Calabi-Yau deformations and negative cyclic homology (Q1731728): Difference between revisions

Let \(k\) be a field, \(\operatorname{char}k=0.\) In very short terms, the authors study the deformation theory of Calabi-Yau \(k\)-algebras in the sense of Ginzburg. An \(A\)-module \(M\) is called perfect, if it is has a finite resolution by finitely generated projective \(A\)-modules. For an \(R\)-algebra \(B,\) let \(B^\circ\) be the opposite algebra, and let \(B^e=B\otimes_R B^\circ.\) A \(k\)-algebra \(A\) is \(d\)-Calabi-Yau (CY) if is perfect as \(A^e\)-module, and if there is an isomorphism in \(D(A^e),\) \(\eta:\operatorname{RHom}_{A^e}(A,A^e)\rightarrow\sum^{-d}A.\) Fix a \(d\)-CY algebra \(A\) and an isomorphism \(\eta\) as above. The authors use the terminology that the definition of a \(d\) CY algebra can be relativized so that there exists an associated deformation theory, and the first main result of the article is to construct a DG-Lie algebra controlling the deformation theory: Let \(\operatorname{Nilp}\) be the category of commutative, finite dimensional, local \(k\)-algebras \((R,m)\) with \(R/m=k.\) For \(R\in\operatorname{Nilp},\) \(\operatorname{Def}_{A,\eta}(R)\) is the \textit{category} of \(R\)-algebras \(B\) which are \(d\)-Calabi-Yau with respect to \(R\) and such that \(B\otimes_Rk\cong A\) respecting \(\eta.\) With this definition of deformations as a category (not only a set), the authors can view \(\operatorname{Def}_{A,\eta}\) as a pseudo-functor from \(\operatorname{Nilp}\) to the category of groupoids \(\operatorname{Gp}.\) Let \(\mathfrak h\) be a nilpotent DG-Lie algebra. Let \(\operatorname{MC}(\mathfrak h)\) be the groupoid of solutions to the Maurer-Cartan equation in \(\mathfrak h.\) For \((R,m)\in\operatorname{Nilp}\) and a DG-Lie algebra \(\mathfrak g^\bullet\) over \(k,\) the vector space \(\mathfrak g^\bullet\otimes_k m\) is a nilpotent DG-Lie algebra. Then \(\mathcal{MC}(R)=\operatorname{MC}(\mathfrak g^\bullet\otimes_k m)\) equipped with a specified groupoid structure gives a pseudo-functor \(\mathcal{MC}(\mathfrak g):\operatorname{Nilp}\rightarrow\operatorname{Gd}:(R,m)\mapsto\operatorname{MC}(\mathfrak g\otimes_k m)\) which is the announced deformation functor. The main objective of this paper is to introduce a DG-Lie algebra \(\mathcal D^\bullet(A,\eta)\) which controls the deformation theory of the \(d\)-CA algebra \((A,\eta).\) Verbatim: Theorem 1.1. There is a morphism of pseudo-functors \(\pi:\mathcal{MC}(\mathcal D^\bullet(A,\eta))\rightarrow\operatorname{Def}_{A,\eta}\) which when evaluated on an arbitrary \(R\in\operatorname{Nilp}\) is essentially surjective on objects and surjective on morphisms. In particular, for \(R\in\operatorname{Nilp}\) this gives a bijection \(\operatorname{MC}(\mathcal D^\bullet(A)\otimes_k m)/\cong\overset\sim\rightarrow\operatorname{Def}_{A,\eta}(R)/\cong\) so that indeed the deformation theory of \(A\) is controlled by the DG-Lie algebra \(\mathcal D^\bullet(A,\eta).\) The authors construct \(\mathcal D^\bullet(A,\eta)\) as a twisted semi-direct product of the Hochschild cochain complex with the negative cyclic chain complex of \(A,\) and this is done in a way similar to using finite-dimensional \(A_\infty\)-algebras with a non-degenerate inner product, except that the present algebra is not finite dimensional, nor does it have an inner product. The paper includes a section with the theory of Hochschild and cyclic complexes in local and relative versions, which makes the definitions very explicit and the paper self-content. The construction of \(\mathcal D^\bullet(A,\eta)\) induce a morphism \(\phi:\mathcal D^\bullet(A,\eta)\rightarrow\overline{\mathfrak C}^\bullet(A),\) where \(\overline{\mathfrak C}^\bullet(A)\) is the shifted Hochschild cochain complex of \(A.\) Now, \(\overline{\mathfrak C}^\bullet(A)\) controls the deformation theory of the algebra \(A,\) so that \(\phi\) corresponds to forgetting \(\eta.\) In the second main result, the authors explicitly construct a quasi-isomorphism of complexes \(\mathcal D^\bullet(A,\eta)\longrightarrow\sum^{-d+1}\operatorname{CC}_\bullet^-(A)\) to the shifted negative cyclic complex \(\operatorname{CC}_\bullet^-(A).\) This proves that the tangent space to the deformation space of a \(d\)-Calabi-Yau algebra is \(\operatorname{HC}^-_{d-2}(A),\) and the obstructions against deforming a \(d\)-Calabi-Yau algebra are in \(\operatorname{ker}(\operatorname{HC}^-_{d-3}(A)\rightarrow\operatorname{HC}^{\text{per}}_{d-3}(A)).\) In particular, if the morphism above is injective, then the deformation theory of \(A\) as Calabi-Yau algebra is unobstructed. This happens in the case where \(d\leq 3.\) The authors prove that the Lie bracket on \(\operatorname{HC}_\bullet(A)\) induced by the quasi-isomorphism of complexes above is the \textit{string topology} introduced by Menichi. Specializing to the case where \(A\) is commutative, letting \(T^{\text{Poly},\bullet}(A)\) be the Lie algebra of poly-vector fields on \(A,\) \(\eta:\operatorname{RHom}_{A^e}(A,A^e)\rightarrow\sum^{-d}A\) can be interpreted as a volume form. When \(\operatorname{div}\) is the divergence operator on \(T^{\text{Poly},\bullet}(A)\) associated to \(\eta\) the authors show by using the Willwacker's \textit{formality for cyclic chains} that there is an isomorphism \((T^{\text{Poly},\bullet}[[u]],-u\operatorname{div})\overset\cong\longrightarrow\mathcal D^\bullet(A,\eta)\;(|u|=2)\) in the category of DG-Lie algebras, commuting with a globalized version of Kontsevich's formality quasi-isomorphism. Thus they give a conceptual explanation of Dolgushev's result that the Kontsevich's \(\ast\)-product associated to a divergence free Poisson bracket is Calabi-Yau. In addition to presenting interesting results, the article contains very readable preliminaries on Hochschild and cyclic complexes, and Calabi-Yau algebras. In particular, the following definition is used: A Calabi-Yau algebra of dimension \(d\) over \(R\) is a couple \((B,\eta)\) where \(B\) is a homologically smooth \(R\)-algebra and \(\eta\) is a non-degenerate element of \(\operatorname{HC}^-_d(B).\) This is essential for the development of the results given in this article.