k())assignsuperscript𝐃𝐑𝑡𝐴superscript𝐃𝐑𝐴𝑡similar-to-or-equals𝐃𝐑𝐴𝑘\mathbf{DR}^{t}(A):=|\mathbf{DR}(A)|^{t}\simeq\mathbf{DR}(A(\infty)/k(\infty))

in ϵ𝐜𝐝𝐠𝐚kgritalic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\epsilon-\mathbf{cdga}^{gr}_{k}. If moreover 𝕃Asubscriptsuperscript𝕃𝐴\mathbb{L}^{\mathcal{M}}_{A} is dualizable in A𝐌𝐨𝐝𝐴subscript𝐌𝐨𝐝A-\mathbf{Mod}_{\mathcal{M}}, then we have a canonical equivalence

𝐏𝐨𝐥t(A,n):=|𝐏𝐨𝐥(A,n)|t𝐏𝐨𝐥(A()/k())assignsuperscript𝐏𝐨𝐥𝑡𝐴𝑛superscript𝐏𝐨𝐥𝐴𝑛𝑡similar-to-or-equals𝐏𝐨𝐥𝐴𝑘\mathbf{Pol}^{t}(A,n):=|\mathbf{Pol}(A,n)|^{t}\simeq\mathbf{Pol}(A(\infty)/k(\infty))

in n+1𝐜𝐝𝐠𝐚kgrsubscript𝑛1superscriptsubscript𝐜𝐝𝐠𝐚𝑘𝑔𝑟\mathbb{P}_{n+1}-\mathbf{cdga}_{k}^{gr}.

Remark 2.1.21

As already observed in the linear case, in general, none of the morphisms 𝐃𝐑(A)𝐃𝐑t(A)𝐃𝐑𝐴superscript𝐃𝐑𝑡𝐴\mathbf{DR}(A)\to\mathbf{DR}^{t}(A), 𝐏𝐨𝐥(A,n)𝐏𝐨𝐥t(A,n)𝐏𝐨𝐥𝐴𝑛superscript𝐏𝐨𝐥𝑡𝐴𝑛\mathbf{Pol}(A,n)\to\mathbf{Pol}^{t}(A,n) are equivalences. If Aϵ𝐂𝐀𝐥𝐠gr𝐴italic-ϵsubscriptsuperscript𝐂𝐀𝐥𝐠𝑔𝑟A\in\epsilon-\mathbf{CAlg}^{gr}_{\mathcal{M}} happens to have only non-negative weights (this will be the case in our application to derived stacks), then also 𝕃Asubscript𝕃𝐴\mathbb{L}_{A} will have the same property, and 𝐃𝐑(A)𝐃𝐑t(A)𝐃𝐑𝐴superscript𝐃𝐑𝑡𝐴\mathbf{DR}(A)\to\mathbf{DR}^{t}(A) will indeed be an equivalence. However, the dual to 𝕃Asubscript𝕃𝐴\mathbb{L}_{A}, if existing, will have postive weights, so that 𝐏𝐨𝐥(A,n)𝐏𝐨𝐥t(A,n)𝐏𝐨𝐥𝐴𝑛superscript𝐏𝐨𝐥𝑡𝐴𝑛\mathbf{Pol}(A,n)\to\mathbf{Pol}^{t}(A,n) will not be an equivalence, even in this case. So, at least for applications to derived stacks, while the introduction of the Tate realization will not make any difference for 𝐃𝐑𝐃𝐑\mathbf{DR}, it will definitely do for 𝐏𝐨𝐥𝐏𝐨𝐥\mathbf{Pol}, and indeed the interesting realization will be 𝐏𝐨𝐥tsuperscript𝐏𝐨𝐥𝑡\mathbf{Pol}^{t} rather than the standard one.

2.2 Formal derived stacks and formal localization

A crucial ingredient in the theory of shifted Poisson structures on general derived Artin stacks is the method of formal localization. Formal localization is interesting in its own right as a new, very powerful tool that will prove useful in many other situations inside derived geometry, especially in order to globalize tricky constructions and results, whose extension from the local case presents obstructions that only vanish formally locally. An example is given by obstructions living in de Rham cohomology (even, say, on a smooth scheme X𝑋X). Suppose we wish to glue some construction that we can perform “locally” on X𝑋X, and we know that obstructions to globalize live in de Rham cohomology of X𝑋X (e.g. we would like to globalize an algebraic version of the Darboux lemma). Since de Rham cohomology never vanishes Zariski or étale locally, it is going to be hard or impossible, depending on the specific problem, to suppress the obstructions and glue with respect to these topologies. On the other hand, for any xX𝑥𝑋x\in X, the de Rham cohomology of the formal completion Xx^^subscript𝑋𝑥\widehat{X_{x}} does vanish, so we may try to glue the construction performed on the family {Xx^|xX}conditional-set^subscript𝑋𝑥𝑥𝑋\{\widehat{X_{x}}\,|\,x\in X\} to a construction on X𝑋X. But in order to do this we need a result telling us when and how we are able to glue objects along the family of formal completions. This is exactly the content of formal localization and below we will sketch how it works and why it is useful.

The following conventions will be adopted throughout this section. A cdga A𝐴A is called almost finitely presented if H0(A)superscript𝐻0𝐴H^{0}(A) is a k𝑘k-algebra of finite type, and each Hi(A)superscript𝐻𝑖𝐴H^{i}(A) is a finitely presented H0(A)superscript𝐻0𝐴H^{0}(A)-module. We will write 𝐝𝐀𝐟𝐟ksubscript𝐝𝐀𝐟𝐟𝑘\mathbf{dAff}_{k} for the opposite \infty-category of almost finitely presented cdga’s, and we will simply refer to its objects as derived affine schemes without mentioning the finite presentation condition. In particular, when writing 𝐒𝐩𝐞𝐜A𝐒𝐩𝐞𝐜𝐴\mathbf{Spec}\,A, we implicitly assume that 𝐒𝐩𝐞𝐜A𝐒𝐩𝐞𝐜𝐴\mathbf{Spec}\,A is an object of 𝐝𝐀𝐟𝐟ksubscript𝐝𝐀𝐟𝐟𝑘\mathbf{dAff}_{k}. The \infty-category 𝐝𝐀𝐟𝐟ksubscript𝐝𝐀𝐟𝐟𝑘\mathbf{dAff}_{k} is equipped with its usual étale topology of [HAG-II, Definition 2.2.2.3], and the corresponding \infty-topos of derived stacks will be denoted by 𝐝𝐒𝐭ksubscript𝐝𝐒𝐭𝑘\mathbf{dSt}_{k}. Its objects will simply be called derived stacks, instead of the more precise but longer locally almost finitely presented derived stacks over k𝑘k. With these conventions, an algebraic derived n𝑛n-stack X𝑋X will have a smooth atlas by objects in 𝐝𝐀𝐟𝐟ksubscript𝐝𝐀𝐟𝐟𝑘\mathbf{dAff}_{k}, and in particular, it will have a bounded above cotangent complex in 𝖢𝗈𝗁(X)𝖢𝗈𝗁𝑋\mathsf{Coh}(X).

2.2.1 Formal derived stacks

As the name suggests, formal localization deals with formal derived stacks, which we now define.

Definition 2.2.1

A formal derived stack is a derived stack F𝐹F satisfying the following conditions.

  1. 1.

    F𝐹F is nilcomplete i.e. for all 𝐒𝐩𝐞𝐜A𝐝𝐀𝐟𝐟k𝐒𝐩𝐞𝐜𝐴subscript𝐝𝐀𝐟𝐟𝑘\mathbf{Spec}\,A\in\mathbf{dAff}_{k}, the canonical map

    F(A)limkF(Ak)𝐹𝐴subscript𝑘𝐹subscript𝐴absent𝑘F(A)\longrightarrow\lim_{k}F(A_{\leq k})

    (induced by the Postnikov tower of A𝐴A) is an equivalence in 𝐓𝐓\mathbf{T}.

  2. 2.

    F𝐹F is infinitesimally cohesive i.e. for all cartesian squares of almost finitely presented k𝑘k-cdga’s in non-positive degrees

    B𝐵\textstyle{B\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}B1subscript𝐵1\textstyle{B_{1}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}B2subscript𝐵2\textstyle{B_{2}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}B0,subscript𝐵0\textstyle{B_{0},}

    such that each π0(Bi)π0(B0)subscript𝜋0subscript𝐵𝑖subscript𝜋0subscript𝐵0\pi_{0}(B_{i})\longrightarrow\pi_{0}(B_{0}) is surjective with nilpotent kernel, then the induced square

    F(B)𝐹𝐵\textstyle{F(B)\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}F(B1)𝐹subscript𝐵1\textstyle{F(B_{1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}F(B2)𝐹subscript𝐵2\textstyle{F(B_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}F(B0),𝐹subscript𝐵0\textstyle{F(B_{0}),}

    is cartesian in 𝐓𝐓\mathbf{T}.

Remark 2.2.2

(1) The property of being infinitesimally cohesive is a derived variation of the Schlessinger condition in classical deformation theory ([Sch]). In particular, one can show that any derived Artin stack F𝐹F is a formal derived stack ([HAG-II, Appendix]), and it is actually cohesive i.e. sends any diagram as in 2.2.1 (2)2(2), with the nilpotency condition possibly omitted, to pullbacks in 𝐓𝐓\mathbf{T} ([Lu3, DAG IX, Corollary 6.5] and [Lu3, DAG XIV, Lemma 2.1.7]).

(2) A small limit of formal derived stacks is a formal derived stack.

Let the \infty-functor i:𝐚𝐥𝐠kred𝐜𝐝𝐠𝐚k:𝑖subscriptsuperscript𝐚𝐥𝐠red𝑘subscript𝐜𝐝𝐠𝐚𝑘i:\mathbf{alg}^{\textrm{red}}_{k}\longrightarrow\mathbf{cdga}_{k} be the inclusion of the full reflective sub \infty-category of reduced discrete objects (i.e. R𝐜𝐝𝐠𝐚k𝑅subscript𝐜𝐝𝐠𝐚𝑘R\in\mathbf{cdga}_{k} such that R𝑅R is discrete and RH0(R)similar-to-or-equals𝑅superscript𝐻0𝑅R\simeq H^{0}(R) is a usual reduced k𝑘k-algebra). The \infty-functor i𝑖i has a left adjoint

()red:𝐜𝐝𝐠𝐚k𝐚𝐥𝐠kred,AAred:=H0(A)/Nilp(H0(A),(-)^{red}:\mathbf{cdga}_{k}\longrightarrow\mathbf{alg}^{\textrm{red}}_{k}\,,\,\,A\longmapsto A^{red}:=H^{0}(A)/\textrm{Nilp}(H^{0}(A),

and it is easy to verify that we get an induced \infty-functor i:𝐝𝐒𝐭k𝐒𝐭red,k:superscript𝑖subscript𝐝𝐒𝐭𝑘subscript𝐒𝐭red𝑘i^{*}:\mathbf{dSt}_{k}\longrightarrow\mathbf{St}_{\textrm{red},k}, where 𝐒𝐭red,ksubscript𝐒𝐭red𝑘\mathbf{St}_{\textrm{red},k} is the \infty-category of stacks on (𝐚𝐥𝐠kred)opsuperscriptsubscriptsuperscript𝐚𝐥𝐠red𝑘𝑜𝑝(\mathbf{alg}^{\textrm{red}}_{k})^{op} for the étale topology. Now isuperscript𝑖i^{*} has both a right adjoint isubscript𝑖i_{*}, and a left adjoint i!subscript𝑖i_{!}, both fully faithful, and i!isubscript𝑖superscript𝑖i_{!}i^{*} is left adjoint to iisubscript𝑖superscript𝑖i_{*}i^{*}.

Definition 2.2.3
  1. 1.

    The functor ()DR:=ii:𝐝𝐒𝐭k𝐝𝐒𝐭k:assignsubscript𝐷𝑅subscript𝑖superscript𝑖subscript𝐝𝐒𝐭𝑘subscript𝐝𝐒𝐭𝑘(-)_{DR}:=i_{*}i^{*}:\mathbf{dSt}_{k}\longrightarrow\mathbf{dSt}_{k} is called the de Rham stack functor. By adjunction, for any F𝐝𝐒𝐭k𝐹subscript𝐝𝐒𝐭𝑘F\in\mathbf{dSt}_{k}, we have a canonical natural map qF:FFDR:subscript𝑞𝐹𝐹subscript𝐹𝐷𝑅q_{F}:F\to F_{DR}.

  2. 2.

    The functor ()red:=i!i:𝐝𝐒𝐭k𝐝𝐒𝐭k:assignsubscriptredsubscript𝑖superscript𝑖subscript𝐝𝐒𝐭𝑘subscript𝐝𝐒𝐭𝑘(-)_{\textrm{red}}:=i_{!}i^{*}:\mathbf{dSt}_{k}\longrightarrow\mathbf{dSt}_{k} is called the reduced stack functor. By adjunction, for any F𝐝𝐒𝐭k𝐹subscript𝐝𝐒𝐭𝑘F\in\mathbf{dSt}_{k}, we have a canonical natural map ιF:FredF:subscript𝜄𝐹subscript𝐹red𝐹\iota_{F}:F_{\textrm{red}}\to F.

  3. 3.

    Let f:FG:𝑓𝐹𝐺f:F\longrightarrow G be a morphism in 𝐝𝐒𝐭ksubscript𝐝𝐒𝐭𝑘\mathbf{dSt}_{k}. We define the formal completion G^fsubscript^𝐺𝑓\widehat{G}_{f} of G𝐺G along the morphism f𝑓f as the fibered product in 𝐝𝐒𝐭ksubscript𝐝𝐒𝐭𝑘\mathbf{dSt}_{k}:

    G^fsubscript^𝐺𝑓\textstyle{\widehat{G}_{f}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}FDRsubscript𝐹𝐷𝑅\textstyle{F_{DR}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}fDRsubscript𝑓𝐷𝑅\scriptstyle{f_{DR}}G𝐺\textstyle{G\ignorespaces\ignorespaces\ignorespaces\ignorespaces}qGsubscript𝑞𝐺\scriptstyle{q_{G}}GDR.subscript𝐺𝐷𝑅\textstyle{G_{DR}.}

Since the left adjoint to i𝑖i is ()redsuperscript𝑟𝑒𝑑(-)^{red}, then it is easy to see that FDR(A)F(Ared)similar-to-or-equalssubscript𝐹𝐷𝑅𝐴𝐹superscript𝐴redF_{DR}(A)\simeq F(A^{\textrm{red}}), and (𝐒𝐩𝐞𝐜A)red𝐒𝐩𝐞𝐜(Ared),similar-to-or-equalssubscript𝐒𝐩𝐞𝐜𝐴red𝐒𝐩𝐞𝐜superscript𝐴red(\mathbf{Spec}\,A)_{\textrm{red}}\simeq\mathbf{Spec}\,(A^{\textrm{red}}), for any A𝐜𝐝𝐠𝐚k𝐴subscript𝐜𝐝𝐠𝐚𝑘A\in\mathbf{cdga}_{k}. Therefore G^f(A)=G(A)×G(Ared)F(Ared)subscript^𝐺𝑓𝐴subscript𝐺superscript𝐴𝑟𝑒𝑑𝐺𝐴𝐹superscript𝐴𝑟𝑒𝑑\widehat{G}_{f}(A)=G(A)\times_{G(A^{red})}F(A^{red}), for f:FG:𝑓𝐹𝐺f:F\to G in 𝐝𝐒𝐭ksubscript𝐝𝐒𝐭𝑘\mathbf{dSt}_{k}. We already observed that ()DRsubscript𝐷𝑅(-)_{DR} is right adjoint to ()redsubscriptred(-)_{\textrm{red}}, as functors 𝐝𝐒𝐭k𝐝𝐒𝐭ksubscript𝐝𝐒𝐭𝑘subscript𝐝𝐒𝐭𝑘\mathbf{dSt}_{k}\to\mathbf{dSt}_{k}. We list a few elementary properties of these constructions:

  • FDRsubscript𝐹𝐷𝑅F_{DR} is a formal derived stack for any F𝐝𝐒𝐭k𝐹subscript𝐝𝐒𝐭𝑘F\in\mathbf{dSt}_{k}.

  • If G𝐺G is a formal derived stack, the formal completion G^fsubscript^𝐺𝑓\widehat{G}_{f}, along any map f:FG:𝑓𝐹𝐺f:F\to G in 𝐝𝐒𝐭ksubscript𝐝𝐒𝐭𝑘\mathbf{dSt}_{k}, is again a formal derived stack.

  • For any F𝐝𝐒𝐭k𝐹subscript𝐝𝐒𝐭𝑘F\in\mathbf{dSt}_{k}, if j:𝗍0FF:𝑗subscript𝗍0𝐹𝐹j:\mathsf{t}_{0}F\to F denotes the canonical map in 𝐝𝐒𝐭ksubscript𝐝𝐒𝐭𝑘\mathbf{dSt}_{k} from the truncation of F𝐹F to F𝐹F, then the canonical map F^jFsubscript^𝐹𝑗𝐹\widehat{F}_{j}\to F is an equivalence.

Our main object of study in the next section, will be the map q:FFDR:𝑞𝐹subscript𝐹𝐷𝑅q:F\to F_{DR}. If 𝕂𝕂\mathbb{K} is a field, and x:𝐒𝐩𝐞𝐜𝕂FDR:𝑥𝐒𝐩𝐞𝐜𝕂subscript𝐹𝐷𝑅x:\mathbf{Spec}\,\mathbb{K}\to F_{DR} is a point (by adjunction, this is the same thing as a 𝕂𝕂\mathbb{K}-point of F𝐹F, since 𝕂𝕂\mathbb{K} is reduced), then the fiber of q𝑞q at x𝑥x is exactly the (classical) formal completion Xx^^subscript𝑋𝑥\widehat{X_{x}} of X𝑋X at the closed point x𝑥x. More generally, we have

Proposition 2.2.4

Let F𝐝𝐒𝐭k𝐹subscript𝐝𝐒𝐭𝑘F\in\mathbf{dSt}_{k}, 𝐒𝐩𝐞𝐜A𝐝𝐀𝐟𝐟k𝐒𝐩𝐞𝐜𝐴subscript𝐝𝐀𝐟𝐟𝑘\mathbf{Spec}\,A\in\mathbf{dAff}_{k}, and u¯:𝐒𝐩𝐞𝐜AFDR:¯𝑢𝐒𝐩𝐞𝐜𝐴subscript𝐹𝐷𝑅\overline{u}:\mathbf{Spec}\,A\longrightarrow F_{DR}, corresponding to a morphism u:𝐒𝐩𝐞𝐜AredF:𝑢𝐒𝐩𝐞𝐜superscript𝐴𝑟𝑒𝑑𝐹u:\mathbf{Spec}\,A^{red}\longrightarrow F. Then the base-change derived stack F×FDR𝐒𝐩𝐞𝐜Asubscriptsubscript𝐹𝐷𝑅𝐹𝐒𝐩𝐞𝐜𝐴F\times_{F_{DR}}\mathbf{Spec}\,A is equivalent to the formal completion (𝐒𝐩𝐞𝐜A×F)^(i,u)subscript^𝐒𝐩𝐞𝐜𝐴𝐹𝑖𝑢\widehat{(\mathbf{Spec}\,A\times F)}_{(i,u)} of the graph morphism

(i,u):𝐒𝐩𝐞𝐜Ared𝐒𝐩𝐞𝐜A×F,:𝑖𝑢𝐒𝐩𝐞𝐜superscript𝐴𝑟𝑒𝑑𝐒𝐩𝐞𝐜𝐴𝐹(i,u):\mathbf{Spec}\,A^{red}\longrightarrow\mathbf{Spec}\,A\times F,

where i:𝐒𝐩𝐞𝐜Ared𝐒𝐩𝐞𝐜A:𝑖𝐒𝐩𝐞𝐜superscript𝐴𝑟𝑒𝑑𝐒𝐩𝐞𝐜𝐴i:\mathbf{Spec}\,A^{red}\longrightarrow\mathbf{Spec}\,A is the natural map.

Remark 2.2.5

Suppose that F𝐝𝐒𝐭k𝐹subscript𝐝𝐒𝐭𝑘F\in\mathbf{dSt}_{k} has a cotangent complex (e.g. F𝐹F is a derived Artin stack). Then 𝕃F/FDRsubscript𝕃𝐹subscript𝐹𝐷𝑅\mathbb{L}_{F/F_{DR}} exists, and we have an equivalence 𝕃F𝕃F/FDRsimilar-to-or-equalssubscript𝕃𝐹subscript𝕃𝐹subscript𝐹𝐷𝑅\mathbb{L}_{F}\simeq\mathbb{L}_{F/F_{DR}} in 𝐐𝐂𝐨𝐡(F)𝐐𝐂𝐨𝐡𝐹\mathbf{QCoh}(F). In fact, by the transitivity sequence for the map q:FFDR:𝑞𝐹subscript𝐹𝐷𝑅q:F\to F_{DR}. it is enough to show that 𝕃FDR0similar-to-or-equalssubscript𝕃subscript𝐹𝐷𝑅0\mathbb{L}_{F_{DR}}\simeq 0. But this follows immediately from the equivalences: FDR(AE)F((AE)red)F(Ared)FDR(A),similar-to-or-equalssubscript𝐹𝐷𝑅direct-sum𝐴𝐸𝐹superscriptdirect-sum𝐴𝐸redsimilar-to-or-equals𝐹superscript𝐴redsimilar-to-or-equalssubscript𝐹𝐷𝑅𝐴F_{DR}(A\oplus E)\simeq F((A\oplus E)^{\textrm{red}})\simeq F(A^{\textrm{red}})\simeq F_{DR}(A), for any A𝐜𝐝𝐠𝐚k0𝐴superscriptsubscript𝐜𝐝𝐠𝐚𝑘absent0A\in\mathbf{cdga}_{k}^{\leq 0}, and any E𝐝𝐠A0𝐸subscriptsuperscript𝐝𝐠absent0𝐴E\in\mathbf{dg}^{\leq 0}_{A}.

2.2.2 Formal localization for XXDR𝑋subscript𝑋𝐷𝑅X\to X_{DR}

The general theory of formal localization is developed in detail in Section 2 of [CPTVV]. Instead of giving a complete account, we will content ourselves with the application of the general theory to our main case of interest, i.e. to the map q:XXDR:𝑞𝑋subscript𝑋𝐷𝑅q:X\to X_{DR}. Throughout this Section, X𝑋X will be derived Artin stack lfp over k𝑘k (hence with a perfect cotangent complex).

The map XXDR𝑋subscript𝑋𝐷𝑅X\to X_{DR} as a family of formal derived stacks. First of all, let us observe that q:XXDR:𝑞𝑋subscript𝑋𝐷𝑅q:X\to X_{DR} is an algebraisable family of perfect formal derived stacks, i.e. for any derived point 𝐒𝐩𝐞𝐜AXDR𝐒𝐩𝐞𝐜𝐴subscript𝑋𝐷𝑅\mathbf{Spec}\,A\to X_{DR} the corresponding fiber XA𝐒𝐩𝐞𝐜Asubscript𝑋𝐴𝐒𝐩𝐞𝐜𝐴X_{A}\to\mathbf{Spec}\,A of q𝑞q has the following properties:

  1. 1.

    XAsubscript𝑋𝐴X_{A} is a formal derived stack, and the canonical map (XA)red𝐒𝐩𝐞𝐜Aredsubscriptsubscript𝑋𝐴red𝐒𝐩𝐞𝐜subscript𝐴red(X_{A})_{\textrm{red}}\to\mathbf{Spec}\,A_{\textrm{red}} is an equivalence in 𝐝𝐒𝐭ksubscript𝐝𝐒𝐭𝑘\mathbf{dSt}_{k}.

  2. 2.

    The relative cotangent complex 𝕃XA/Asubscript𝕃subscript𝑋𝐴𝐴\mathbb{L}_{X_{A}/A} is perfect (by Remark 2.2.5 and base change)

  3. 3.

    XAsubscript𝑋𝐴X_{A} has a cohomologically bounded above coherent cotangent complex 𝕃XAsubscript𝕃subscript𝑋𝐴\mathbb{L}_{X_{A}} (i.e. for any xB:𝐒𝐩𝐞𝐜BXA:subscript𝑥𝐵𝐒𝐩𝐞𝐜𝐵subscript𝑋𝐴x_{B}:\mathbf{Spec}\,B\to X_{A}, the fiber xB𝕃XAsuperscriptsubscript𝑥𝐵subscript𝕃subscript𝑋𝐴x_{B}^{*}\mathbb{L}_{X_{A}} is a cohomologically bounded above coherent B𝐵B-dg-module).

  4. 4.

    XAsubscript𝑋𝐴X_{A} is equivalent to the formal completion of X×𝐒𝐩𝐞𝐜A𝑋𝐒𝐩𝐞𝐜𝐴X\times\mathbf{Spec}\,A along the map 𝐒𝐩𝐞𝐜AredX×𝐒𝐩𝐞𝐜A𝐒𝐩𝐞𝐜subscript𝐴red𝑋𝐒𝐩𝐞𝐜𝐴\mathbf{Spec}\,A_{\textrm{red}}\to X\times\mathbf{Spec}\,A, induced by the chosen derived point 𝐒𝐩𝐞𝐜AXDR𝐒𝐩𝐞𝐜𝐴subscript𝑋𝐷𝑅\mathbf{Spec}\,A\to X_{DR}, and the canonical map 𝐒𝐩𝐞𝐜Ared𝐒𝐩𝐞𝐜A𝐒𝐩𝐞𝐜subscript𝐴red𝐒𝐩𝐞𝐜𝐴\mathbf{Spec}\,A_{\textrm{red}}\to\mathbf{Spec}\,A (see Proposition 2.2.4).

Properties 1-3 caracterize a family of perfect formal derived stacks over A𝐴A, while property 4 says that the family is algebraisable ([CPTVV, Section 2.1, 2.2.]).

Crystalline structure sheaf and shifted principal parts. Let us consider the \infty-functor

𝔻:𝐝𝐀𝐟𝐟kopϵ𝐜𝐝𝐠𝐚kgr,A𝐃𝐑(Ared/A):𝔻formulae-sequencesubscriptsuperscript𝐝𝐀𝐟𝐟𝑜𝑝𝑘italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘𝐴𝐃𝐑subscript𝐴red𝐴\mathbb{D}:\mathbf{dAff}^{op}_{k}\longrightarrow\epsilon-\mathbf{cdga}^{gr}_{k}\,,\qquad A\longmapsto\mathbf{DR}(A_{\textrm{red}}/A)

(where 𝐃𝐑(Ared/A)𝐃𝐑superscript𝐴red𝐴\mathbf{DR}(A^{\textrm{red}}/A) is defined in Definition 1.1.1, via Remark 1.1.2). Note that 𝔻(A)SymAred(𝕃Ared/A[1])similar-to-or-equals𝔻𝐴𝑆𝑦subscript𝑚subscript𝐴redsubscript𝕃subscript𝐴red𝐴delimited-[]1\mathbb{D}(A)\simeq Sym_{A_{\textrm{red}}}(\mathbb{L}_{A_{\textrm{red}}/A}[-1]) in 𝐜𝐝𝐠𝐚kgrsuperscriptsubscript𝐜𝐝𝐠𝐚𝑘𝑔𝑟\mathbf{cdga}_{k}^{gr} (Section 1.1). The functor 𝔻𝔻\mathbb{D} satisfies étale descent, and thus we have an induced \infty-functor

𝔻:𝐝𝐒𝐭kopϵ𝐜𝐝𝐠𝐚kgr,Flim𝐒𝐩𝐞𝐜AF𝔻(A).:𝔻formulae-sequencesubscriptsuperscript𝐝𝐒𝐭𝑜𝑝𝑘italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘𝐹subscript𝐒𝐩𝐞𝐜𝐴𝐹𝔻𝐴\mathbb{D}:\mathbf{dSt}^{op}_{k}\longrightarrow\epsilon-\mathbf{cdga}^{gr}_{k}\,,\,\,F\longmapsto\lim_{\mathbf{Spec}\,A\to F}\mathbb{D}(A).

We consider the following prestacks of mixed graded cdga’s on 𝐝𝐀𝐟𝐟/XDR𝐝𝐀𝐟𝐟subscript𝑋𝐷𝑅\mathbf{dAff}/X_{DR}

𝔻XDR:=𝔻(𝒪XDR)::assignsubscript𝔻subscript𝑋𝐷𝑅𝔻subscript𝒪subscript𝑋𝐷𝑅absent\displaystyle\mathbb{D}_{X_{DR}}:=\mathbb{D}(\mathcal{O}_{X_{DR}}): (𝐝𝐀𝐟𝐟k/XDR)opϵ𝐜𝐝𝐠𝐚kgr,(𝐒𝐩𝐞𝐜AXDR)𝔻(A),formulae-sequencesuperscriptsubscript𝐝𝐀𝐟𝐟𝑘subscript𝑋𝐷𝑅𝑜𝑝italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘𝐒𝐩𝐞𝐜𝐴subscript𝑋𝐷𝑅𝔻𝐴\displaystyle(\mathbf{dAff}_{k}/X_{DR})^{op}\longrightarrow\epsilon-\mathbf{cdga}^{gr}_{k},\qquad(\mathbf{Spec}\,A\rightarrow X_{DR})\longmapsto\mathbb{D}(A),
𝒫X::subscript𝒫𝑋absent\displaystyle\mathcal{P}_{X}: (𝐝𝐀𝐟𝐟k/XDR)opϵ𝐜𝐝𝐠𝐚kgr,(𝐒𝐩𝐞𝐜AXDR)𝔻(XA).formulae-sequencesuperscriptsubscript𝐝𝐀𝐟𝐟𝑘subscript𝑋𝐷𝑅𝑜𝑝italic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘𝐒𝐩𝐞𝐜𝐴subscript𝑋𝐷𝑅𝔻subscript𝑋𝐴\displaystyle(\mathbf{dAff}_{k}/X_{DR})^{op}\longrightarrow\epsilon-\mathbf{cdga}^{gr}_{k},\qquad(\mathbf{Spec}\,A\rightarrow X_{DR})\longmapsto\mathbb{D}(X_{A}).

Note that there is a natural equivalence 𝒫X(𝐒𝐩𝐞𝐜AXDR)SymAred(𝕃𝐒𝐩𝐞𝐜Ared/XA[1])similar-to-or-equalssubscript𝒫𝑋𝐒𝐩𝐞𝐜𝐴subscript𝑋𝐷𝑅𝑆𝑦subscript𝑚subscript𝐴redsubscript𝕃𝐒𝐩𝐞𝐜subscript𝐴redsubscript𝑋𝐴delimited-[]1\mathcal{P}_{X}(\mathbf{Spec}\,A\rightarrow X_{DR})\simeq Sym_{A_{\textrm{red}}}(\mathbb{L}_{\mathbf{Spec}\,A_{\textrm{red}}/X_{A}}[-1]) in 𝐜𝐝𝐠𝐚kgrsuperscriptsubscript𝐜𝐝𝐠𝐚𝑘𝑔𝑟\mathbf{cdga}_{k}^{gr} ([CPTVV, Proposition 2.2.6]).

Definition 2.2.6

The prestack 𝔻XDRsubscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}} on XDRsubscript𝑋𝐷𝑅X_{DR} is called the crystalline structure sheaf for X𝑋X. The prestack 𝒫Xsubscript𝒫𝑋\mathcal{P}_{X} on XDRsubscript𝑋𝐷𝑅X_{DR} is called the prestack of principal parts for X𝑋X.

Remark 2.2.7

An alternative interpretation of 𝒫Xsubscript𝒫𝑋\mathcal{P}_{X} can be given as follows. As already observed, the canonical map XXDR𝑋subscript𝑋𝐷𝑅X\longrightarrow X_{DR} realizes X𝑋X as a family of formal derived stacks over XDRsubscript𝑋𝐷𝑅X_{DR}, namely as the family of formal completions at closed points of X𝑋X. By [Lu2] these formal completions are determined by a dg-Lie algebra xsubscript𝑥\ell_{x}. The dg-Lie algebra xsubscript𝑥\ell_{x} itself does not extend globally as a sheaf of dg-Lie algebras over XDRsubscript𝑋𝐷𝑅X_{DR}, simply because its underlying complex is 𝕋X[1]subscript𝕋𝑋delimited-[]1\mathbb{T}_{X}[-1], the shifted tangent complex of X𝑋X ([Hen]), does not descend to XDRsubscript𝑋𝐷𝑅X_{DR}. However, a remarkable consequence of derived formal localization (Theorem 2.2.12) is that the Chevalley-Eilenberg complexes of xsubscript𝑥\ell_{x}, for xX𝑥𝑋x\in X, viewed as a graded mixed commutative dg-algebras, do glue to a global object over XDRsubscript𝑋𝐷𝑅X_{DR}. This is exactly 𝒫Xsubscript𝒫𝑋\mathcal{P}_{X}. Then, the Formal Localization Theorem 2.2.12 tells us exactly how to express global geometric objects on X𝑋X as correspondingly sheafified objects on XDRsubscript𝑋𝐷𝑅X_{DR} related to 𝒫Xsubscript𝒫𝑋\mathcal{P}_{X}.

Note that, by functoriality of 𝔻𝔻\mathbb{D}, we have a natural morphism 𝔻XDR𝒫Xsubscript𝔻subscript𝑋𝐷𝑅subscript𝒫𝑋\mathbb{D}_{X_{DR}}\to\mathcal{P}_{X} of prestacks of mixed graded cdga’s on XDRsubscript𝑋𝐷𝑅X_{DR}. In particular, if we consider the \infty-category superscript\mathcal{M}^{\prime} of prestacks on 𝐝𝐀𝐟𝐟k/XDRsubscript𝐝𝐀𝐟𝐟𝑘subscript𝑋𝐷𝑅\mathbf{dAff}_{k}/X_{DR} with values in ϵ𝐝𝐠kgritalic-ϵsubscriptsuperscript𝐝𝐠𝑔𝑟𝑘\epsilon-\mathbf{dg}^{gr}_{k}, then 𝔻XDR𝐂𝐀𝐥𝐠()subscript𝔻subscript𝑋𝐷𝑅𝐂𝐀𝐥𝐠superscript\mathbb{D}_{X_{DR}}\in\mathbf{CAlg}(\mathcal{M}^{\prime}), and 𝒫X𝔻XDR/𝐂𝐀𝐥𝐠()subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝐂𝐀𝐥𝐠superscript\mathcal{P}_{X}\in\mathbb{D}_{X_{DR}}/\mathbf{CAlg}(\mathcal{M}^{\prime}). We let :=𝔻XDR𝐌𝐨𝐝assignsubscript𝔻subscript𝑋𝐷𝑅subscript𝐌𝐨𝐝superscript\mathcal{M}:=\mathbb{D}_{X_{DR}}-\mathbf{Mod}_{\mathcal{M}^{\prime}}. Then 𝒫X𝐂𝐀𝐥𝐠()𝔻XDR/𝐂𝐀𝐥𝐠()subscript𝒫𝑋𝐂𝐀𝐥𝐠similar-to-or-equalssubscript𝔻subscript𝑋𝐷𝑅𝐂𝐀𝐥𝐠superscript\mathcal{P}_{X}\in\mathbf{CAlg}(\mathcal{M})\simeq\mathbb{D}_{X_{DR}}/\mathbf{CAlg}(\mathcal{M}^{\prime}), and, for any n𝑛n\in\mathbb{Z}, we may consider (Section 2.1)

𝐏𝐨𝐥(𝒫X/𝔻XDR,n)=𝐏𝐨𝐥(𝒫X,n),n+1gr𝐀𝐥𝐠,𝐃𝐑(𝒫X/𝔻XDR)=𝐃𝐑(𝒫X)ϵ𝐂𝐀𝐥𝐠gr.formulae-sequencesuperscript𝐏𝐨𝐥superscriptsubscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛superscript𝐏𝐨𝐥subscript𝒫𝑋𝑛subscriptsuperscript𝑔𝑟𝑛1𝐀𝐥𝐠superscript𝐃𝐑superscriptsubscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅superscript𝐃𝐑subscript𝒫𝑋italic-ϵsuperscriptsubscript𝐂𝐀𝐥𝐠𝑔𝑟\mathbf{Pol}^{\mathcal{M}^{\prime}}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}},n)=\mathbf{Pol}^{\mathcal{M}}(\mathcal{P}_{X},n)\in\mathbb{P}^{gr}_{\mathcal{M},n+1}-\mathbf{Alg}\,,\qquad\mathbf{DR}^{\mathcal{M}^{\prime}}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}})=\mathbf{DR}^{\mathcal{M}}(\mathcal{P}_{X})\in\epsilon-\mathbf{CAlg}_{\mathcal{M}}^{gr}\,.

We will also consider the following prestacks on 𝐝𝐀𝐟𝐟/XDR𝐝𝐀𝐟𝐟subscript𝑋𝐷𝑅\mathbf{dAff}/X_{DR} obtained by Tate realizations:

𝐏𝐨𝐥¯t(𝒫X/𝔻XDR,n):(𝐝𝐀𝐟𝐟k/XDR)opn+1𝐜𝐝𝐠𝐚kgr,(𝐒𝐩𝐞𝐜AXDR)|𝐏𝐨𝐥ϵ𝐝𝐠gr(𝒫X(A)/𝔻XDR(A),n)|t:superscript¯𝐏𝐨𝐥𝑡subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛formulae-sequencesuperscriptsubscript𝐝𝐀𝐟𝐟𝑘subscript𝑋𝐷𝑅𝑜𝑝subscript𝑛1superscriptsubscript𝐜𝐝𝐠𝐚𝑘𝑔𝑟maps-to𝐒𝐩𝐞𝐜𝐴subscript𝑋𝐷𝑅superscriptsuperscript𝐏𝐨𝐥italic-ϵsuperscript𝐝𝐠𝑔𝑟subscript𝒫𝑋𝐴subscript𝔻subscript𝑋𝐷𝑅𝐴𝑛𝑡\mathbf{\underline{Pol}}^{t}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}},n):(\mathbf{dAff}_{k}/X_{DR})^{op}\longrightarrow\mathbb{P}_{n+1}-\mathbf{cdga}_{k}^{gr}\,,(\mathbf{Spec}\,A\to X_{DR})\mapsto|\mathbf{Pol}^{\epsilon-\mathbf{dg}^{gr}}(\mathcal{P}_{X}(A)/\mathbb{D}_{X_{DR}}(A),n)|^{t}
𝐃𝐑¯t(𝒫X/𝔻XDR):(𝐝𝐀𝐟𝐟k/XDR)opϵ𝐜𝐝𝐠𝐚kgr,(𝐒𝐩𝐞𝐜AXDR)|𝐃𝐑ϵ𝐝𝐠gr(𝒫X(A)/𝔻XDR(A))|t:superscript¯𝐃𝐑𝑡subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅formulae-sequencesuperscriptsubscript𝐝𝐀𝐟𝐟𝑘subscript𝑋𝐷𝑅𝑜𝑝italic-ϵsuperscriptsubscript𝐜𝐝𝐠𝐚𝑘𝑔𝑟maps-to𝐒𝐩𝐞𝐜𝐴subscript𝑋𝐷𝑅superscriptsuperscript𝐃𝐑italic-ϵsuperscript𝐝𝐠𝑔𝑟subscript𝒫𝑋𝐴subscript𝔻subscript𝑋𝐷𝑅𝐴𝑡\underline{\mathbf{DR}}^{t}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}}):(\mathbf{dAff}_{k}/X_{DR})^{op}\longrightarrow\epsilon-\mathbf{cdga}_{k}^{gr}\,,(\mathbf{Spec}\,A\to X_{DR})\mapsto|\mathbf{DR}^{\epsilon-\mathbf{dg}^{gr}}(\mathcal{P}_{X}(A)/\mathbb{D}_{X_{DR}}(A))|^{t}
Remark 2.2.8

It is worth pointing out that while 𝒫Xsubscript𝒫𝑋\mathcal{P}_{X} and DXDRsubscript𝐷subscript𝑋𝐷𝑅D_{X_{DR}} are not stacks, all 𝐏𝐨𝐥¯t(𝒫X/𝔻XDR,n)superscript¯𝐏𝐨𝐥𝑡subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛\mathbf{\underline{Pol}}^{t}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}},n), 𝐃𝐑(𝒫X/𝔻XDR)𝐃𝐑subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\mathbf{DR}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}}), and 𝐃𝐑¯t(𝒫X/𝔻XDR)superscript¯𝐃𝐑𝑡subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\underline{\mathbf{DR}}^{t}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}}) are stacks ([CPTVV, Corollary 2.4.9]).

Analogously (see Section 2.1), if we consider the \infty-category IndsubscriptsuperscriptInd\mathcal{M}^{\prime}_{\textrm{Ind}} of prestacks on 𝐝𝐀𝐟𝐟k/XDRsubscript𝐝𝐀𝐟𝐟𝑘subscript𝑋𝐷𝑅\mathbf{dAff}_{k}/X_{DR} with values in Ind(ϵ𝐝𝐠kgr)Inditalic-ϵsubscriptsuperscript𝐝𝐠𝑔𝑟𝑘\mathrm{Ind}(\epsilon-\mathbf{dg}^{gr}_{k}), then (see Section 2.1) 𝔻XDR()𝐂𝐀𝐥𝐠(Ind)subscript𝔻subscript𝑋𝐷𝑅𝐂𝐀𝐥𝐠subscriptsuperscriptInd\mathbb{D}_{X_{DR}}(\infty)\in\mathbf{CAlg}(\mathcal{M}^{\prime}_{\textrm{Ind}}), and 𝒫X()𝔻XDR()/𝐂𝐀𝐥𝐠(Ind)subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝐂𝐀𝐥𝐠subscriptsuperscriptInd\mathcal{P}_{X}(\infty)\in\mathbb{D}_{X_{DR}}(\infty)/\mathbf{CAlg}(\mathcal{M}^{\prime}_{\textrm{Ind}}). We let Ind:=𝔻XDR()𝐌𝐨𝐝IndassignsubscriptIndsubscript𝔻subscript𝑋𝐷𝑅subscript𝐌𝐨𝐝subscriptsuperscriptInd\mathcal{M}_{\textrm{Ind}}:=\mathbb{D}_{X_{DR}}(\infty)-\mathbf{Mod}_{\mathcal{M}^{\prime}_{\textrm{Ind}}}. Then 𝒫X()𝐂𝐀𝐥𝐠(Ind)𝔻XDR()/𝐂𝐀𝐥𝐠(Ind)subscript𝒫𝑋𝐂𝐀𝐥𝐠subscriptIndsimilar-to-or-equalssubscript𝔻subscript𝑋𝐷𝑅𝐂𝐀𝐥𝐠subscriptsuperscriptInd\mathcal{P}_{X}(\infty)\in\mathbf{CAlg}(\mathcal{M}_{\textrm{Ind}})\simeq\mathbb{D}_{X_{DR}}(\infty)/\mathbf{CAlg}(\mathcal{M}^{\prime}_{\textrm{Ind}}), and, for any n𝑛n\in\mathbb{Z}, we may consider

𝐏𝐨𝐥Ind(𝒫X()/𝔻XDR(),n)=𝐏𝐨𝐥Ind(𝒫X(),n)Ind,n+1gr𝐀𝐥𝐠,superscript𝐏𝐨𝐥subscriptsuperscriptIndsubscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛superscript𝐏𝐨𝐥subscriptIndsubscript𝒫𝑋𝑛subscriptsuperscript𝑔𝑟subscriptInd𝑛1𝐀𝐥𝐠\mathbf{Pol}^{\mathcal{M}^{\prime}_{\textrm{Ind}}}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty),n)=\mathbf{Pol}^{\mathcal{M}_{\textrm{Ind}}}(\mathcal{P}_{X}(\infty),n)\in\mathbb{P}^{gr}_{\mathcal{M}_{\textrm{Ind}},n+1}-\mathbf{Alg}\,,
𝐃𝐑Ind(𝒫X()/𝔻XDR())=𝐃𝐑Ind(𝒫X())ϵ𝐂𝐀𝐥𝐠Indgr.superscript𝐃𝐑subscriptsuperscriptIndsubscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅superscript𝐃𝐑subscriptIndsubscript𝒫𝑋italic-ϵsuperscriptsubscript𝐂𝐀𝐥𝐠subscriptInd𝑔𝑟\mathbf{DR}^{\mathcal{M}^{\prime}_{\textrm{Ind}}}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty))=\mathbf{DR}^{\mathcal{M}_{\textrm{Ind}}}(\mathcal{P}_{X}(\infty))\in\epsilon-\mathbf{CAlg}_{\mathcal{M}_{\textrm{Ind}}}^{gr}.

And we also have the following prestacks on 𝐝𝐀𝐟𝐟/XDR𝐝𝐀𝐟𝐟subscript𝑋𝐷𝑅\mathbf{dAff}/X_{DR} obtained by standard realizations:

𝐏𝐨𝐥¯(𝒫X()/𝔻XDR(),n):(𝐝𝐀𝐟𝐟k/XDR)op:¯𝐏𝐨𝐥subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛superscriptsubscript𝐝𝐀𝐟𝐟𝑘subscript𝑋𝐷𝑅𝑜𝑝\textstyle{\mathbf{\underline{Pol}}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty),n):(\mathbf{dAff}_{k}/X_{DR})^{op}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}n+1𝐜𝐝𝐠𝐚kgr,subscript𝑛1superscriptsubscript𝐜𝐝𝐠𝐚𝑘𝑔𝑟\textstyle{\mathbb{P}_{n+1}-\mathbf{cdga}_{k}^{gr},}(𝐒𝐩𝐞𝐜AXDR)𝐒𝐩𝐞𝐜𝐴subscript𝑋𝐷𝑅\textstyle{(\mathbf{Spec}\,A\to X_{DR})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}|𝐏𝐨𝐥Ind(ϵ𝐝𝐠gr)(𝒫X()(A)/𝔻XDR()(A),n)|superscript𝐏𝐨𝐥Inditalic-ϵsuperscript𝐝𝐠𝑔𝑟subscript𝒫𝑋𝐴subscript𝔻subscript𝑋𝐷𝑅𝐴𝑛\textstyle{|\mathbf{Pol}^{\mathrm{Ind}(\epsilon-\mathbf{dg}^{gr})}(\mathcal{P}_{X}(\infty)(A)/\mathbb{D}_{X_{DR}}(\infty)(A),n)|}

and

𝐃𝐑¯(𝒫X()/𝔻XDR()):(𝐝𝐀𝐟𝐟k/XDR)op:¯𝐃𝐑subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅superscriptsubscript𝐝𝐀𝐟𝐟𝑘subscript𝑋𝐷𝑅𝑜𝑝\textstyle{\underline{\mathbf{DR}}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty)):(\mathbf{dAff}_{k}/X_{DR})^{op}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϵ𝐜𝐝𝐠𝐚kgritalic-ϵsuperscriptsubscript𝐜𝐝𝐠𝐚𝑘𝑔𝑟\textstyle{\epsilon-\mathbf{cdga}_{k}^{gr}}(𝐒𝐩𝐞𝐜AXDR)𝐒𝐩𝐞𝐜𝐴subscript𝑋𝐷𝑅\textstyle{(\mathbf{Spec}\,A\to X_{DR})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}|𝐃𝐑Ind(ϵ𝐝𝐠gr)(𝒫X()(A)/𝔻XDR()(A))|superscript𝐃𝐑Inditalic-ϵsuperscript𝐝𝐠𝑔𝑟subscript𝒫𝑋𝐴subscript𝔻subscript𝑋𝐷𝑅𝐴\textstyle{|\mathbf{DR}^{\mathrm{Ind}(\epsilon-\mathbf{dg}^{gr})}(\mathcal{P}_{X}(\infty)(A)/\mathbb{D}_{X_{DR}}(\infty)(A))|}
Remark 2.2.9

By Proposition 2.1.20 and Remark 2.1.21, we get equivalences

𝐃𝐑(𝒫X/𝔻XDR)superscript𝐃𝐑superscriptsubscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\displaystyle\mathbf{DR}^{\mathcal{M}^{\prime}}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}}) 𝐃𝐑¯t(𝒫X/𝔻XDR)𝐃𝐑¯(𝒫X()/𝔻XDR()),similar-to-or-equalsabsentsuperscript¯𝐃𝐑𝑡subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅similar-to-or-equals¯𝐃𝐑subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\displaystyle\simeq\underline{\mathbf{DR}}^{t}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}})\simeq\underline{\mathbf{DR}}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty))\,,
𝐏𝐨𝐥¯t(𝒫X/𝔻XDR,n)superscript¯𝐏𝐨𝐥𝑡subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛\displaystyle\mathbf{\underline{Pol}}^{t}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}},n) 𝐏𝐨𝐥¯(𝒫X()/𝔻XDR(),n)similar-to-or-equalsabsent¯𝐏𝐨𝐥subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛\displaystyle\simeq\mathbf{\underline{Pol}}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty),n)

but notice that 𝐏𝐨𝐥(𝒫X/𝔻XDR,n)superscript𝐏𝐨𝐥superscriptsubscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛\mathbf{Pol}^{\mathcal{M}^{\prime}}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}},n) is not in general equivalent to the previous ones.

The Formal Localization theorem. We have already defined the mixed graded k𝑘k-cdgas 𝐃𝐑(X/k)𝐃𝐑𝑋𝑘\mathbf{DR}(X/k), and 𝐃𝐑(X/XDR)𝐃𝐑𝑋subscript𝑋𝐷𝑅\mathbf{DR}(X/X_{DR}) (Definition 1.1.6). It is an easy consequence of the equivalence 𝕃X𝕃X/XDRsimilar-to-or-equalssubscript𝕃𝑋subscript𝕃𝑋subscript𝑋𝐷𝑅\mathbb{L}_{X}\simeq\mathbb{L}_{X/X_{DR}} (Remark 2.2.5), that 𝐃𝐑(X/k)𝐃𝐑(X/XDR)similar-to-or-equals𝐃𝐑𝑋𝑘𝐃𝐑𝑋subscript𝑋𝐷𝑅\mathbf{DR}(X/k)\simeq\mathbf{DR}(X/X_{DR}) in ϵ𝐜𝐝𝐠𝐚kgritalic-ϵsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\epsilon-\mathbf{cdga}^{gr}_{k}. We can give a similar, general definition of shifted polyvectors on X𝑋X, at least as a graded k𝑘k-cdga.

Definition 2.2.10

Let FG𝐹𝐺F\to G be a map between derived stacks, both having cotangent complexes (so that 𝕃F/Gsubscript𝕃𝐹𝐺\mathbb{L}_{F/G} exists, too). We define the graded k𝑘k-dg module of n𝑛n-shifted relative polyvectors as

Pol(F/G,n)p0(Hom¯𝐐𝐂𝐨𝐡(F)(p𝕃F/G,𝒪F[pn]))hΣp𝐜𝐝𝐠𝐚kgr.Pol(F/G,n)\simeq\bigoplus_{p\geq 0}(\underline{Hom}_{\mathbf{QCoh}(F)}(\otimes^{p}\mathbb{L}_{F/G},\mathcal{O}_{F}[-pn]))^{h\Sigma_{p}}\in\mathbf{cdga}^{gr}_{k}\,.

In the above definition, 𝐐𝐂𝐨𝐡(F)𝐐𝐂𝐨𝐡𝐹\mathbf{QCoh}(F) is regarded as a dg-category over k𝑘k, and Hom¯𝐐𝐂𝐨𝐡(F)subscript¯𝐻𝑜𝑚𝐐𝐂𝐨𝐡𝐹\underline{Hom}_{\mathbf{QCoh}(F)} denotes its k𝑘k-dg-module of morphisms; the ΣpsubscriptΣ𝑝\Sigma_{p}-action on Hom¯𝐐𝐂𝐨𝐡(F)(p𝕃F/G,𝒪X[pn])\underline{Hom}_{\mathbf{QCoh}(F)}(\otimes^{p}\mathbb{L}_{F/G},\mathcal{O}_{X}[pn]) is induced by ΣpsubscriptΣ𝑝\Sigma_{p} acting in the standard way on p𝕃F/Gsuperscripttensor-product𝑝absentsubscript𝕃𝐹𝐺\otimes^{p}\mathbb{L}_{F/G}, and via (1)nsuperscript1𝑛(-1)^{n} times the sign representation on 𝒪F[pn]=𝒪F[n]psubscript𝒪𝐹delimited-[]𝑝𝑛subscript𝒪𝐹superscriptdelimited-[]𝑛superscripttensor-product𝑝\mathcal{O}_{F}[-pn]=\mathcal{O}_{F}[-n]^{\otimes^{p}}.

Remark 2.2.11

(1) Again by Remark 2.2.5, we have an equivalence Pol(X/XDR,n)Pol(X/k,n)similar-to-or-equals𝑃𝑜𝑙𝑋subscript𝑋𝐷𝑅𝑛𝑃𝑜𝑙𝑋𝑘𝑛Pol(X/X_{DR},n)\simeq Pol(X/k,n) in 𝐜𝐝𝐠𝐚kgrsubscriptsuperscript𝐜𝐝𝐠𝐚𝑔𝑟𝑘\mathbf{cdga}^{gr}_{k}.

(2) When 𝕃F/Gsubscript𝕃𝐹𝐺\mathbb{L}_{F/G} is perfect over F𝐹F (e.g. for F=X𝐹𝑋F=X derived Artin stack lfp over k𝑘k, and G=XDR𝐺subscript𝑋𝐷𝑅G=X_{DR}), then we may express Pol(F/G,n)𝑃𝑜𝑙𝐹𝐺𝑛Pol(F/G,n) using the dual relative tangent complex 𝕋F/Gsubscript𝕋𝐹𝐺\mathbb{T}_{F/G} as (see Remark 2.1.5)

Pol(F/G,n)p0Γ(F,Symp(𝕋F/G[n])𝐜𝐝𝐠𝐚kgrPol(F/G,n)\simeq\bigoplus_{p\geq 0}\Gamma(F,Sym^{p}(\mathbb{T}_{F/G}[-n])\in\mathbf{cdga}_{k}^{gr}

The problem with Definition 2.2.10 is that, in general, it is impossible to directly endow Pol(F(/G,n)Pol(F(/G,n), as defined, with a bracket and give it the structure of a graded n+1subscript𝑛1\mathbb{P}_{n+1}-differential graded algebras over k𝑘k. This is where the next result comes to rescue.

Theorem 2.2.12

(Formal localization for XXDR𝑋subscript𝑋𝐷𝑅X\to X_{DR})

Let X𝑋X be an Artin derived stack locally of finite presentation over k𝑘k.

  1. 1.

    There is a natural equivalence of \infty-categories

    𝐏𝐞𝐫𝐟(X)𝒫X𝐌𝐨𝐝𝗉𝖾𝗋𝖿,similar-to-or-equals𝐏𝐞𝐫𝐟𝑋subscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝𝗉𝖾𝗋𝖿\mathbf{Perf}(X)\simeq\mathcal{P}_{X}-\mathbf{Mod}_{\mathcal{M}}^{\mathsf{perf}}\,,

    where \mathcal{M} was defined right after Remark 2.2.7, and 𝒫X𝐌𝐨𝐝𝗉𝖾𝗋𝖿subscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝𝗉𝖾𝗋𝖿\mathcal{P}_{X}-\mathbf{Mod}_{\mathcal{M}}^{\mathsf{perf}} is the full sub-\infty-category of 𝒫X𝐌𝐨𝐝subscript𝒫𝑋subscript𝐌𝐨𝐝\mathcal{P}_{X}-\mathbf{Mod}_{\mathcal{M}}, consisting of prestacks E𝐸E of graded mixed 𝒫Xsubscript𝒫𝑋\mathcal{P}_{X}-modules on 𝐝𝐀𝐟𝐟/XDR𝐝𝐀𝐟𝐟subscript𝑋𝐷𝑅\mathbf{dAff}/X_{DR} satisfying the following two conditions:

    • For all 𝐒𝐩𝐞𝐜AXDR𝐒𝐩𝐞𝐜𝐴subscript𝑋𝐷𝑅\mathbf{Spec}\,A\longrightarrow X_{DR}, the graded mixed 𝒫X(A)subscript𝒫𝑋𝐴\mathcal{P}_{X}(A)-module E(A)𝐸𝐴E(A) is equivalent, just as a graded 𝒫X(A)subscript𝒫𝑋𝐴\mathcal{P}_{X}(A)-module, to 𝒫X(A)AredE0subscripttensor-productsubscript𝐴redsubscript𝒫𝑋𝐴subscript𝐸0\mathcal{P}_{X}(A)\otimes_{A_{\textrm{red}}}E_{0}, for some E0𝐏𝐞𝐫𝐟(Ared)subscript𝐸0𝐏𝐞𝐫𝐟subscript𝐴redE_{0}\in\mathbf{Perf}(A_{\textrm{red}}).

    • E𝐸E is quasi-coherent in the sense that: for all 𝐒𝐩𝐞𝐜B𝐒𝐩𝐞𝐜A𝐒𝐩𝐞𝐜𝐵𝐒𝐩𝐞𝐜𝐴\mathbf{Spec}\,B\longrightarrow\mathbf{Spec}\,A in 𝐝𝐀𝐟𝐟k/XDRsubscript𝐝𝐀𝐟𝐟𝑘subscript𝑋𝐷𝑅\mathbf{dAff}_{k}/X_{DR}, the induced morphism E(A)𝒫X(A)𝒫X(B)E(B)subscripttensor-productsubscript𝒫𝑋𝐴𝐸𝐴subscript𝒫𝑋𝐵𝐸𝐵E(A)\otimes_{\mathcal{P}_{X}(A)}\mathcal{P}_{X}(B)\longrightarrow E(B) is an equivalence.

  2. 2.

    There are natural equivalences of graded mixed cdga’s over k𝑘k

    𝐃𝐑(X/XDR)𝐃𝐑(X/k)Γ(XDR,𝐃𝐑¯(𝒫X/𝔻XDR))Γ(XDR,𝐃𝐑¯t(𝒫X/𝔻XDR)).similar-to-or-equals𝐃𝐑𝑋subscript𝑋𝐷𝑅𝐃𝐑𝑋𝑘similar-to-or-equalsΓsubscript𝑋𝐷𝑅¯𝐃𝐑subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅similar-to-or-equalsΓsubscript𝑋𝐷𝑅superscript¯𝐃𝐑𝑡subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\mathbf{DR}(X/X_{DR})\simeq\mathbf{DR}(X/k)\simeq\Gamma(X_{DR},\underline{\mathbf{DR}}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}}))\simeq\Gamma(X_{DR},\underline{\mathbf{DR}}^{t}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}})).

    (where ΓΓ\Gamma denotes derived global sections , i.e. Γ(XDR,)=lim𝐒𝐩𝐞𝐜AXDR(A)Γsubscript𝑋𝐷𝑅subscript𝐒𝐩𝐞𝐜𝐴subscript𝑋𝐷𝑅𝐴\Gamma(X_{DR},\mathcal{F})=\lim_{\mathbf{Spec}\,A\to X_{DR}}\mathcal{F}(A), the limit being taken in the \infty-category where \mathcal{F} is valued).

  3. 3.

    For each n𝑛n\in\mathbb{Z}, there are natural equivalences of graded dg-modules over k𝑘k

    Pol(X/XDR,n)Pol(X/k,n)Γ(XDR,𝐏𝐨𝐥¯t(𝒫X/𝔻XDR,n))Γ(XDR,𝐏𝐨𝐥¯(𝒫X()/𝔻XDR(),n)).similar-to-or-equals𝑃𝑜𝑙𝑋subscript𝑋𝐷𝑅𝑛𝑃𝑜𝑙𝑋𝑘𝑛similar-to-or-equalsΓsubscript𝑋𝐷𝑅superscript¯𝐏𝐨𝐥𝑡subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛similar-to-or-equalsΓsubscript𝑋𝐷𝑅¯𝐏𝐨𝐥subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛Pol(X/X_{DR},n)\simeq Pol(X/k,n)\simeq\Gamma(X_{DR},\underline{\mathbf{Pol}}^{t}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}},n))\simeq\Gamma(X_{DR},\underline{\mathbf{Pol}}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty),n)).
  4. 4.

    Let IndsubscriptInd\mathcal{M}_{\mathrm{Ind}} be the \infty-category defined right after Remark 2.2.8. The natural \infty-functor

    𝒫X𝐌𝐨𝐝𝗉𝖾𝗋𝖿𝒫X()𝐌𝐨𝐝Ind𝗉𝖾𝗋𝖿,subscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝𝗉𝖾𝗋𝖿subscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝subscriptInd𝗉𝖾𝗋𝖿\mathcal{P}_{X}-\mathbf{Mod}_{\mathcal{M}}^{\mathsf{perf}}\longrightarrow\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}}^{\mathsf{perf}},

    induced by the base change ()k()tensor-product𝑘(-)\otimes k(\infty), is an equivalence.

  5. 5.

    For each n𝑛n\in\mathbb{Z}, there are canonical equivalences in 𝐓𝐓\mathbf{T}

    𝖲𝗒𝗆𝗉(X,n)𝖲𝗒𝗆𝗉(𝒫X/𝔻XDR,n)𝖲𝗒𝗆𝗉(𝒫X()/𝔻XDR(),n)similar-to-or-equals𝖲𝗒𝗆𝗉𝑋𝑛𝖲𝗒𝗆𝗉subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛similar-to-or-equals𝖲𝗒𝗆𝗉subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛\mathsf{Symp}(X,n)\simeq\mathsf{Symp}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}},n)\simeq\mathsf{Symp}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty),n)

    where 𝖲𝗒𝗆𝗉(𝒫X/𝔻XDR,n)𝖲𝗒𝗆𝗉subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛\mathsf{Symp}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}},n) and 𝖲𝗒𝗆𝗉(𝒫X()/𝔻XDR(),n)𝖲𝗒𝗆𝗉subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛\mathsf{Symp}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty),n) are defined as in Definition 2.1.2 (with \mathcal{M} and IndsubscriptInd\mathcal{M}_{\textrm{Ind}}, respectively, as defined in the previous paragraph).

The proof of Theorem 2.2.12 can be found in [CPTVV, Corollary 2.4.12, Proposition 2.4.15].

Remark 2.2.13

Let us explain why the formal localization Theorem 2.2.12 is important and useful.

(1) Points 1 and 4 in Theorem 2.2.12 are absolutely crucial since they allow us to completely recover perfect complexes on X𝑋X as certain, explicitly identified, mixed graded modules over 𝒫Xsubscript𝒫𝑋\mathcal{P}_{X} or 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty). This makes manipulations on perfect complexes much easier, and will allow us to go from a quantization of 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty) to a quantization of 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) (see Section 3.1).

(2) Point 2 in the above Theorem lets us completely recover (with its full structure of mixed graded algebra) the descent-theoretic definition 1.1.6 of the de Rham algebra on X𝑋X in terms of 𝒫X/𝔻XDRsubscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\mathcal{P}_{X}/\mathbb{D}_{X_{DR}}.

(3) Point 3 is important because it allows us not only to recover the ’geometrical” polyvectors of Definition 2.2.10, but also to establish a full graded n+1subscript𝑛1\mathbb{P}_{n+1}-algebra structure on them. This is essential in order to be able to define shifted Poisson structures on X𝑋X (Definition 2.3.1).

(4) The last point of Theorem 2.2.12 lets us completely recover shifted symplectic forms on X𝑋X in terms of shifted symplectic form on 𝒫X/𝔻XDRsubscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\mathcal{P}_{X}/\mathbb{D}_{X_{DR}} (or 𝒫X()/𝔻XDR()subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty)). This have the effect to enable a definition of shifted Poisson and shifted symplectic structures in terms of the very same object 𝒫X/𝔻XDRsubscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\mathcal{P}_{X}/\mathbb{D}_{X_{DR}} (or 𝒫X()/𝔻XDR()subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty)), thus opening the way for a comparison between them (see Theorem 2.4.2).

2.3 Shifted Poisson structures

We are finally ready to define shifted Poisson structures on a derived Artin stack X𝑋X lfp over k𝑘k. In the previous Section (Definition 2.2.6), we constructed two prestacks 𝔻XDRsubscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}}, and 𝒫X:=𝔻X/XDRassignsubscript𝒫𝑋subscript𝔻𝑋subscript𝑋𝐷𝑅\mathcal{P}_{X}:=\mathbb{D}_{X/X_{DR}} of graded mixed cdga’s on XDRsubscript𝑋𝐷𝑅X_{DR} , together with a map of prestacks 𝔻XDR𝒫Xsubscript𝔻subscript𝑋𝐷𝑅subscript𝒫𝑋\mathbb{D}_{X_{DR}}\to\mathcal{P}_{X}, exhibiting 𝒫Xsubscript𝒫𝑋\mathcal{P}_{X} as a prestack of 𝔻XDRsubscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}}-linear graded mixed cdga’s on XDRsubscript𝑋𝐷𝑅X_{DR}. By passing to Tate realization, we obtain the prestack 𝐏𝐨𝐥¯t(𝒫X/𝔻XDR,n)superscript¯𝐏𝐨𝐥𝑡subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛\mathbf{\underline{Pol}}^{t}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}},n) of graded n+1subscript𝑛1\mathbb{P}_{n+1}-cdga’ on 𝐝𝐀𝐟𝐟/XDR𝐝𝐀𝐟𝐟subscript𝑋𝐷𝑅\mathbf{dAff}/X_{DR}.

Definition 2.3.1

If X𝑋X is a derived Artin stack lfp over k𝑘k, and n𝑛n\in\mathbb{Z}, we define

  • the graded n+1subscript𝑛1\mathbb{P}_{n+1}-cdga over k𝑘k, 𝖯𝗈𝗅(X,n):=Γ(XDR,𝐏𝐨𝐥¯t(𝒫X/𝔻XDR,n))assign𝖯𝗈𝗅𝑋𝑛Γsubscript𝑋𝐷𝑅superscript¯𝐏𝐨𝐥𝑡subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛\mathsf{Pol}(X,n):=\Gamma(X_{DR},\mathbf{\underline{Pol}}^{t}(\mathcal{P}_{X}/\mathbb{D}_{X_{DR}},n)) of n𝑛n-shifted polyvectors on X𝑋X;

  • the space 𝖯𝗈𝗂𝗌𝗌(X,n):=𝖬𝖺𝗉𝐝𝐠𝐋𝐢𝐞kgr(k(2)[1],𝖯𝗈𝗅(X,n+1)[n+1])assign𝖯𝗈𝗂𝗌𝗌𝑋𝑛subscript𝖬𝖺𝗉subscriptsuperscript𝐝𝐠𝐋𝐢𝐞𝑔𝑟𝑘𝑘2delimited-[]1𝖯𝗈𝗅𝑋𝑛1delimited-[]𝑛1\mathsf{Poiss}(X,n):=\mathsf{Map}_{\mathbf{dgLie}^{gr}_{k}}(k(2)[-1],\mathsf{Pol}(X,n+1)[n+1]) of n𝑛n-shifted Poisson structures on X𝑋X. An n𝑛n-shifted Poisson structure on X𝑋X is an element ππ0(𝖯𝗈𝗂𝗌𝗌(X,n))𝜋subscript𝜋0𝖯𝗈𝗂𝗌𝗌𝑋𝑛\pi\in\pi_{0}(\mathsf{Poiss}(X,n)).

In the second item of the previous definition, 𝐝𝐠𝐋𝐢𝐞kgrsuperscriptsubscript𝐝𝐠𝐋𝐢𝐞𝑘𝑔𝑟\mathbf{dgLie}_{k}^{gr} is the \infty-category of graded k𝑘k-linear dg-Lie algebras, and k(2)[1]𝑘2delimited-[]1k(2)[-1] denotes k𝑘k sitting in cohomological degree 111, in weight degree 222, endowed with trivial bracket and trivial differential.

Since 𝕃Xsubscript𝕃𝑋\mathbb{L}_{X} is perfect, by Theorem 2.2.12 (3) and Remark 2.2.11, there is an equivalence

𝖯𝗈𝗅(X,n)Pol(X/k,n)p0Γ(X,Symp(𝕋X[n])\mathsf{Pol}(X,n)\simeq Pol(X/k,n)\simeq\bigoplus_{p\geq 0}\Gamma(X,Sym^{p}(\mathbb{T}_{X}[-n])

of graded mixed dg-modules over k𝑘k. This justifies the use of the word polyvectors for 𝖯𝗈𝗅(X,n)𝖯𝗈𝗅𝑋𝑛\mathsf{Pol}(X,n).

The intuition behind our definition of 𝖯𝗈𝗂𝗌𝗌(X,n)𝖯𝗈𝗂𝗌𝗌𝑋𝑛\mathsf{Poiss}(X,n) is that if X𝑋X is a smooth scheme, n=0𝑛0n=0, and we replace 𝖬𝖺𝗉𝐝𝐠𝐋𝐢𝐞kgrsubscript𝖬𝖺𝗉subscriptsuperscript𝐝𝐠𝐋𝐢𝐞𝑔𝑟𝑘\mathsf{Map}_{\mathbf{dgLie}^{gr}_{k}} with the usual, strict Hom𝖽𝗀𝗅𝗂𝖾kgr𝐻𝑜subscript𝑚subscriptsuperscript𝖽𝗀𝗅𝗂𝖾𝑔𝑟𝑘Hom_{\mathsf{dglie}^{gr}_{k}}, one gets that an element in Hom𝖽𝗀𝗅𝗂𝖾kgr(k(2)[1],𝖯𝗈𝗅(X,n+1)[n+1])𝐻𝑜subscript𝑚subscriptsuperscript𝖽𝗀𝗅𝗂𝖾𝑔𝑟𝑘𝑘2delimited-[]1𝖯𝗈𝗅𝑋𝑛1delimited-[]𝑛1Hom_{\mathsf{dglie}^{gr}_{k}}(k(2)[-1],\mathsf{Pol}(X,n+1)[n+1]) is exactly a bivector field π𝜋\pi, such that [π,π]=0𝜋𝜋0[\pi,\pi]=0, [,][-,-] being the usual Schouten-Nijenhuis bracket on algebraic polyvector fields on X𝑋X; in other words, such a π𝜋\pi is a usual algebraic Poisson bivector on X𝑋X. See also Example 2.3.3 (1) below.

We can give an alternative description of the space 𝖯𝗈𝗂𝗌𝗌(X,n)𝖯𝗈𝗂𝗌𝗌𝑋𝑛\mathsf{Poiss}(X,n). Recall from the previous section that the stabilized versions 𝔻XDR()subscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}}(\infty) and 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty) of 𝔻XDRsubscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}} and 𝒫Xsubscript𝒫𝑋\mathcal{P}_{X} are both prestacks of commutative monoid objects in 𝖨𝗇𝖽(ϵ𝐝𝐠gr)𝖨𝗇𝖽italic-ϵsuperscript𝐝𝐠𝑔𝑟\mathsf{Ind}(\epsilon-\mathbf{dg}^{gr}) on XDRsubscript𝑋𝐷𝑅X_{DR}, and that there is an analogous canonical map 𝔻XDR()𝒫X()subscript𝔻subscript𝑋𝐷𝑅subscript𝒫𝑋\mathbb{D}_{X_{DR}}(\infty)\to\mathcal{P}_{X}(\infty). We let n+1(𝒫X()/𝔻XDR())subscript𝑛1subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\mathbb{P}_{n+1}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty)) be the space of those n+1subscript𝑛1\mathbb{P}_{n+1}-algebras structures on 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty), in the \infty-category \mathcal{M} of prestacks, on XDRsubscript𝑋𝐷𝑅X_{DR}, of 𝔻XDR()subscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}}(\infty)-modules inside 𝖨𝗇𝖽(ϵ𝐝𝐠gr)𝖨𝗇𝖽italic-ϵsuperscript𝐝𝐠𝑔𝑟\mathsf{Ind}(\epsilon-\mathbf{dg}^{gr}), which are compatible the given commutative algebra structure on 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty) in \mathcal{M}. An elaboration of Corollary 2.1.11 yields the following comparison

Theorem 2.3.2

([CPTVV, Theorem 3.1.2]) For any derived Artin stack X𝑋X lfp over k𝑘k, and any n𝑛n\in\mathbb{Z}, we have a canonical equivalence 𝖯𝗈𝗂𝗌𝗌(X,n),n+1(𝒫X()/𝔻XDR())similar-to-or-equals𝖯𝗈𝗂𝗌𝗌𝑋𝑛subscript𝑛1subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\mathsf{Poiss}(X,n)\simeq\mathbb{P}_{\mathcal{M},n+1}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty)) in 𝐓𝐓\mathbf{T}.

This theorem should be viewed as a vast generalization to derived Artin stacks of the equivalence between the notion of Poisson bivectors on X𝑋X and Poisson brackets on 𝒪Xsubscript𝒪𝑋\mathcal{O}_{X}, well-known for smooth schemes X𝑋X.

Examples 2.3.3

(1) If X𝑋X is a smooth scheme, then the space 𝖯𝗈𝗂𝗌𝗌(X,0)𝖯𝗈𝗂𝗌𝗌𝑋0\mathsf{Poiss}(X,0) is discrete and equivalent to the vector space of usual algebraic Poisson brackets on 𝒪Xsubscript𝒪𝑋\mathcal{O}_{X}.

(2) If G𝐺G is a reductive group scheme over k𝑘k, and 𝔤𝔤\mathfrak{g} its Lie algebra, then one has ([CPTVV, 3.1])

π0(𝖯𝗈𝗂𝗌𝗌(𝖡G,n)){k3(𝔤)G,n=1Symk2(𝔤)G,n=20,n1,2similar-to-or-equalssubscript𝜋0𝖯𝗈𝗂𝗌𝗌𝖡𝐺𝑛casessubscriptsuperscript3𝑘superscript𝔤𝐺𝑛1𝑆𝑦superscriptsubscript𝑚𝑘2superscript𝔤𝐺𝑛20𝑛12\pi_{0}(\mathsf{Poiss}(\mathsf{B}G,n))\simeq\begin{cases}\wedge^{3}_{k}(\mathfrak{g})^{G}\,,&n=1\\ Sym_{k}^{2}(\mathfrak{g})^{G}\,,&n=2\\ 0\,,&n\neq 1,2\end{cases}

The comparison theorem in the next section, together with the coisotropic (combined with the results of Section 1.2), and the intersection theorem in Section 2.5, will provide more examples of shifted Poisson structures.

2.4 Comparison between non degenerate shifted Poisson structures and shifted symplectic structures

In this Section, we establish a derived analog of the usual equivalence between classical non-degenerate Poisson structures and symplectic structures.

Let X𝑋X be a derived Artin stack lfp over k𝑘k, n𝑛n\in\mathbb{Z}, and ππ0(𝖯𝗈𝗂𝗌𝗌(X,n))𝜋subscript𝜋0𝖯𝗈𝗂𝗌𝗌𝑋𝑛\pi\in\pi_{0}(\mathsf{Poiss}(X,n)) be an n𝑛n-shifted Poisson structure on X𝑋X (Definition 2.3.1). By considering the “forgetful” map

𝖬𝖺𝗉𝐝𝐠𝐋𝐢𝐞kgr(k(2)[1],𝖯𝗈𝗅(X,n+1)[n+1])𝖬𝖺𝗉𝐝𝐠kgr(k(2)[1],𝖯𝗈𝗅(X,n+1)[n+1]),subscript𝖬𝖺𝗉subscriptsuperscript𝐝𝐠𝐋𝐢𝐞𝑔𝑟𝑘𝑘2delimited-[]1𝖯𝗈𝗅𝑋𝑛1delimited-[]𝑛1subscript𝖬𝖺𝗉subscriptsuperscript𝐝𝐠𝑔𝑟𝑘𝑘2delimited-[]1𝖯𝗈𝗅𝑋𝑛1delimited-[]𝑛1\mathsf{Map}_{\mathbf{dgLie}^{gr}_{k}}(k(2)[-1],\mathsf{Pol}(X,n+1)[n+1])\longrightarrow\mathsf{Map}_{\mathbf{dg}^{gr}_{k}}(k(2)[-1],\mathsf{Pol}(X,n+1)[n+1]),

together with the equivalence in 𝐝𝐠kgrsubscriptsuperscript𝐝𝐠𝑔𝑟𝑘\mathbf{dg}^{gr}_{k}

𝖯𝗈𝗅(X,n+1)[n+1]pΓ(X,Sym𝒪Xp(𝕋X[n1])[n+1]\mathsf{Pol}(X,n+1)[n+1]\simeq\bigoplus_{p}\Gamma(X,Sym^{p}_{\mathcal{O}_{X}}(\mathbb{T}_{X}[-n-1])[n+1]

(see Remark 2.2.11), π𝜋\pi induces a morphism k(2)pΓ(X,Sym𝒪Xp(𝕋X[n1])[n+2]k(2)\to\oplus_{p}\Gamma(X,Sym^{p}_{\mathcal{O}_{X}}(\mathbb{T}_{X}[-n-1])[n+2] in 𝐝𝐠kgrsubscriptsuperscript𝐝𝐠𝑔𝑟𝑘\mathbf{dg}^{gr}_{k}, and thus defines an element απHn(X,Φn(2)(𝕋X)),subscript𝛼𝜋superscript𝐻𝑛𝑋superscriptsubscriptΦ𝑛2subscript𝕋𝑋\alpha_{\pi}\in H^{-n}(X,\Phi_{n}^{(2)}(\mathbb{T}_{X})), where

Φn(2)(𝕋X):={Sym𝒪X2𝕋X,ifnisodd𝒪X2𝕋X,ifniseven.assignsuperscriptsubscriptΦ𝑛2subscript𝕋𝑋cases𝑆𝑦subscriptsuperscript𝑚2subscript𝒪𝑋subscript𝕋𝑋ifnisoddsubscriptsuperscript2subscript𝒪𝑋subscript𝕋𝑋ifniseven\Phi_{n}^{(2)}(\mathbb{T}_{X}):=\left\{\begin{array}[]{c}Sym^{2}_{\mathcal{O}_{X}}\mathbb{T}_{X},\,\,\mathrm{if\;n\;is\;odd}\\ \wedge^{2}_{\mathcal{O}_{X}}\mathbb{T}_{X},\,\,\mathrm{if\;n\;is\;even.}\end{array}\right.

We denote by πsuperscript𝜋\pi^{\sharp} the map 𝕃X𝕋X[n]subscript𝕃𝑋subscript𝕋𝑋delimited-[]𝑛\mathbb{L}_{X}\to\mathbb{T}_{X}[-n] induced, via adjunction, by απsubscript𝛼𝜋\alpha_{\pi}.

Definition 2.4.1

Let X𝑋X be a derived Artin stack lfp over k𝑘k, and n𝑛n\in\mathbb{Z}. An n𝑛n-shifted Poisson structure ππ0(𝖯𝗈𝗂𝗌𝗌(X,n))𝜋subscript𝜋0𝖯𝗈𝗂𝗌𝗌𝑋𝑛\pi\in\pi_{0}(\mathsf{Poiss}(X,n)) is non-degenerate if the induced map π:𝕃X𝕋X[n]:superscript𝜋subscript𝕃𝑋subscript𝕋𝑋delimited-[]𝑛\pi^{\sharp}:\mathbb{L}_{X}\to\mathbb{T}_{X}[-n] is an equivalence. We denote by 𝖯𝗈𝗂𝗌𝗌nd(X,n)superscript𝖯𝗈𝗂𝗌𝗌𝑛𝑑𝑋𝑛\mathsf{Poiss}^{nd}(X,n) the subspace of 𝖯𝗈𝗂𝗌𝗌(X,n)𝖯𝗈𝗂𝗌𝗌𝑋𝑛\mathsf{Poiss}(X,n) whose connected components are non-degenerate n𝑛n-shifted Poisson structures on X𝑋X.

We are now ready to construct the comparison map between the space of shifted Poisson structures and the space of shifted symplectic structures on derived Artin stacks.
Recalling from the previous Section, let IndsuperscriptsubscriptInd\mathcal{M}_{\textrm{Ind}}^{\prime} is the category of prestacks on 𝐝𝐀𝐟𝐟/XDR𝐝𝐀𝐟𝐟subscript𝑋𝐷𝑅\mathbf{dAff}/X_{DR} with values in 𝖨𝗇𝖽(ϵ𝐝𝐠kgr)𝖨𝗇𝖽italic-ϵsubscriptsuperscript𝐝𝐠𝑔𝑟𝑘\mathsf{Ind}(\epsilon-\mathbf{dg}^{gr}_{k}), and IndsubscriptInd\mathcal{M}_{\textrm{Ind}} the category of DXDR()subscript𝐷subscript𝑋𝐷𝑅D_{X_{DR}}(\infty)-modules in superscript\mathcal{M}^{\prime}. Then A:=𝒫X()assign𝐴subscript𝒫𝑋A:=\mathcal{P}_{X}(\infty) belongs to 𝐂𝐀𝐥𝐠subscript𝐂𝐀𝐥𝐠\mathbf{CAlg}_{\mathcal{M}}. By Proposition 2.1.18 and Th. 2.3.2, we have an equivalence ,n+1nd(𝒫X()/𝔻XDR())𝖯𝗈𝗂𝗌𝗌nd(X,n)similar-to-or-equalssubscriptsuperscript𝑛𝑑𝑛1subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅superscript𝖯𝗈𝗂𝗌𝗌𝑛𝑑𝑋𝑛\mathbb{P}^{nd}_{\mathcal{M},n+1}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty))\simeq\mathsf{Poiss}^{nd}(X,n), where n+1nd(𝒫X()/𝔻XDR())subscriptsuperscript𝑛𝑑𝑛1subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\mathbb{P}^{nd}_{n+1}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty)) is the subspace of non degenerate n+1subscript𝑛1\mathbb{P}_{n+1}-algebra structures on 𝒫X()/𝔻XDR()subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty), compatible with the underlying commutative 𝔻XDR()subscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}}(\infty)-algebra structure on 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty).

Recall from Proposition 2.1.17, that for any \mathcal{M}, and any A𝐂𝐀𝐥𝐠𝐴subscript𝐂𝐀𝐥𝐠A\in\mathbf{CAlg}_{\mathcal{M}} with a dualizable cotangent complex, we have a map ,n+1nd(A)𝖲𝗒𝗆𝗉𝗅(A,n)subscriptsuperscript𝑛𝑑𝑛1𝐴𝖲𝗒𝗆𝗉𝗅𝐴𝑛\mathbb{P}^{nd}_{\mathcal{M},n+1}(A)\to\mathsf{Sympl}(A,n), from the moduli space of those non-degenerate n+1subscript𝑛1\mathbb{P}_{n+1}-algebra structures in \mathcal{M} on A𝐴A which are compatible with the given commutative algebra structure on A𝐴A, to the moduli space of n𝑛n-shifted symplectic structures on A𝐴A. With our current choice of \mathcal{M}, we thus get a map

𝖯𝗈𝗂𝗌𝗌nd(X,n),n+1nd(𝒫X()/𝔻XDR())𝖲𝗒𝗆𝗉𝗅(𝒫X()/𝔻XDR(),n).similar-to-or-equalssuperscript𝖯𝗈𝗂𝗌𝗌𝑛𝑑𝑋𝑛subscriptsuperscript𝑛𝑑𝑛1subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝖲𝗒𝗆𝗉𝗅subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛\mathsf{Poiss}^{nd}(X,n)\simeq\mathbb{P}^{nd}_{\mathcal{M},n+1}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty))\longrightarrow\mathsf{Sympl}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty),n).

But by Theorem 2.2.12, we have 𝖲𝗒𝗆𝗉𝗅(𝒫X()/𝔻XDR(),n)𝖲𝗒𝗆𝗉𝗅(X,n)similar-to-or-equals𝖲𝗒𝗆𝗉𝗅subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝑛𝖲𝗒𝗆𝗉𝗅𝑋𝑛\mathsf{Sympl}(\mathcal{P}_{X}(\infty)/\mathbb{D}_{X_{DR}}(\infty),n)\simeq\mathsf{Sympl}(X,n), so we obtain a comparison map

ψ:𝖯𝗈𝗂𝗌𝗌nd(X,n)𝖲𝗒𝗆𝗉𝗅(X,n):𝜓superscript𝖯𝗈𝗂𝗌𝗌𝑛𝑑𝑋𝑛𝖲𝗒𝗆𝗉𝗅𝑋𝑛\psi:\mathsf{Poiss}^{nd}(X,n)\longrightarrow\mathsf{Sympl}(X,n)

from the space of non-degenerate n𝑛n-shifted Poisson structures on X𝑋X, to the space of n𝑛n-shifted symplectic structures on X𝑋X. One of the main result of [CPTVV], and the key to quantize all the shifted symplectic moduli spaces constructed in Section 1.2, is the following

Theorem 2.4.2

[CPTVV, Theorem 3.2.4] Let X𝑋X be a derived Artin stack lfp over k𝑘k, and n𝑛n\in\mathbb{Z}. The canonical map ψ:𝖯𝗈𝗂𝗌𝗌nd(X,n)𝖲𝗒𝗆𝗉𝗅(X,n):𝜓superscript𝖯𝗈𝗂𝗌𝗌𝑛𝑑𝑋𝑛𝖲𝗒𝗆𝗉𝗅𝑋𝑛\psi:\mathsf{Poiss}^{nd}(X,n)\to\mathsf{Sympl}(X,n) is an equivalence in 𝐓𝐓\mathbf{T}.

In spite of being expected, and somehow very natural, Theorem 2.4.2 has a rather technical and non-trivial proof. It is not too difficult showing that ψ𝜓\psi induces isomorphisms on all the homotopy groups πisubscript𝜋𝑖\pi_{i}’s for i1𝑖1i\geq 1. More difficult is proving that ψ𝜓\psi is an isomorphism on π0subscript𝜋0\pi_{0}: this is achieved in [CPTVV] by first showing that the functors A,n+1nd(𝒫X()(A)/𝔻XDR()(A))maps-to𝐴subscriptsuperscript𝑛𝑑𝑛1subscript𝒫𝑋𝐴subscript𝔻subscript𝑋𝐷𝑅𝐴A\mapsto\mathbb{P}^{nd}_{\mathcal{M},n+1}(\mathcal{P}_{X}(\infty)(A)/\mathbb{D}_{X_{DR}}(\infty)(A)), and A𝖲𝗒𝗆𝗉𝗅(𝒫X()(A)/𝔻XDR()(A),n)maps-to𝐴𝖲𝗒𝗆𝗉𝗅subscript𝒫𝑋𝐴subscript𝔻subscript𝑋𝐷𝑅𝐴𝑛A\mapsto\mathsf{Sympl}(\mathcal{P}_{X}(\infty)(A)/\mathbb{D}_{X_{DR}}(\infty)(A),n) are both formal derived stacks (Definition 2.2.1), and then showing that it is enough to prove the isomorphism on 𝐝𝐀𝐟𝐟red/Xredsuperscript𝐝𝐀𝐟𝐟redsubscript𝑋red\mathbf{dAff}^{\textrm{red}}/X_{\textrm{red}}. This reduced case is then specifically handled by using pairings and copairings on Lsubscript𝐿L_{\infty}-algebras, the Lsubscript𝐿L_{\infty}-algebra being given, for any u:𝐒𝐩𝐞𝐜AX:𝑢𝐒𝐩𝐞𝐜𝐴𝑋u:\mathbf{Spec}\,A\to X, with A𝐴A reduced, by an Lsubscript𝐿L_{\infty}-model for (u𝕃X)[1]superscriptsuperscript𝑢subscript𝕃𝑋delimited-[]1(u^{*}\mathbb{L}_{X})^{\vee}[-1]. Similar techniques have been used in [Co-Gwi].

2.5 Coisotropic structures

In this section we discuss briefly the notion of a coisotrpic structure on a map to a general n𝑛n-shifted Poisson target. This is analogous to the notion of a lagrangian structure from Definition 1.2.3. A new feature of the Poisson context is that the definition of coisotropic structure itself requires a non-trivial statement - Rozenblyum’s additivity theorem. This additivity theorem asserts that for every n1𝑛1n\geq 1 and every symmetric monoidal \infty-category \mathcal{M}, satisfying our standard hypotheses, there is a natural equivalence between the \infty-category ,n+1𝐀𝐥𝐠subscript𝑛1𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg} and the \infty-category 𝐀𝐥𝐠(,n𝐂𝐀𝐥𝐠)𝐀𝐥𝐠subscript𝑛𝐂𝐀𝐥𝐠\mathbf{Alg}(\mathbb{P}_{\mathcal{M},n}-\mathbf{CAlg}) of unital and associative algebras in the category ,n𝐀𝐥𝐠subscript𝑛𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},n}-\mathbf{Alg} 666The same result holds for the operads 𝔼nsubscript𝔼𝑛\mathbb{E}_{n} of little n𝑛n-disks, and it is known as Dunn-Lurie additivity [Lu4, 5.1.2.2].(note that ,nsubscript𝑛\mathbb{P}_{\mathcal{M},n} is a Hopf operad, hence ,n𝐀𝐥𝐠subscript𝑛𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},n}-\mathbf{Alg} inherits a natural symmetric monoidal structure).

The additivity equivalence is functorial in \mathcal{M}, with respect to symmetric monoidal \infty-functors, and commutes with the forgetful functors to \mathcal{M} (see [CPTVV, Remark 3.4.2] for more details). The main utility of the additivity theorem is that it allows us to make sense of n+1subscript𝑛1\mathbb{P}_{n+1}-structures on morphisms between commutative algebras in \mathcal{M}. More precisely, if we write ,(n+1,n)𝐀𝐥𝐠subscript𝑛1𝑛𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},(n+1,n)}-\mathbf{Alg} for the \infty-category of pairs (A,B)𝐴𝐵(A,B) consisting of an object A𝐀𝐥𝐠(,n𝐀𝐥𝐠)𝐴𝐀𝐥𝐠subscript𝑛𝐀𝐥𝐠A\in\mathbf{Alg}(\mathbb{P}_{\mathcal{M},n}-\mathbf{Alg}) and an A𝐴A-module B𝐵B in ,n𝐀𝐥𝐠subscript𝑛subscript𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},n}-\mathbf{Alg}_{\mathcal{M}} , then by the additivity theorem ,(n+1,n)𝐀𝐥𝐠subscript𝑛1𝑛𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},(n+1,n)}-\mathbf{Alg} comes equipped with two forgetful \infty-functors (A,B)Amaps-to𝐴𝐵𝐴(A,B)\mapsto A, and (A,B)Bmaps-to𝐴𝐵𝐵(A,B)\mapsto B to ,n+1𝐀𝐥𝐠subscript𝑛1𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg}, and ,n𝐀𝐥𝐠subscript𝑛𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},n}-\mathbf{Alg}, respectively. Furthermore, the forgetful functor ,n𝐀𝐥𝐠𝐀𝐥𝐠subscript𝑛𝐀𝐥𝐠subscript𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},n}-\mathbf{Alg}\to\mathbf{Alg}_{\mathcal{M}} induces a natural forgetful functor from ,(n+1,n)𝐀𝐥𝐠subscript𝑛1𝑛subscript𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},(n+1,n)}-\mathbf{Alg}_{\mathcal{M}} to the \infty-category of pairs (A,B)𝐴𝐵(A,B) where A𝐀𝐥𝐠(𝐂𝐀𝐥𝐠)𝐴𝐀𝐥𝐠subscript𝐂𝐀𝐥𝐠A\in\mathbf{Alg}(\mathbf{CAlg}_{\mathcal{M}}) and B𝐵B is an A𝐴A-module in 𝐂𝐀𝐥𝐠subscript𝐂𝐀𝐥𝐠\mathbf{CAlg}_{\mathcal{M}}. The \infty-category of such pairs is naturally equivalent to the \infty-category 𝐌𝐨𝐫(𝐂𝐀𝐥𝐠)𝐌𝐨𝐫subscript𝐂𝐀𝐥𝐠\mathbf{Mor}(\mathbf{CAlg}_{\mathcal{M}}) of morphisms in 𝐂𝐀𝐥𝐠subscript𝐂𝐀𝐥𝐠\mathbf{CAlg}_{\mathcal{M}}. In particular, given a morphism φ:AB:𝜑𝐴𝐵\varphi:A\to B between commutative algebras in \mathcal{M}, we can define the space of ,(n+1,n)subscript𝑛1𝑛\mathbb{P}_{\mathcal{M},(n+1,n)}-structures on φ𝜑\varphi as the fiber of the \infty-functor ,(n+1,n)𝐀𝐥𝐠𝐌𝐨𝐫(𝐀𝐥𝐠)subscript𝑛1𝑛𝐀𝐥𝐠𝐌𝐨𝐫subscript𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},(n+1,n)}-\mathbf{Alg}\to\mathbf{Mor}(\mathbf{Alg}_{\mathcal{M}}) over φ𝜑\varphi. We will write

,(n+1,n)(φ):=,(n+1,n)𝐀𝐥𝐠×𝐌𝐨𝐫(𝐂𝐀𝐥𝐠){φ}assignsubscript𝑛1𝑛𝜑subscript𝑛1𝑛subscript𝐌𝐨𝐫subscript𝐂𝐀𝐥𝐠𝐀𝐥𝐠𝜑\mathbb{P}_{\mathcal{M},(n+1,n)}(\varphi):=\mathbb{P}_{\mathcal{M},(n+1,n)}-\mathbf{Alg}\times_{\mathbf{Mor}(\mathbf{CAlg}_{\mathcal{M}})}\{\varphi\}

for this space. Note that by construction the space ,(n+1,n)(φ)subscript𝑛1𝑛𝜑\mathbb{P}_{\mathcal{M},(n+1,n)}(\varphi) projects naturally both to the space ,n+1(A)subscript𝑛1𝐴\mathbb{P}_{\mathcal{M},n+1}(A) of n+1subscript𝑛1\mathbb{P}_{n+1}-structures on the source A𝐴A, and to the space ,n(B)subscript𝑛𝐵\mathbb{P}_{\mathcal{M},n}(B) of nsubscript𝑛\mathbb{P}_{n}-structures on the target B𝐵B.

Let f:XY:𝑓𝑋𝑌f:X\to Y is a morphism of derived stacks locally of finite presentation over k𝑘k. We specialize the construction above to the following case:

  • :=X,Indassignsubscript𝑋Ind\mathcal{M}:=\mathcal{M}_{X,\textrm{Ind}} is the \infty-category defined in Section 2.4 (i.e. if X,Indsuperscriptsubscript𝑋Ind\mathcal{M}_{X,\textrm{Ind}}^{\prime} is the \infty-category of prestacks on 𝐝𝐀𝐟𝐟/XDR𝐝𝐀𝐟𝐟subscript𝑋𝐷𝑅\mathbf{dAff}/X_{DR} with values in 𝖨𝗇𝖽(ϵ𝐝𝐠kgr)𝖨𝗇𝖽italic-ϵsubscriptsuperscript𝐝𝐠𝑔𝑟𝑘\mathsf{Ind}(\epsilon-\mathbf{dg}^{gr}_{k}), then X,Indsubscript𝑋Ind\mathcal{M}_{X,\textrm{Ind}} is the \infty-category of DXDR()subscript𝐷subscript𝑋𝐷𝑅D_{X_{DR}}(\infty)-modules in X,Indsuperscriptsubscript𝑋Ind\mathcal{M}_{X,\textrm{Ind}}^{\prime}).

  • φ𝜑\varphi is the induced map f𝒫:fDR(𝒫Y())𝒫X():superscriptsubscript𝑓𝒫superscriptsubscript𝑓𝐷𝑅subscript𝒫𝑌subscript𝒫𝑋f_{\mathcal{P}}^{*}:f_{DR}^{*}(\mathcal{P}_{Y}(\infty))\to\mathcal{P}_{X}(\infty).

Note that the map fDR(𝔻YDR())𝔻XDR()subscriptsuperscript𝑓𝐷𝑅subscript𝔻subscript𝑌𝐷𝑅subscript𝔻subscript𝑋𝐷𝑅f^{*}_{DR}(\mathbb{D}_{Y_{DR}}(\infty))\to\mathbb{D}_{X_{DR}}(\infty) is an equivalence and so f𝒫superscriptsubscript𝑓𝒫f_{\mathcal{P}}^{*} may indeed be considered as a morphism in 𝐂𝐀𝐥𝐠subscript𝐂𝐀𝐥𝐠\mathbf{CAlg}_{\mathcal{M}}.

Now, if Y𝑌Y is endowed with an n𝑛n-shifted Poisson structure π𝜋\pi, then 𝒫Y()subscript𝒫𝑌\mathcal{P}_{Y}(\infty) is canonically an object in Y,Ind,n+1𝐀𝐥𝐠subscriptsubscript𝑌Ind𝑛1𝐀𝐥𝐠\mathbb{P}_{\mathcal{M}_{Y,\mathrm{Ind}},n+1}-\mathbf{Alg} (Theorem 2.3.2), and thus its pull-back fDR(𝒫Y())superscriptsubscript𝑓𝐷𝑅subscript𝒫𝑌f_{DR}^{*}(\mathcal{P}_{Y}(\infty)) is canonically an object in ,n+1𝐀𝐥𝐠subscript𝑛1𝐀𝐥𝐠\mathbb{P}_{\mathcal{M},n+1}-\mathbf{Alg}. We denote this object by 𝒫Y,fπsubscriptsuperscript𝒫𝜋𝑌𝑓\mathcal{P}^{\pi}_{Y,f} (in order to distinguish it from fDR(𝒫Y())superscriptsubscript𝑓𝐷𝑅subscript𝒫𝑌f_{DR}^{*}(\mathcal{P}_{Y}(\infty)) as an object in 𝐂𝐀𝐥𝐠subscript𝐂𝐀𝐥𝐠\mathbf{CAlg}_{\mathcal{M}}). To ease notation, we will write (n+1,n)(f𝒫)subscript𝑛1𝑛superscriptsubscript𝑓𝒫\mathbb{P}_{(n+1,n)}(f_{\mathcal{P}}^{*}) for ,(n+1,n)(f𝒫)subscript𝑛1𝑛superscriptsubscript𝑓𝒫\mathbb{P}_{\mathcal{M},(n+1,n)}(f_{\mathcal{P}}^{*}), and n+1(fDR(𝒫Y()))subscript𝑛1superscriptsubscript𝑓𝐷𝑅subscript𝒫𝑌\mathbb{P}_{n+1}(f_{DR}^{*}(\mathcal{P}_{Y}(\infty))) for ,n+1(fDR(𝒫Y()))subscript𝑛1superscriptsubscript𝑓𝐷𝑅subscript𝒫𝑌\mathbb{P}_{\mathcal{M},n+1}(f_{DR}^{*}(\mathcal{P}_{Y}(\infty))). So it makes sense to consider the fiber of the projection map (n+1,n)(f𝒫)n+1(fDR(𝒫Y()))subscript𝑛1𝑛superscriptsubscript𝑓𝒫subscript𝑛1superscriptsubscript𝑓𝐷𝑅subscript𝒫𝑌\mathbb{P}_{(n+1,n)}(f_{\mathcal{P}}^{*})\to\mathbb{P}_{n+1}(f_{DR}^{*}(\mathcal{P}_{Y}(\infty))) over 𝒫Y,fπsubscriptsuperscript𝒫𝜋𝑌𝑓\mathcal{P}^{\pi}_{Y,f}.

Definition 2.5.1

Let f:XY:𝑓𝑋𝑌f:X\to Y be a morphism of derived stacks locally of finite presentation over k𝑘k and assume that Y𝑌Y is equipped with an n𝑛n-shifted Poisson structure π𝜋\pi. The space of coisotropic structures on f𝑓f relative to π𝜋\pi is the fiber

𝖢𝗈𝗂𝗌(f,π):=(n+1,n)(f𝒫)×n+1(fDR(𝒫Y())){𝒫Y,fπ}.assign𝖢𝗈𝗂𝗌𝑓𝜋subscriptsubscript𝑛1superscriptsubscript𝑓𝐷𝑅subscript𝒫𝑌subscript𝑛1𝑛superscriptsubscript𝑓𝒫subscriptsuperscript𝒫𝜋𝑌𝑓\mathsf{Cois}(f,\pi):=\mathbb{P}_{(n+1,n)}(f_{\mathcal{P}}^{*})\times_{\mathbb{P}_{n+1}(f_{DR}^{*}(\mathcal{P}_{Y}(\infty)))}\{\mathcal{P}^{\pi}_{Y,f}\}.

A coisotropic structure on f𝑓f relative to π𝜋\pi is an element in π0𝖢𝗈𝗂𝗌(f,π)subscript𝜋0𝖢𝗈𝗂𝗌𝑓𝜋\pi_{0}\,\mathsf{Cois}(f,\pi).

In other words, a coisotropic structure on f:XY:𝑓𝑋𝑌f:X\to Y consists of the datum of a 𝔻XDR()subscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}}(\infty)-linear nsubscript𝑛\mathbb{P}_{n}-algebra structure on 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty) (the target of f𝒫superscriptsubscript𝑓𝒫f_{\mathcal{P}}^{*}), together with a suitably compatible structure of module over 𝒫Y,fπsubscriptsuperscript𝒫𝜋𝑌𝑓\mathcal{P}^{\pi}_{Y,f} (the source of f𝒫superscriptsubscript𝑓𝒫f_{\mathcal{P}}^{*}), inside the \infty-category of 𝔻XDR()subscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}}(\infty)-linear graded mixed nsubscript𝑛\mathbb{P}_{n}-algebras on XDRsubscript𝑋𝐷𝑅X_{DR}.

Remark 2.5.2

This notion of coisotropic structure has the expected geometric behavior:

  • (i)

    Using the other projection map (n+1,n)(f𝒫)n(𝒫X())subscript𝑛1𝑛superscriptsubscript𝑓𝒫subscript𝑛subscript𝒫𝑋\mathbb{P}_{(n+1,n)}(f_{\mathcal{P}}^{*})\to\mathbb{P}_{n}(\mathcal{P}_{X}(\infty)) (i.e. the map keeping only the target of f𝒫superscriptsubscript𝑓𝒫f_{\mathcal{P}}^{*} ), and Theorem 2.3.2, we get that a choice of a coisotropic structure on f:XY:𝑓𝑋𝑌f:X\to Y relative to an n𝑛n-shifted Poisson structure π𝜋\pi on Y𝑌Y, tautologically induces an (n1)𝑛1(n-1)-shifted Poisson structure on the source X𝑋X.

  • (ii)

    If πωsubscript𝜋𝜔\pi_{\omega} is a non-degenerate Poisson structure corresponding to an n𝑛n-shifted symplectic structure ω𝜔\omega on Y𝑌Y, then, keeping the notations of Definition 2.5.1, one expects a natural equivalence of spaces 𝖫𝖺𝗀(f,ω)𝖢𝗈𝗂𝗌(f,π)nd𝖫𝖺𝗀𝑓𝜔𝖢𝗈𝗂𝗌superscript𝑓𝜋nd\mathsf{Lag}(f,\omega)\cong\mathsf{Cois}(f,\pi)^{\text{nd}} between the space of lagrangian structures on f𝑓f (see Definition 1.2.3) and the space of suitably non-degenerate coisotropic structures on f𝑓f. This is being investigated by Melani and Safronov (see [Me-Sa]).

  • (iii)

    The Lagrangian intersection theorem, Theorem 1.2.5 was recently extended to the Poisson context in [Me-Sa]. Suppose (Y,π)𝑌𝜋(Y,\pi) be an n𝑛n-shifted Poisson Artin stack locally of finite presentation over k𝑘k. Let fi:XiY:subscript𝑓𝑖subscript𝑋𝑖𝑌f_{i}:X_{i}\to Y, i=1,2𝑖12i=1,2 be maps of derived Artin stacks each endowed with coisotropic structures relative to π𝜋\pi. Then, Melani and Safronov prove that the derived fiber product X1×YX2subscript𝑌subscript𝑋1subscript𝑋2X_{1}\times_{Y}X_{2} has a natural, induced (n1)𝑛1(n-1)-shifted Poisson structure such that the natural map X1×YX2X1×X2subscript𝑌subscript𝑋1subscript𝑋2subscript𝑋1subscript𝑋2X_{1}\times_{Y}X_{2}\to X_{1}\times X_{2} is a morphism of (n1)𝑛1(n-1)-shifted Poisson stacks, where in the target X2subscript𝑋2X_{2} is endowed with the (n1)𝑛1(n-1)-shifted Poisson structure from point (i) above, and X1subscript𝑋1X_{1} with the corresponding opposite (n1)𝑛1(n-1)-shifted Poisson structure (see [Me-Sa] for details). A classical, i.e. 00-shifted, and purely cohomological precursor of this result was proved in [Gi-Ba]. Aside from its conceptual significance, the coisotropic intersection theorem of [Me-Sa] has many purely utilitarian corollaries. It allows us to extend the list of examples at the end of Section 2.3, by providing many more examples of shifted Poisson structures on moduli stacks, hence of moduli stacks admitting natural deformation quantizations (see Section 3). For instance, recently Spaide [Spa] applied coisotropic inersections to construct and characterize shifted Poisson structures on moduli spaces of framed sheaves in arbitrary dimension as well as on the moduli of monopoles.

3 Deformation quantization

Recall that for an ordinary smooth scheme X𝑋X over k𝑘k, a classical (unshifted) Poisson structure π𝜋\pi can be viewed as the infinitesimal to the deformation of 𝒪Xsubscript𝒪𝑋\mathcal{O}_{X} as a sheaf of associative algebras on X𝑋X. According to the algebraic deformation quantization results of Kontsevich [Ko1] and Yekutieli [Ye] every ordinary Poisson scheme (X,π)𝑋𝜋(X,\pi) admits a quantization as a stack of algebroids. That is we can always find a stack of algebroids 𝒳𝒳\mathcal{X} defined over k[[]]𝑘delimited-[]delimited-[]Planck-constant-over-2-pik[[\hbar]] with (𝒳 mod )=X𝒳 mod Planck-constant-over-2-pi𝑋(\mathcal{X}\text{ mod }\hbar)=X and with infinitesimal π𝜋\pi. Moreover [Ko1, Ye] all possible quantizations with a given infinitesimal depend on a choice of a formality quasi-isomorphism (Drinfeld associator) and are classified by deformation of (X,π)𝑋𝜋(X,\pi) as a Poisson scheme over k[[]]𝑘delimited-[]delimited-[]Planck-constant-over-2-pik[[\hbar]]. In particular the trivial Poisson deformation corresponding to the k[[]]𝑘delimited-[]delimited-[]Planck-constant-over-2-pik[[\hbar]]-linear Poisson bivector πPlanck-constant-over-2-pi𝜋\hbar\cdot\pi gives rise to a preferred quantization of (X,π)𝑋𝜋(X,\pi). This preferred quantization is Kontsevich’s canonical quantization, or in the case of a non-degenerate π𝜋\pi is the algebraic Fedosov canonical quantization of Bezrukavnikov-Kaledin [Bez-Ka].

In this section we discuss the extension of the deformation quantization problem to shifted Poisson structures on derived Artin stacks. We argue that the canonical n𝑛n-shifted quantization always exists as long as n0𝑛0n\neq 0 and again depends on the choice of a Drinfeld associator. Ineterestingly enough the special case when n=0𝑛0n=0 remains the hardest case and the best existing quantization results are still those of [Ko1, Ye]. The natural question of extending the [Ko1, Ye] quantization of smooth Poisson schemes to 00-shifted Poisson derived Artin stacks requires new ideas and will not be treated here.

3.1 Weak and strong quantization

Informally, shifted Poisson structures arise when we study deformations of X𝑋X in which we allow only partial non-commutativity in the deformed product structure. More precisely, an n𝑛n-shifted Poisson structure can be viewed as the infinitesimal for deforming the commutative (=𝔼absentsubscript𝔼=\mathbb{E}_{\infty}) algebra structure on 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty) to an 𝔼n+1subscript𝔼𝑛1\mathbb{E}_{n+1}-algebra structure.

To spell this out, recall that for n1𝑛1n\geq 1 the operad 𝔼nsubscript𝔼𝑛\mathbb{E}_{n} of chains little n𝑛n-dimensional disks is a k𝑘k-linear dg-operad which is given by the chain complexes C(FMn,k)subscript𝐶𝐹subscript𝑀𝑛𝑘C_{\bullet}(FM_{n},k) of the Fulton-MacPherson’s topological operad FMn𝐹subscript𝑀𝑛FM_{n}. By definition, the space of operations of FMn𝐹subscript𝑀𝑛FM_{n} labeled by a finite set I𝐼I is the Fulton-MacPherson compactified configuration space FMn(I)𝐹subscript𝑀𝑛𝐼FM_{n}(I) of I𝐼I-labeled configurations of points in nsuperscript𝑛\mathbb{R}^{n}. For n2𝑛2n\geq 2 the Postnikov tower of the spaces FMn(I)𝐹subscript𝑀𝑛𝐼FM_{n}(I) induces a filtration of 𝔼nsubscript𝔼𝑛\mathbb{E}_{n} whose associated graded is the graded k𝑘k-linear homology operad H(FMn,k)subscript𝐻𝐹subscript𝑀𝑛𝑘H_{\bullet}(FM_{n},k) of FMn𝐹subscript𝑀𝑛FM_{n} which is known to be the operad nsubscript𝑛\mathbb{P}_{n} controlling (n1)𝑛1(n-1)-shifted Poisson algebras. We can now apply the Rees construction to the filtration on 𝔼nsubscript𝔼𝑛\mathbb{E}_{n} to obtain a dg-operad 𝔹𝔻n𝔹subscript𝔻𝑛\mathbb{BD}_{n} (for Beilinson-Drinfeld) which is linear over k[h]𝑘delimited-[]k[h] and deforms the filtered operad 𝔼nsubscript𝔼𝑛\mathbb{E}_{n} to its associated graded nsubscript𝑛\mathbb{P}_{n}. This deformation makes sense for n=1𝑛1n=1 as well. In this case, 𝔼1subscript𝔼1\mathbb{E}_{1} is the operad controlling associative algebra structures. For every finite I𝐼I, the k𝑘k-module of operations 𝔼1(I)subscript𝔼1𝐼\mathbb{E}_{1}(I) labeled by I𝐼I is the non-commutative polynomial algebra kxi|iI𝑘inner-productsubscript𝑥𝑖𝑖𝐼k\langle x_{i}|i\in I\rangle which is naturally filtered by monomial degree. The associated graded to this filtration is the commutative polynomial algebra k[xi|iI]𝑘delimited-[]conditionalsubscript𝑥𝑖𝑖𝐼k[x_{i}|i\in I] equipped with the induced Lie bracket, i.e. we have gr𝔼1(I)=P1(I)grsubscript𝔼1𝐼subscript𝑃1𝐼\text{gr}\mathbb{E}_{1}(I)=P_{1}(I). Again applying the Rees construction to the filtration gives a k[h]𝑘delimited-[]k[h]-linear operad 𝔹𝔻1𝔹subscript𝔻1\mathbb{BD}_{1} which interpolates between 𝔼1subscript𝔼1\mathbb{E}_{1} and 1subscript1\mathbb{P}_{1}. The difference between this case and the case n2𝑛2n\geq 2 is that 1subscript1\mathbb{P}_{1} is not the homology of 𝔼1subscript𝔼1\mathbb{E}_{1}. In fact 𝔼1subscript𝔼1\mathbb{E}_{1} is already formal and isomorphic to H(FM1,k)subscript𝐻𝐹subscript𝑀1𝑘H_{\bullet}(FM_{1},k).

Nevertheless, for any n1𝑛1n\geq 1 we constructed a k[]𝑘delimited-[]Planck-constant-over-2-pik[\hbar]-linear dg-operad operad 𝔹𝔻n𝔹subscript𝔻𝑛\mathbb{BD}_{n} such that 𝔹𝔻nk[]knsubscripttensor-product𝑘delimited-[]Planck-constant-over-2-pi𝔹subscript𝔻𝑛𝑘subscript𝑛\mathbb{BD}_{n}\otimes_{k[\hbar]}k\cong\mathbb{P}_{n} and 𝔹𝔻nk[]k[,1]𝔼n[,1]subscripttensor-product𝑘delimited-[]Planck-constant-over-2-pi𝔹subscript𝔻𝑛𝑘Planck-constant-over-2-pisuperscriptPlanck-constant-over-2-pi1subscript𝔼𝑛Planck-constant-over-2-pisuperscriptPlanck-constant-over-2-pi1\mathbb{BD}_{n}\otimes_{k[\hbar]}k[\hbar,\hbar^{-1}]\cong\mathbb{E}_{n}[\hbar,\hbar^{-1}]. With this in place, we are now ready to formulate the quantization problem in the shifted setting. Suppose X𝑋X is a derived Artin stack, locally of finite presentation over k𝑘k. We use again the notations from Section 2.2.2: IndsubscriptsuperscriptInd\mathcal{M}^{\prime}_{\textrm{Ind}} is the \infty-category of prestacks on 𝐝𝐀𝐟𝐟k/XDRsubscript𝐝𝐀𝐟𝐟𝑘subscript𝑋𝐷𝑅\mathbf{dAff}_{k}/X_{DR} with values in Ind(ϵ𝐝𝐠kgr)Inditalic-ϵsubscriptsuperscript𝐝𝐠𝑔𝑟𝑘\mathrm{Ind}(\epsilon-\mathbf{dg}^{gr}_{k}), so that 𝔻XDR()𝐂𝐀𝐥𝐠(Ind)subscript𝔻subscript𝑋𝐷𝑅𝐂𝐀𝐥𝐠subscriptsuperscriptInd\mathbb{D}_{X_{DR}}(\infty)\in\mathbf{CAlg}(\mathcal{M}^{\prime}_{\textrm{Ind}}), 𝒫X()𝔻XDR()/𝐂𝐀𝐥𝐠(Ind)subscript𝒫𝑋subscript𝔻subscript𝑋𝐷𝑅𝐂𝐀𝐥𝐠subscriptsuperscriptInd\mathcal{P}_{X}(\infty)\in\mathbb{D}_{X_{DR}}(\infty)/\mathbf{CAlg}(\mathcal{M}^{\prime}_{\textrm{Ind}}), and we define Ind:=𝔻XDR()𝐌𝐨𝐝IndassignsubscriptIndsubscript𝔻subscript𝑋𝐷𝑅subscript𝐌𝐨𝐝subscriptsuperscriptInd\mathcal{M}_{\textrm{Ind}}:=\mathbb{D}_{X_{DR}}(\infty)-\mathbf{Mod}_{\mathcal{M}^{\prime}_{\textrm{Ind}}}.

By Theorem 2.3.2 specifying an n𝑛n-shifted Poisson structure π𝜋\pi on X𝑋X is equivalent to specifying a Ind,n+1subscriptsubscriptInd𝑛1\mathbb{P}_{\mathcal{M}_{\textrm{Ind}},n+1}-algebra structure on the Tate stack of principal parts 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty), compatible with its given commutative 𝔻XDR()subscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}}(\infty)-algebra structure. Suppose n0𝑛0n\geq 0, then we have two flavors of the quantization problem:

(strong quantization)

Construct a 𝔻XDR()subscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}}(\infty)-linear 𝔹𝔻n+1𝔹subscript𝔻𝑛1\mathbb{BD}_{n+1}-algebra structure on 𝒫X()k[]tensor-productsubscript𝒫𝑋𝑘delimited-[]Planck-constant-over-2-pi\mathcal{P}_{X}(\infty)\otimes k[\hbar], such that after tensoring with k[]ksubscripttensor-product𝑘delimited-[]Planck-constant-over-2-piabsent𝑘\otimes_{k[\hbar]}k we get the n+1subscript𝑛1\mathbb{P}_{n+1}-structure given by π𝜋\pi.

(weak quantization)

Construct a 𝔹𝔻n𝔹subscript𝔻𝑛\mathbb{BD}_{n}-monoidal structure on the \infty-category (Theorem 2.2.12, (1) and (4))

𝐏𝐞𝐫𝐟(X)kk[]𝒫X()𝐌𝐨𝐝Ind𝗉𝖾𝗋𝖿kk[]subscripttensor-product𝑘𝐏𝐞𝐫𝐟𝑋𝑘delimited-[]Planck-constant-over-2-pisubscript𝒫𝑋subscripttensor-product𝑘superscriptsubscript𝐌𝐨𝐝subscriptInd𝗉𝖾𝗋𝖿𝑘delimited-[]Planck-constant-over-2-pi\mathbf{Perf}(X)\otimes_{k}k[\hbar]\cong\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}}^{\mathsf{perf}}\otimes_{k}k[\hbar]

which after k[]ksubscripttensor-product𝑘delimited-[]Planck-constant-over-2-piabsent𝑘\otimes_{k[\hbar]}k recovers the nsubscript𝑛\mathbb{P}_{n}-monoidal structure on 𝒫X()𝐌𝐨𝐝Ind𝗉𝖾𝗋𝖿subscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝subscriptInd𝗉𝖾𝗋𝖿\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}}^{\mathsf{perf}} corresponding to π𝜋\pi via Rozenblyum’s additivity theorem.

Remark 3.1.1

It is natural to expect that a solution to the strong quantization problem yields a solution to the weak quantization problem by passing to the category of perfect complexes over the 𝔹𝔻n+1𝔹subscript𝔻𝑛1\mathbb{BD}_{n+1}-algebra provided by the strong quantization. For this to make sense we need a 𝔹𝔻𝔹𝔻\mathbb{BD}-version of the additivity theorem. In other words we need to know that, for any k𝑘k-linear presentable stable symmetric monoidal \infty-category 𝒩𝒩\mathcal{N}, there exists a natural equivalence of \infty-categories:

(3.1) 𝔹𝔻n+1𝐀𝐥𝐠𝒩𝐀𝐥𝐠(𝔹𝔻n𝐀𝐥𝐠𝒩)𝔹subscript𝔻𝑛1subscript𝐀𝐥𝐠𝒩𝐀𝐥𝐠𝔹subscript𝔻𝑛subscript𝐀𝐥𝐠𝒩\mathbb{BD}_{n+1}-\mathbf{Alg}_{\mathcal{N}}\cong\mathbf{Alg}\left(\mathbb{BD}_{n}-\mathbf{Alg}_{\mathcal{N}}\right)

which specializes to Rozenblyum’s additivity at =0Planck-constant-over-2-pi0\hbar=0 and to Dunn-Lurie’s additivity from [Lu4, 5.1.2.2] at =1Planck-constant-over-2-pi1\hbar=1. Rozenblyum recently proved that the additivity equivalence (3.1) exists and so to any the strong shifted quantization we can indeed associate a weak shifted quantization.

Our main result in this setting is the following unobstructedness theorem

Theorem 3.1.2

([CPTVV, Theorem 3.5.4]) Let X𝑋X be a derived Artin stack locally of finite presentation over k𝑘k, equipped with an n𝑛n-shifted Poisson structure π𝜋\pi. If n>0𝑛0n>0 then there is a canonical strong quantization.

The above theorem is analogous to the existence of canonical quantization for unshifted smooth schemes. In fact, at this stage, the proof of the theorem is almost a tautology. Since the operad 𝔼n+1subscript𝔼𝑛1\mathbb{E}_{n+1} is formal and for n>0𝑛0n>0 its homology is n+1subscript𝑛1\mathbb{P}_{n+1}, we can choose a formality equivalence of k𝑘k-dg-operads αn+1:𝔼n+1n+1:subscript𝛼𝑛1similar-to-or-equalssubscript𝔼𝑛1subscript𝑛1\alpha_{n+1}:\mathbb{E}_{n+1}\simeq\mathbb{P}_{n+1}. The map αn+1subscript𝛼𝑛1\alpha_{n+1} induces an equivalence 𝔹𝔻n+1n+1kk[]similar-to-or-equals𝔹subscript𝔻𝑛1subscripttensor-product𝑘subscript𝑛1𝑘delimited-[]Planck-constant-over-2-pi\mathbb{BD}_{n+1}\simeq\mathbb{P}_{n+1}\otimes_{k}k[\hbar] which is the identity mod Planck-constant-over-2-pi\hbar. Therefore one can consider 𝒫X()kk[]subscripttensor-product𝑘subscript𝒫𝑋𝑘delimited-[]Planck-constant-over-2-pi\mathcal{P}_{X}(\infty)\otimes_{k}k[\hbar] as a stack of 𝔻XDR()subscript𝔻subscript𝑋𝐷𝑅\mathbb{D}_{X_{DR}}(\infty)-linear graded mixed 𝔹𝔻n+1𝔹subscript𝔻𝑛1\mathbb{BD}_{n+1}-algebras on XDRsubscript𝑋𝐷𝑅X_{DR}, and by construction this stack is a strong deformation quantization of 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty).

By specializing the 𝔹𝔻n+1𝔹subscript𝔻𝑛1\mathbb{BD}_{n+1}-algebra structure at =1Planck-constant-over-2-pi1\hbar=1 we can then view 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty) as a 𝔼n+1subscript𝔼𝑛1\mathbb{E}_{n+1}-algebra in IndsubscriptInd\mathcal{M}_{\textrm{Ind}}. By [Lu4, 5.1.2.2 and 5.1.2.7], the \infty-category 𝒫X()𝐌𝐨𝐝Indsubscript𝒫𝑋subscript𝐌𝐨𝐝subscriptInd\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}} has an induced 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-monoidal structure, and one checks that its full sub-category 𝒫X()𝐌𝐨𝐝Ind𝗉𝖾𝗋𝖿subscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝subscriptInd𝗉𝖾𝗋𝖿\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}}^{\mathsf{perf}} inherits an 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-monoidal structure that we will denote by (𝒫X()𝐌𝐨𝐝Ind𝗉𝖾𝗋𝖿)𝔼n,πsubscriptsubscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝subscriptInd𝗉𝖾𝗋𝖿subscript𝔼𝑛𝜋(\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}}^{\mathsf{perf}})_{\mathbb{E}_{n},\,\pi}. The subscript π𝜋\pi indicates that the 𝔼n+1subscript𝔼𝑛1\mathbb{E}_{n+1}-algebra structure on 𝒫X()subscript𝒫𝑋\mathcal{P}_{X}(\infty), hence the induced 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-monoidal structure on 𝒫X()𝐌𝐨𝐝Ind𝗉𝖾𝗋𝖿subscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝subscriptInd𝗉𝖾𝗋𝖿\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}}^{\mathsf{perf}}, depends on π𝜋\pi, while the subscript 𝔼nsubscript𝔼𝑛\mathbb{E}_{n} records the 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-monoidal structure. This is exactly the deformation of 𝐏𝐞𝐫𝐟(X)𝒫X()𝐌𝐨𝐝Ind𝗉𝖾𝗋𝖿𝐏𝐞𝐫𝐟𝑋subscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝subscriptInd𝗉𝖾𝗋𝖿\mathbf{Perf}(X)\cong\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}}^{\mathsf{perf}} that we were looking for. We record this fact in the following

Definition 3.1.3

With the notation above, and n>0𝑛0n>0, the weak quantization of X𝑋X with infinitesimal π𝜋\pi is the 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-monoidal \infty-category

𝐏𝐞𝐫𝐟(X,π):=(𝒫X()𝐌𝐨𝐝Ind𝗉𝖾𝗋𝖿)𝔼n,πassign𝐏𝐞𝐫𝐟𝑋𝜋subscriptsubscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝subscriptInd𝗉𝖾𝗋𝖿subscript𝔼𝑛𝜋\mathbf{Perf}(X,\pi):=(\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}}^{\mathsf{perf}})_{\mathbb{E}_{n},\,\pi}

Quantization for n<0𝑛0n<0.  The quantization problem for n𝑛n-shifted Poisson structures with n<0𝑛0n<0 can be understood by looking at parameter spaces which are themselves dg schemes. Concretely, let n<0𝑛0n<0 and let π𝜋\pi be an n𝑛n-shifted Poisson structure on some derived stack X𝑋X lfp over k𝑘k. Let 2nsubscriptPlanck-constant-over-2-pi2𝑛\hbar_{2n} a formal variable of cohomological degree 2n2𝑛2n, and consider the stack 𝒫X()[2n]subscript𝒫𝑋delimited-[]subscriptPlanck-constant-over-2-pi2𝑛\mathcal{P}_{X}(\infty)[\hbar_{2n}] of Ind-objects in graded k()[2n]𝑘delimited-[]subscriptPlanck-constant-over-2-pi2𝑛k(\infty)[\hbar_{2n}]-linear mixed cdgas over XDRsubscript𝑋𝐷𝑅X_{DR}. Because of the homological shift it is equipped with a k()[2n]𝑘delimited-[]subscriptPlanck-constant-over-2-pi2𝑛k(\infty)[\hbar_{2n}]-linear 1nsubscript1𝑛\mathbb{P}_{1-n}-structure, induced by 2nπsubscriptPlanck-constant-over-2-pi2𝑛𝜋\hbar_{2n}\cdot\pi (Theorem 2.3.2). Since n<0𝑛0n<0, this brings us back to the situation of positively shifted Poisson structures.

Proceeding as before, we choose a formality equivalence of k𝑘k-dg-operads α1n:𝔼1n1n:subscript𝛼1𝑛similar-to-or-equalssubscript𝔼1𝑛subscript1𝑛\alpha_{1-n}:\mathbb{E}_{1-n}\simeq\mathbb{P}_{1-n}, and thus view 𝒫X()[2n]subscript𝒫𝑋delimited-[]subscriptPlanck-constant-over-2-pi2𝑛\mathcal{P}_{X}(\infty)[\hbar_{2n}] as a k()[2n]𝑘delimited-[]subscriptPlanck-constant-over-2-pi2𝑛k(\infty)[\hbar_{2n}]-linear 𝔼1nsubscript𝔼1𝑛\mathbb{E}_{1-n}-algebra. Again by using Dunn-Lurie’s additivity [Lu4, 5.1.2.2 and 5.1.2.7], the \infty-category 𝒫X()𝐌𝐨𝐝Ind𝗉𝖾𝗋𝖿subscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝subscriptInd𝗉𝖾𝗋𝖿\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}}^{\mathsf{perf}} comes equipped with an induced 𝔼nsubscript𝔼𝑛\mathbb{E}_{-n}-monoidal structure. We will write (𝒫X()𝐌𝐨𝐝Ind𝗉𝖾𝗋𝖿)𝔼n,πsubscriptsubscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝subscriptInd𝗉𝖾𝗋𝖿subscript𝔼𝑛𝜋(\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}}^{\mathsf{perf}})_{\mathbb{E}_{-n},\,\pi} for this 𝔼nsubscript𝔼𝑛\mathbb{E}_{-n}-monoidal category. Thus for n<0𝑛0n<0 we can define the weak quantization of X𝑋X with infinitesimal π𝜋\pi as the 𝔼nsubscript𝔼𝑛\mathbb{E}_{-n}-monoidal \infty-category

𝐏𝐞𝐫𝐟(X,π):=(𝒫X()𝐌𝐨𝐝Ind𝗉𝖾𝗋𝖿)𝔼n,π.assign𝐏𝐞𝐫𝐟𝑋𝜋subscriptsubscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝subscriptInd𝗉𝖾𝗋𝖿subscript𝔼𝑛𝜋\mathbf{Perf}(X,\pi):=(\mathcal{P}_{X}(\infty)-\mathbf{Mod}_{\mathcal{M}_{\textrm{Ind}}}^{\mathsf{perf}})_{\mathbb{E}_{-n},\,\pi}.

As before, the underlying \infty-category of 𝐏𝐞𝐫𝐟(X,π)𝐏𝐞𝐫𝐟𝑋𝜋\mathbf{Perf}(X,\pi) is 𝐏𝐞𝐫𝐟(X)kk[2n]=:𝐏𝐞𝐫𝐟(X)[2n]\mathbf{Perf}(X)\otimes_{k}k[\hbar_{2n}]=:\mathbf{Perf}(X)[\hbar_{2n}]. Hence for n<0𝑛0n<0 our weak quantization of X𝑋X consists then of the datum of a 𝔼nsubscript𝔼𝑛\mathbb{E}_{-n}-monoidal structure on 𝐏𝐞𝐫𝐟(X)[2n]𝐏𝐞𝐫𝐟𝑋delimited-[]subscriptPlanck-constant-over-2-pi2𝑛\mathbf{Perf}(X)[\hbar_{2n}], and by the strong version of Rozenblyum’s additivity, such a quantization can be considered as a deformation of the standard symmetric (i.e. 𝔼subscript𝔼\mathbb{E}_{\infty}-) monoidal structure on 𝐏𝐞𝐫𝐟(X)[2n]𝐏𝐞𝐫𝐟𝑋delimited-[]subscriptPlanck-constant-over-2-pi2𝑛\mathbf{Perf}(X)[\hbar_{2n}]. Note that this standard symmetric monoidal structure on 𝐏𝐞𝐫𝐟(X)[2n]𝐏𝐞𝐫𝐟𝑋delimited-[]subscriptPlanck-constant-over-2-pi2𝑛\mathbf{Perf}(X)[\hbar_{2n}] recovers the standard symmetric monoidal structure on 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) after base change along the canonical map k[2n]k𝑘delimited-[]subscriptPlanck-constant-over-2-pi2𝑛𝑘k[\hbar_{2n}]\rightarrow k.

Remark 3.1.4

This quantization answers a conjecture of Kapustin [Kap, 3.2] which concerns the n=1𝑛1n=-1 case. Note that Kapustin considers /22\mathbb{Z}/2-graded derived categories, and therefore the fact that we work over k[2n]𝑘delimited-[]subscriptPlanck-constant-over-2-pi2𝑛k[\hbar_{2n}] is immaterial: we really obtain a quantization of the /22\mathbb{Z}/2 perfect derived category of X𝑋X, since 𝐏𝐞𝐫𝐟(X)[2n]𝐏𝐞𝐫𝐟𝑋delimited-[]subscriptPlanck-constant-over-2-pi2𝑛\mathbf{Perf}(X)[\hbar_{2n}] and 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) coincide after /22\mathbb{Z}/2-periodization.

3.2 Examples and vistas

(a)  Quantization of moduli. The equivalence of shifted symplectic and non-degenerate shifted Poisson structures from Theorem 2.4.2 combined with the n>0𝑛0n>0 (or n<0𝑛0n<0) quantization scheme described in the previous section provides a canonical 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-monoidal (or 𝔼nsubscript𝔼𝑛\mathbb{E}_{-n}-monoidal) deformation of the \infty-category of perfect complexes on the various shifted symplectic moduli stacks listed at the end of section 2.3. For example:

  • For a derived Artin stack X𝑋X locally of finite presentation we obtain a canonical quantization of the shifted cotangent stack TX[n]superscript𝑇𝑋delimited-[]𝑛T^{*}X[n] for n0𝑛0n\neq 0. The shifted cotangent stack TX[n]superscript𝑇𝑋delimited-[]𝑛T^{*}X[n] has a natural n𝑛n-shifted symplectic form [Cal2]. If we denote the corresponding non-degenerate n𝑛n-shifted Poisson structure by πnsubscript𝜋𝑛\pi_{n}, then the modules over the 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-monoidal (or 𝔼nsubscript𝔼𝑛\mathbb{E}_{-n}-monoidal) category 𝐏𝐞𝐫𝐟(TX[n],πn)𝐏𝐞𝐫𝐟superscript𝑇𝑋delimited-[]𝑛subscript𝜋𝑛\mathbf{Perf}(T^{*}X[n],\pi_{n}) will be the modules over the n𝑛n-shifted differential operators on X𝑋X.

  • For a complex reductive group G𝐺G we obtain canonical quantizations of:

    • the derived stack 𝖫𝗈𝖼G(M)subscript𝖫𝗈𝖼𝐺𝑀\mathbb{R}\mathsf{Loc}_{G}(M) of G𝐺G-local systems on a compact oriented topological manifold M𝑀M of dimension 2absent2\neq 2;

    • the derived stack 𝖫𝗈𝖼GDR(X)superscriptsubscript𝖫𝗈𝖼𝐺𝐷𝑅𝑋\mathbb{R}\mathsf{Loc}_{G}^{DR}(X) of algebraic G𝐺G-local systems on a smooth complex projective variety X𝑋X of dimension >1absent1>1;

    • the derived stack 𝖧𝗂𝗀𝗀𝗌G(X)subscript𝖧𝗂𝗀𝗀𝗌𝐺𝑋\mathbb{R}\mathsf{Higgs}_{G}(X) of algebraic G𝐺G-Higgs bundles on a smooth complex projective variety X𝑋X of dimension >1absent1>1;

    • the derived stack 𝖡𝗎𝗇G(X)subscript𝖡𝗎𝗇𝐺𝑋\mathbb{R}\mathsf{Bun}_{G}(X) of algebraic G𝐺G-torsors on a smooth compact Calabi-Yau variety X𝑋X of dimension 2absent2\neq 2.

    Similarly we get quantizations of the stack of perfect complexes on a compact oriented topological manifold M𝑀M of dimension 2absent2\neq 2, of the derived stack of perfect complexes over XDRsubscript𝑋𝐷𝑅X_{DR} for a smooth complex projective variety X𝑋X of dimension >1absent1>1, of the derived stack of perfect complexes of Higgs bundles on a smooth complex projective variety of dimension >1absent1>1, and on the derived stack of perfect complexes on a smooth compact Calabi-Yau variety of dimension 2absent2\neq 2.

  • For a smooth compact Calabi-Yau dg category T𝑇T of dimension 2absent2\neq 2 we get a canonical weak quantization of the derived moduli stack Tsubscript𝑇\mathcal{M}_{T} of compact objects in T𝑇T. For instance we can take T𝑇T to be the category of graded matrix factorizations of a cubic polynomial f𝑓f in 3n3𝑛3n variables with n2𝑛2n\neq 2. Applying the general quantization procedure to this setting we get an (n2)𝑛2(n-2)-shifted quantization of the moduli stack of graded matrix factorizations of f𝑓f.

(b)  Quantization formally at a point.  Let (X,π)𝑋𝜋(X,\pi) be an n𝑛n-shifted Poisson derived Artin stack locally of finite presentation and let x:=𝐒𝐩𝐞𝐜kXx:*=\mathbf{Spec}\,k\to X be a closed point. It can be checked [CPTVV, Lemma 3.6.1] that any n𝑛n-shifted Poisson structure on X𝑋X induces an n𝑛n-shifted Poisson structure on the formal completion X^xsubscript^𝑋𝑥\widehat{X}_{x} at x𝑥x.

As a (non-mixed) graded cdga over k𝑘k, 𝒫X^xsubscript𝒫subscript^𝑋𝑥\mathcal{P}_{\widehat{X}_{x}} is equivalent to

Sym(𝕃/X^x[1])Sym(x𝕃X^x)Sym(x𝕃X).𝑆𝑦𝑚subscript𝕃absentsubscript^𝑋𝑥delimited-[]1𝑆𝑦𝑚superscript𝑥subscript𝕃subscript^𝑋𝑥𝑆𝑦𝑚superscript𝑥subscript𝕃𝑋Sym(\mathbb{L}_{*/\widehat{X}_{x}}[-1])\cong Sym(x^{*}\mathbb{L}_{\widehat{X}_{x}})\cong Sym(x^{*}\mathbb{L}_{X})\,.

We therefore get a graded mixed n+1subscript𝑛1\mathbb{P}_{n+1}-algebra structure on Sym(x𝕃X)𝑆𝑦𝑚superscript𝑥subscript𝕃𝑋Sym(x^{*}\mathbb{L}_{X}), whose underlying graded mixed cdgas is 𝒫X^xsubscript𝒫subscript^𝑋𝑥\mathcal{P}_{\widehat{X}_{x}}. After a choice of formality αn+1subscript𝛼𝑛1\alpha_{n+1}, we get a graded mixed 𝔼n+1subscript𝔼𝑛1\mathbb{E}_{n+1}-structure on Sym(x𝕃X)𝑆𝑦𝑚superscript𝑥subscript𝕃𝑋Sym(x^{*}\mathbb{L}_{X}) whenever n>0𝑛0n>0. When π𝜋\pi is non-degenerate at x𝑥x and the induced n𝑛n-shifted Poisson structure on 𝒫X^xsubscript𝒫subscript^𝑋𝑥\mathcal{P}_{\widehat{X}_{x}} is strict and constant, then the graded mixed 𝔼n+1subscript𝔼𝑛1\mathbb{E}_{n+1}-structure on Sym(x𝕃X)𝑆𝑦𝑚superscript𝑥subscript𝕃𝑋Sym(x^{*}\mathbb{L}_{X}) can be described explicitly in terms of Kontsevich’s graph complex [CPTVV, 3.6.1]. When the underlying dg-Lie algebra is formal, the explicit formula then identifies the 𝔼n+1subscript𝔼𝑛1\mathbb{E}_{n+1}-structure on Sym(x𝕃X)𝑆𝑦𝑚superscript𝑥subscript𝕃𝑋Sym(x^{*}\mathbb{L}_{X}) with the Weyl n𝑛n-algebra recently introduced by Markarian [Mar].

(c)  Quantization of BG𝐵𝐺BG.  Suppose G𝐺G is an affine group scheme, and let X=BG𝑋𝐵𝐺X=BG be the classifying stack of G𝐺G. Note that XDR=B(GDR)subscript𝑋𝐷𝑅𝐵subscript𝐺𝐷𝑅X_{DR}=B(G_{DR}). Let x:BGx:*\to BG be the classifying map of the unit e:Ge:*\to G. We have a fiber sequence of groups G^eGGDRsubscript^𝐺𝑒𝐺subscript𝐺𝐷𝑅\widehat{G}_{e}\longrightarrow G\longrightarrow G_{DR}, and hence BG^xB(G^e)similar-to-or-equalssubscript^𝐵𝐺𝑥𝐵subscript^𝐺𝑒\widehat{BG}_{x}\simeq B(\widehat{G}_{e}).

As we noted in (b) the pull-back of 𝒫Xsubscript𝒫𝑋\mathcal{P}_{X} along xDR:BGDRx_{DR}:*\to BG_{DR} is 𝒫X^xsubscript𝒫subscript^𝑋𝑥\mathcal{P}_{\widehat{X}_{x}}. Thus the symmetric monoidal \infty-category

𝐏𝐞𝐫𝐟(BG)𝒫X𝐌𝐨𝐝ϵ𝐝𝐠gr𝐏𝐞𝐫𝐟similar-to-or-equals𝐏𝐞𝐫𝐟𝐵𝐺subscript𝒫𝑋superscriptsubscript𝐌𝐨𝐝italic-ϵsuperscript𝐝𝐠𝑔𝑟𝐏𝐞𝐫𝐟\mathbf{Perf}(BG)\simeq\mathcal{P}_{X}-\mathbf{Mod}_{\epsilon-\mathbf{dg}^{gr}}^{\mathbf{Perf}}

is equivalent to the symmetric monoidal \infty-category of GDRsubscript𝐺𝐷𝑅G_{DR}-equivariant objects in

𝐏𝐞𝐫𝐟(BG^e)𝒫X^x𝐌𝐨𝐝ϵ𝐝𝐠gr𝐏𝐞𝐫𝐟.similar-to-or-equals𝐏𝐞𝐫𝐟𝐵subscript^𝐺𝑒subscript𝒫subscript^𝑋𝑥superscriptsubscript𝐌𝐨𝐝italic-ϵsuperscript𝐝𝐠𝑔𝑟𝐏𝐞𝐫𝐟\mathbf{Perf}(B\widehat{G}_{e})\simeq\mathcal{P}_{\widehat{X}_{x}}-\mathbf{Mod}_{\epsilon-\mathbf{dg}^{gr}}^{\mathbf{Perf}}\,.

In view of this the quantization of an n𝑛n-shifted Poisson structure on BG𝐵𝐺BG will be determined completely by the GDRsubscript𝐺𝐷𝑅G_{DR}-equivariant graded mixed 𝔼n+1subscript𝔼𝑛1\mathbb{E}_{n+1}-algebra structure on 𝒫X^xsubscript𝒫subscript^𝑋𝑥\mathcal{P}_{\widehat{X}_{x}} obtained from the equivalence αn+1:n+1𝔼n+1:subscript𝛼𝑛1similar-to-or-equalssubscript𝑛1subscript𝔼𝑛1\alpha_{n+1}:\mathbb{P}_{n+1}\simeq\mathbb{E}_{n+1}.

This algebra structure can be analyzed in concrete terms. Before we look more closely at the 111 and 222-shifted cases it is useful to observe that as a graded mixed cdga over k𝑘k the algebra 𝒫X^x𝔻(BG^e)similar-to-or-equalssubscript𝒫subscript^𝑋𝑥𝔻𝐵subscript^𝐺𝑒\mathcal{P}_{\widehat{X}_{x}}\simeq\mathbb{D}(B\widehat{G}_{e}) admits an explicit description. Indeed, in [CPTVV, 3.6.2] it is proven that 𝔻(BG^e)𝔻𝐵subscript^𝐺𝑒\mathbb{D}(B\widehat{G}_{e}) is actually equivalent to the Chevalley-Eilenberg graded mixed cdga 𝖢𝖤(𝔤)𝖢𝖤𝔤\mathsf{CE}(\mathfrak{g}) of the Lie algebra 𝔤=𝖫𝗂𝖾(G)𝔤𝖫𝗂𝖾𝐺\mathfrak{g}=\mathsf{Lie}(G).

The case n=1𝑛1n=1 for a reductive G𝐺G. For a reductive group G𝐺G the 111-shifted Poisson structures on BG𝐵𝐺BG are simply elements in 3(𝔤)Gsuperscript3superscript𝔤𝐺\wedge^{3}(\mathfrak{g})^{G}. If π𝜋\pi is such an element, then the induced 111-shifted Poisson structure on the graded mixed cdga 𝖢𝖤(𝔤)𝖢𝖤𝔤\mathsf{CE}(\mathfrak{g}) is given explicitly as a semi-strict 2subscript2\mathbb{P}_{2}-structure (see [Me]): all structure 222-shifted polyvectors are trivial except for the 333-ary one which is constant and given by π𝜋\pi.

The weak 111-shifted deformation quantization in particular gives rise to a deformation of the category 𝖱𝖾𝗉fd(𝔤)superscript𝖱𝖾𝗉𝑓𝑑𝔤\mathsf{Rep}^{fd}(\mathfrak{g}) of finite dimensional representation of 𝔤𝔤\mathfrak{g} as a monoidal category. For specific choices of π𝜋\pi we recover familiar monoidal deformations:

Example 3.2.1

Given a non-degenerate invariant pairing ,\langle\,,\,\rangle on 𝔤𝔤\mathfrak{g}, we can choose π𝜋\pi as the dual of the G𝐺G-invariant linear form

3𝔤k,(x,y,z)x,[y,z].formulae-sequencesuperscript3𝔤𝑘𝑥𝑦𝑧𝑥𝑦𝑧\wedge^{3}\mathfrak{g}\longrightarrow k,\quad\quad(x,y,z)\longmapsto\langle x,[y,z]\rangle\,.

Alternatively, any invariant symmetric 222-tensor tSym2(𝔤)G𝑡𝑆𝑦superscript𝑚2superscript𝔤𝐺t\in Sym^{2}(\mathfrak{g})^{G} leads to such an element π=[t1,2,t2,3]3(𝔤)G𝜋superscript𝑡12superscript𝑡23superscript3superscript𝔤𝐺\pi=[t^{1,2},t^{2,3}]\in\wedge^{3}(\mathfrak{g})^{G}. In these cases the deformation of 𝖱𝖾𝗉fd(𝔤)superscript𝖱𝖾𝗉𝑓𝑑𝔤\mathsf{Rep}^{fd}(\mathfrak{g}) as a monoidal category can be obtained by means of a deformation of the associativity constraint only (see [Dr1]), which then looks like

Φ=13+2π+o(2)U(𝔤)3[[]].Φsuperscript1tensor-productabsent3superscriptPlanck-constant-over-2-pi2𝜋𝑜superscriptPlanck-constant-over-2-pi2𝑈superscript𝔤tensor-productabsent3delimited-[]delimited-[]Planck-constant-over-2-pi\Phi=1^{\otimes 3}+\hbar^{2}\pi+o(\hbar^{2})\in U(\mathfrak{g})^{\otimes 3}[[\hbar]]\,.

The case n=2𝑛2n=2 for a reductive G𝐺G. For a reductive group G𝐺G the equivalences classes of 222-shifted Poisson structures on BG𝐵𝐺BG are in bijection with elements tSym2(𝔤)G𝑡𝑆𝑦superscript𝑚2superscript𝔤𝐺t\in Sym^{2}(\mathfrak{g})^{G}. The induced 222-shifted Poisson structure on the graded mixed cdga 𝖢𝖤(𝔤)𝖢𝖤𝔤\mathsf{CE}(\mathfrak{g}) is strict and constant. The graded mixed 𝔼3subscript𝔼3\mathbb{E}_{3}-structure on 𝖢𝖤(𝔤)𝖢𝖤𝔤\mathsf{CE}(\mathfrak{g}) given by our deformation quantization then takes the form of a Weyl 333-algebra, as described in [Mar].

Note that this graded mixed 𝔼3subscript𝔼3\mathbb{E}_{3}-structure is GDRsubscript𝐺𝐷𝑅G_{DR}-equivariant by construction, so it leads to an 𝔼2subscript𝔼2\mathbb{E}_{2}-monoidal deformation of 𝖯𝖾𝗋𝖿(BG)𝖯𝖾𝗋𝖿𝐵𝐺\mathsf{Perf}(BG). This in particular leads to a braided monoidal deformation of 𝖱𝖾𝗉fd(𝔤)superscript𝖱𝖾𝗉𝑓𝑑𝔤\mathsf{Rep}^{fd}(\mathfrak{g}).

Such deformation quantizations of BG𝐵𝐺BG have already been constructed:

  • when 𝔤𝔤\mathfrak{g} is reductive and t𝑡t is non-degenerate, by means of purely algebraic methods: the quantum group U(𝔤)subscript𝑈Planck-constant-over-2-pi𝔤U_{\hbar}(\mathfrak{g}) is an explicit deformation of the enveloping algebra U(𝔤)𝑈𝔤U(\mathfrak{g}) as a quasi-triangular Hopf algebra.

  • without any assumption, by Drinfeld [Dr2], using transcendental methods similar to the ones that are crucial in the proof of the formality of 𝔼2subscript𝔼2\mathbb{E}_{2}.

It is known that Drinfeld’s quantization is equivalent to the quantum group one in the semi-simple case (see e.g. [Ka] and references therein).

Remark 3.2.2

It is interesting to note that our quantization, in contrast to Drinfeld’s, relies on the formality of 𝔼3subscript𝔼3\mathbb{E}_{3} rather than on the formality of 𝔼2subscript𝔼2\mathbb{E}_{2}.

(d)  Relative and absolute quantization.  An important question that is not addressed in this paper or in [CPTVV] is the question of quantizing lagrangian structures on maps with shifted symplectic targets or quantizing coisotropic structures on maps with shifted Poisson targets. The quantization problem in this relative stting can be formulated in a manner similar to the absolute quantization from section 3.1.

Suppose (Y,π)𝑌𝜋(Y,\pi) is an n𝑛n-shifted derived Artin stack locally of finite presentation, and f:XY:𝑓𝑋𝑌f:X\to Y is a morphism of derived stacks furnished with a coisotropic structure κ𝜅\kappa relative to π𝜋\pi. The \infty categories 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) and 𝐏𝐞𝐫𝐟(Y)𝐏𝐞𝐫𝐟𝑌\mathbf{Perf}(Y) are symmetric monoidal categories and via the pullback functor f:𝐏𝐞𝐫𝐟(Y)𝐏𝐞𝐫𝐟(X):superscript𝑓𝐏𝐞𝐫𝐟𝑌𝐏𝐞𝐫𝐟𝑋f^{*}:\mathbf{Perf}(Y)\to\mathbf{Perf}(X) the category 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) becomes a module over 𝐏𝐞𝐫𝐟(Y)𝐏𝐞𝐫𝐟𝑌\mathbf{Perf}(Y) so that its 𝔼subscript𝔼\mathbb{E}_{\infty}-monoidal structure becomes linear over the 𝔼subscript𝔼\mathbb{E}_{\infty}-monoidal structure on 𝐏𝐞𝐫𝐟(Y)𝐏𝐞𝐫𝐟𝑌\mathbf{Perf}(Y). In other words fsuperscript𝑓f^{*} makes 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) is an 𝔼subscript𝔼\mathbb{E}_{\infty}-algebra over 𝐏𝐞𝐫𝐟(Y)𝐏𝐞𝐫𝐟𝑌\mathbf{Perf}(Y). Assume for simplicity n>0𝑛0n>0. Then the weak quantization of (Y,π)𝑌𝜋(Y,\pi) gives a deformation of 𝐏𝐞𝐫𝐟(Y)𝐏𝐞𝐫𝐟𝑌\mathbf{Perf}(Y) to an 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-monoidal category 𝐏𝐞𝐫𝐟(Y,π)𝐏𝐞𝐫𝐟𝑌𝜋\mathbf{Perf}(Y,\pi). The weak quantization problem for f𝑓f is to find a concurrent deformation of 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) as an algebra over 𝐏𝐞𝐫𝐟(Y)𝐏𝐞𝐫𝐟𝑌\mathbf{Perf}(Y). In other words, we need to deform 𝐏𝐞𝐫𝐟(X)𝐏𝐞𝐫𝐟𝑋\mathbf{Perf}(X) to an 𝔼n1subscript𝔼𝑛1\mathbb{E}_{n-1}-monoidal category 𝐏𝐞𝐫𝐟(X,κ)𝐏𝐞𝐫𝐟𝑋𝜅\mathbf{Perf}(X,\kappa), so that the functor fsuperscript𝑓f^{*} deforms to a functor 𝐪f:𝐏𝐞𝐫𝐟(Y,π)𝐏𝐞𝐫𝐟(X,κ):𝐪superscript𝑓𝐏𝐞𝐫𝐟𝑌𝜋𝐏𝐞𝐫𝐟𝑋𝜅\mathbf{q}f^{*}:\mathbf{Perf}(Y,\pi)\to\mathbf{Perf}(X,\kappa) exhibiting 𝐏𝐞𝐫𝐟(X,κ)𝐏𝐞𝐫𝐟𝑋𝜅\mathbf{Perf}(X,\kappa) as a module, inside 𝔼n1subscript𝔼𝑛1\mathbb{E}_{n-1}-monoidal \infty-categories, over 𝐏𝐞𝐫𝐟(Y,π)𝐏𝐞𝐫𝐟𝑌𝜋\mathbf{Perf}(Y,\pi), viewed as an 𝔼1subscript𝔼1\mathbb{E}_{1}-algebra in 𝔼n1subscript𝔼𝑛1\mathbb{E}_{n-1}-monoidal categories. Here again we use Dunn-Lurie additivity [Lu4, §5.1.2] asserting the equivalence between 𝔼1subscript𝔼1\mathbb{E}_{1}-algberas in 𝔼n1subscript𝔼𝑛1\mathbb{E}_{n-1}-algebras and 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-algebras, in any base symmetric monoidal \infty-category.

Note that the extension from the absolute to the relative case is not tautological as the swiss-cheese operad, which governs the deformations of pairs of an 𝔼nsubscript𝔼𝑛\mathbb{E}_{n} algebra and an 𝔼n1subscript𝔼𝑛1\mathbb{E}_{n-1} module over it, is not formal. Nevertheless we expect that such relative quantizations are again unobstructed for n>0𝑛0n>0 and we are currently investigating the problem.

Another interesting problem in this regard is the question of compatibility of quantizations with our standard constructions. A simple instance of this goes as follows. Suppose Y𝑌Y is an n𝑛n-shifted symplectic derived stack (with n>1𝑛1n>1) and suppose f1:X1Y:subscript𝑓1subscript𝑋1𝑌f_{1}:X_{1}\to Y and f2:X2Y:subscript𝑓2subscript𝑋2𝑌f_{2}:X_{2}\to Y be two morphisms equipped with lagrangian structures. The derived intersection Z=X1×YhX2𝑍subscriptsuperscript𝑌subscript𝑋1subscript𝑋2Z=X_{1}\times^{h}_{Y}X_{2} carries a natural (n1)𝑛1(n-1)-shifted symplectic form. We expect that the derived intersection persists in quantizations, i.e. that absolute quantization of Z𝑍Z is the homotopy fiber product of the relative quantizations of f1subscript𝑓1f_{1} and f2subscript𝑓2f_{2}. More precisely, write π𝜋\pi for the non-degenerate shifted Poisson structure corresponding to the symplectic structure on Y𝑌Y and η𝜂\eta for the induced (n1)𝑛1(n-1)-shifted non-degenerate Poisson structure on Z𝑍Z. Let κ2π0𝖢𝗈𝗂𝗌(f2,π)subscript𝜅2subscript𝜋0𝖢𝗈𝗂𝗌subscript𝑓2𝜋\kappa_{2}\in\pi_{0}\mathsf{Cois}(f_{2},\pi) be the non-degenerate coisotropic structure relative to π𝜋\pi corresponding to the lagrangian structure on f2subscript𝑓2f_{2}, and let κ1π0𝖢𝗈𝗂𝗌(f1,π)subscript𝜅1subscript𝜋0𝖢𝗈𝗂𝗌subscript𝑓1𝜋\kappa_{1}\in\pi_{0}\mathsf{Cois}(f_{1},-\pi) be the non-degenerate coisotropic structure relative to π𝜋-\pi corresponding to the lagrangian structure on f1subscript𝑓1f_{1}. Then 𝐏𝐞𝐫𝐟(X2,κ2)𝐏𝐞𝐫𝐟subscript𝑋2subscript𝜅2\mathbf{Perf}(X_{2},\kappa_{2}) is an 𝔼n1subscript𝔼𝑛1\mathbb{E}_{n-1}-algebra over the 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-algebra 𝐏𝐞𝐫𝐟(Y,π)𝐏𝐞𝐫𝐟𝑌𝜋\mathbf{Perf}(Y,\pi). Similarly 𝐏𝐞𝐫𝐟(X1,κ1)𝐏𝐞𝐫𝐟subscript𝑋1subscript𝜅1\mathbf{Perf}(X_{1},\kappa_{1}) is an 𝔼n1subscript𝔼𝑛1\mathbb{E}_{n-1}-algebra over the 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-algebra 𝐏𝐞𝐫𝐟(Y,π)𝐏𝐞𝐫𝐟𝑌𝜋\mathbf{Perf}(Y,-\pi) or equivalently 𝐏𝐞𝐫𝐟(X1,κ1)𝐏𝐞𝐫𝐟subscript𝑋1subscript𝜅1\mathbf{Perf}(X_{1},\kappa_{1}) is an 𝔼n1subscript𝔼𝑛1\mathbb{E}_{n-1}-algebra over the opposite of the 𝔼nsubscript𝔼𝑛\mathbb{E}_{n}-algebra 𝐏𝐞𝐫𝐟(Y,π)𝐏𝐞𝐫𝐟𝑌𝜋\mathbf{Perf}(Y,\pi). Conjecturally the quantized 𝔼n1subscript𝔼𝑛1\mathbb{E}_{n-1}-monoidal category 𝐏𝐞𝐫𝐟(Z,η)𝐏𝐞𝐫𝐟𝑍𝜂\mathbf{Perf}(Z,\eta) is reconstructed from the 𝔼n1subscript𝔼𝑛1\mathbb{E}_{n-1}-monoidal category 𝐏𝐞𝐫𝐟(X1,κ1)𝐏𝐞𝐫𝐟(Y,π)𝐏𝐞𝐫𝐟(X2,κ2)subscripttensor-product𝐏𝐞𝐫𝐟𝑌𝜋𝐏𝐞𝐫𝐟subscript𝑋1subscript𝜅1𝐏𝐞𝐫𝐟subscript𝑋2subscript𝜅2\mathbf{Perf}(X_{1},\kappa_{1})\otimes_{\mathbf{Perf}(Y,\pi)}\mathbf{Perf}(X_{2},\kappa_{2}).

(e)  Vistas.  We conclude our paper by short list of few other directions of investigation, just to stimulate the reader’s interest.

First of all we would like to mention the proposal of [Cal, 4.2.2] for a category of lagrangian correspondences 𝐋𝐚𝐠𝐫nsubscript𝐋𝐚𝐠𝐫𝑛\mathbf{Lagr}_{n}, based on Thm. 1.2.5. Roughly speaking its objects are n𝑛n-shifted symplectic stacks, and morphisms from X𝑋X to Y𝑌Y are maps LX×Y𝐿𝑋𝑌L\to X\times Y equipped with lagrangian structures (where X𝑋X is considered with the opposite of the given symplectic structure). Some truncated versions of this category were already considered in [Amo-BenB]. The details of a complete construction of 𝐋𝐚𝐠𝐫nsubscript𝐋𝐚𝐠𝐫𝑛\mathbf{Lagr}_{n} as an \infty-category together with its natural symmetric monoidal structure remain to be written down, but we have no doubts that such a construction exists. This will be an important step toward the study of extended TQFT’s (as in [Lu6]) with values in lagrangian correspondences. Building on Remark 2.5.2, one can imagine an analogous construction by replacing shifted symplectic stacks with shifted Poisson stacks, and lagrangian correspondences with coisotropic correspondences. Such a construction is currently being considered by Melani and Safronov, and it might shed some light on Weinstein’s original proposal [We].

Another promising research program related to the topics treated in this review is the one being pursued, since a few years, by D. Joyce and various collaborators. As part of their project, they use shifted symplectic structures to study Donaldson-Thomas moduli spaces of Calabi-Yau fourfolds, vanishing cycles, and various categorifications of Donaldson-Thomas invariants (see e.g. [BBDJS, Jo-Sa]).

The geometry of coisotropic structures for shifted Poisson structures is in a very early stage of development, and a lot of new phenomena need to be properly explored. Just to give one example, the identity map is always endowed with a canonical coisotropic structure, and this produces a map from n𝑛n-shifted Poisson structures to (n1)𝑛1(n-1)-shifted Poisson structures, which is worth investigating. If not trivial, this could e.g. connect the n=2𝑛2n=2 to the n=1𝑛1n=1 examples of quantizations of BG𝐵𝐺BG (c) above. More generally, having a definition of coisotropic structures that is equivalent to Definition 2.5.1 but avoids reference to the additivity theorem, would be very useful, especially in applications. Some important steps in this direction have been done by P. Safronov [Saf], and more recently by V. Melani and P. Safronov [Me-Sa].

Finally, it would be interesting to have a version of Theorem 1.2.1 with target 𝖯𝖾𝗋𝖿𝖯𝖾𝗋𝖿\mathsf{Perf} and a stratified topological space as a source. A possible way to include the stratification in our theory is through the use of the MacPherson-Lurie exit paths \infty-category [Lu4, A.6]. If this can be accomplished, then a corresponding relative version, as in Remark 1.2.7, could be relevant for some ideas and conjectures about moduli spaces of constructible sheaves with singular support in a legendrian knot (see [STZ]). A promising preliminary step in this direction is the theory of left and right Calabi-Yau structures on functors developed recently by Toën, and Brav-Dyckerhoff ([To2, p. 228], [Br-Dyc]).

References

  • [AKSZ] M. Alexandrov, M. Kontsevich, A. Schwarz, O. Zaboronsky, The geometry of the master equation and topological quantum field theory, Int. J. Mod. Phys. A, 12 (1997), 1405?1429.
  • [Amo-BenB] L. Amorim, O. Ben-Bassat, Perversely categorified Lagrangian correspondences, e-Print: arXiv:1601.01536, 56 pages.
  • [Be-Fa] K. Behrend, B. Fantechi, Symmetric obstruction theories and Hilbert schemes of points on threefolds, Algebra Number Theory 2 (2008), 313?345.
  • [Bez-Ka] R. Bezrukavnikov, D. Kaledin, Fedosov quantization in algebraic context. Mosc. Math. J. 4 (2004), no. 3, 559-592.
  • [BBJ] C. Brav, V. Bussi, D. Joyce, A Darboux theorem for derived schemes with shifted symplectic structure, e-Print: arXiv:1305.6302 (submitted)
  • [BBDJS] C. Brav, V. Bussi, D. Dupont, D. Joyce, and B. Szendröi, Symmetries and stabilization for sheaves of vanishing cycles, Journal of Singularities 11 (2015), 85-151.
  • [Br-Dyc] C. Brav, T. Dyckerhoff, Relative Calabi-Yau structures, in preparation.
  • [Cal] D. Calaque, Lagrangian structures on mapping stacks and semi-classical TFT’s, Stacks and Categories in Geometry, Topology, and Algebra, Contemporary Mathematics 643 (2015).
  • [Cal2] D. Calaque, Deformed shifted cotangent stacks in derived Poisson geometry, preprint, 2016, 9 pages.
  • [CPTVV] D. Calaque, T. Pantev, B. Toën, M. Vaquié, G. Vezzosi, Shifted Poisson Structures and Deformation Quantization, e-Print: arXiv:1506.03699v3 (submitted).
  • [Coh] F. Cohen in F. Cohen, T. J. Lada, and J. P. May, The homology of iterated loop spaces, Springer-Verlag, Berlin, 1976.
  • [Co-Gwi] K. Costello, O. Gwilliam, Factorization algebras in quantum field theory Volume II, Book draft, December 2015, available at http://people.mpim-bonn.mpg.de/gwilliam/
  • [Dr1] V. Drinfeld, Quasi-Hopf algebras, Leningr. Math. J. 1 (1990), no. 6, 1419-1457.
  • [Dr2] V. Drinfeld, On quasi-triangular quasi-Hopf algebras and a group closely related with 𝖦𝖺𝗅(¯,)𝖦𝖺𝗅¯\mathsf{Gal}(\overline{\mathbb{Q}},\mathbb{Q}), Leningr. Math. J. 2 (1991), no. 4, 829-860.
  • [Gi-Ba] V. Ginzburg, V. Baranovsky, Gerstenhaber-Batalin-Vilkoviski structures on coisotropic intersections, to appear in Math. Res. Lett.
  • [Hen] B. Hennion, Tangent Lie algebra of derived Artin stacks ., e-Print: arXiv:1312.3167, to appear in Journal für die reine und angewandte Mathematik (Crelles Journal).
  • [Hir] P. S. Hirschhorn, Model categories and their localizations, Mathematical Surveys and Monographs (99), AMS, 2003.
  • [HS] A. Hirschowitz, C. Simpson. Descente pour les n-champs, e-Print: arXiv math/9807049.
  • [Hov] M. Hovey, Model categories, Mathematical Surveys and Monographs (63), AMS, 2007.
  • [Hoy] M. Hoyois, The fixed points of the circle action on Hochschild homology, e-Print: arXiv:1506.07123.
  • [Jo-Sa] D. Joyce and P. Safronov, Lagrangian neighborhood theorem for shifted symplectic derived schemes, e-Print: arXiv:1509.08081
  • [JS] D. Joyce and Y. Song, A theory of generalized Donaldson-Thomas invariants, Mem. Amer. Math. Soc. 217 (2012), no. 1020.
  • [Kap] A. Kapustin, Topological Field Theory, Higher Categories, and Their Applications, Proceedings of the International Congress of Mathematicians Hyderabad, India, 2010.
  • [Ka] C. Kassel, Quantum groups, Graduate Texts in Mathematics 155, 1995, Springer.
  • [Ko1] M. Kontsevich, Deformation quantization of Poisson manifolds, Lett. Math. Phys. 66 (2003), 157-216.
  • [KS] M. Kontsevich and Y. Soibelman, Stability structures, motivic Donaldson-Thomas invariants and cluster transformations, e-Print: arXiv:0811.2435, 2008.
  • [Ill] L. Illusie, Complexe cotangent et déformations II, Lecture Notes in Mathematics 283, Springer Verlag, Berlin, 1972.
  • [Lu1] J. Lurie, Higher Topos Theory, Annals of Mathematics Studies, 170, Princeton University Press, Princeton, NJ, 2009. xviii+925 pp.
  • [Lu2] J. Lurie, Formal Moduli Problems, Vol. 2, Proceedings of the International Congress of Mathematicians 2010 (R. Bhatia, A. Pal, G. Rangarajan, V. Srinivas, M. Vanninathan editors), World Scientific.
  • [Lu3] J. Lurie, DAG series, V to XIV, available at http://www.math.harvard.edu/similar-to\simlurie/
  • [Lu4] J. Lurie, Higher Algebra, September 2014, available at http://www.math.harvard.edu/similar-to\simlurie/.
  • [Lu5] J. Lurie, Spectral algebraic Geometry, draft February 2016, available at http://www.math.harvard.edu/similar-to\simlurie/.
  • [Lu6] J. Lurie, On the Classification of Topological Field Theories, Current Developments in Mathematics Volume 2008 (2009), 129-280.
  • [Mar] N. Markarian, Weyl n-algebras, e-Print: arXiv:1504.01931, 2015.
  • [Me] V. Melani, Poisson bivectors and Poisson brackets on affine derived stacks, to appear in Advances in Mathematics.
  • [Me-Sa] V. Melani, P. Safronov, Shifted Poisson stuctures and coisotropic intersections, in preparation.
  • [Ni-Sa] Niklaus, S. Sagave, Presentably symmetric monoidal infinity-categories are represented by symmetric monoidal model categories, e-Print: arXiv:1506.01475.
  • [PTVV] T. Pantev, B. Toën, M. Vaquié, G. Vezzosi, Shifted symplectic structures, Publ. Math. IHES, June 2013, Volume 117, Issue 1, pp 271-328.
  • [Saf] P. Safronov, Poisson reduction as a coisotropic intersection, e-Print: arXiv:1509.08081.
  • [Sch] M. Schlessinger, Functors of Artin rings, Transactions of the American Mathematical Society 130: 208-222 (1968).
  • [STZ] V. Shende, D. Treumann, E. Zaslow, Legendrian knots and constructible sheaves, e-Print: arXiv:1402.0490.
  • [Spa] T. Spaide, Shifted symplectic structures on spaces of framed maps, thesis, University of Pennsylvania, 2015.
  • [To1] B. Toën, Higher and derived stacks: a global overview. Algebraic geometry - Seattle 2005. Part 1, 435-487, Proc. Sympos. Pure Math., 80, Part 1, Amer. Math. Soc., Providence, RI, 2009.
  • [To2] B. Toën, Derived Algebraic Geometry, EMS Surv. Math. Sci. 1 (2014), no. 2, 153-245.
  • [To-Va] B. Toën, M. Vaquié, Moduli of objects in dg-categories, Ann. Sci. de l’ENS Volume 40 (2007) Issue 3, Pages 387-444.
  • [HAG-II] B. Toën, G. Vezzosi, Homotopical algebraic geometry II: Geometric stacks and applications, Mem. Amer. Math. Soc. 193 (2008), no. 902, x+224 pp.
  • [To-Ve-1] B. Toën, G. Vezzosi, Caractères de Chern, traces équivariantes et géométrie algébrique dérivée, Selecta Mathematica 21 (2015), no. 2, 449–554.
  • [To-Ve-2] B. Toën, G. Vezzosi, Algèbres simpliciales S1superscript𝑆1S^{1}-équivariantes, théorie de de Rham et théorèmes HKR multiplicatifs, Compositio Math., 147 / Issue 06, n. 11, pp. 1979-2000 (2011).
  • [Ty] A. N. Tyurin, Fano versus Calabi-Yau, The Fano Conference, Univ. Torino, Turin, 2004, pp. 701-734.
  • [Vez1] G. Vezzosi, Derived critical loci I - Basics, e-Print: arXiv:1109.5213.
  • [Vez2] G. Vezzosi, Quadratic forms and Clifford algebras on derived stacks, e-Print: arXiv:1309.1879 (submitted).
  • [Vez3] G. Vezzosi, A model structure on relative dg-Lie algebroids, Proceedings of CATS4: Stacks and Categories on Geometry, Topology, and Algebra, Contemporary Mathematics 643, 111–118.
  • [We] A. Weinstein, Coisotropic calculus and Poisson groupoids, J. Math. Soc. Japan 40 (1988), 705-727.
  • [Ye] A. Yekutieli, Twisted deformation quantization of algebraic varieties. Adv. Math. 268 (2015), 241-305.

Tony Pantev, University of Pennsylvania, tpantev@math.upenn.edu

Gabriele Vezzosi, Università di Firenze, gabriele.vezzosi@unifi.it