Morita theory for non-commutative Noetherian schemes (Q2118921)

This paper is concerned with equivalences between categories of quasi-coherent sheaves over so-called \textit{non-commutative noetherian schemes}. Examples of such schemes have appeared frequently as non-commutative generalizations of projective curves under many names: as \textit{weighted projective lines} in the sense of Geigle and Lenzing, \textit{categorical resolutions} of singular commutative curves in work of \textit{I. Burban} et al. [Int. Math. Res. Not. 2017, No. 3, 895--920 (2017; Zbl 1405.14070)], and \textit{stacky chains} or \textit{stacky cycles of projective lines} studied by \textit{Y. Lekili} and \textit{A. Polishchuk} [J. Topol. 11, No. 3, 615--644 (2018; Zbl 1403.53075)] in connection with homological mirror symmetry. There is a strong link between non-commutative curves and finite-dimensional derived-tame algebras via tilting theory. In particular, the first class of examples leads naturally to \textit{canonical algebras} whereas the last class is related to \textit{gentle algebras}. Prototypes for the first two main results of the paper are provided by classical Morita theory. Namely, for any rings \(A\) and \(B\) such that there is an equivalence of categories of left modules \(\mathrm{Mod}(A) \overset{\sim}{\longrightarrow} \mathrm{Mod}(B)\) the following statements hold: \begin{itemize} \item[1.] The equivalence induces a ring isomorphism \(Z(A) \overset{\sim}{\longrightarrow} Z(B)\) of the centers of \(A\) and \(B\). \item[2.] There is a ring isomorphism \(B \cong \mathrm{End}_A(P)\) for a right progenerator \(P\) over the ring \(A\). \end{itemize} A \textit{non-commutative noetherian scheme} \(\mathbb{X}\) is defined as a ringed space \((X,\mathcal{A})\), where \(X\) is a separated noetherian scheme and \(\mathcal{A}\) is a sheaf of certain algebras over the structure sheaf \(\mathcal{O}_X\) of \(X\). There is a natural notion of quasi-coherent sheaves over \(\mathbb{X}\). These form an abelian category \(\mathsf{QCoh}(\mathbb{X})\) which can be thought of as being `glued' from finitely many module categories of certain \(\mathcal{O}_X\)-algebras. Given non-commutative noetherian schemes \(\mathbb{X} = (X,\mathcal{A})\) and \(\mathbb{Y} = (Y,\mathcal{B})\) satisfying certain negligible technical conditions and assuming that these are \textit{Morita equivalent}, that is, there is an equivalence of categories \(\Phi\colon \mathsf{QCoh}(\mathbb{X}) \overset{\sim}{\longrightarrow} \mathsf{QCoh}(\mathbb{Y})\), the first main results of the paper (Theorems~4.4 and~4.6) imply the following statements: \begin{itemize} \item[1.] The equivalence \(\Phi\) induces an isomorphism of the underlying schemes \(\varphi\colon X \overset{\sim}{\longrightarrow} Y\). \item[2.] There is an isomorphism of \(\mathcal{O}_Y\)-algebras \(\mathcal{B} \cong \varphi^*(\mathrm{End}_{\mathcal{A}}(\mathcal{P}))\) for a local right progenerator \(\mathcal{P}\) of \(\mathbb{X}\). \end{itemize} These results yield a non-commutative extension of Gabriel's reconstruction theorem which states that the noetherian schemes \(X\) and \(Y\) are Morita equivalent if and only if they are isomorphic. In fact, the proof of the first statement follows Gabriel's approach and introduces a `cover' \(\mathrm{Sp}(\mathbb{X}) \twoheadrightarrow X\), where the domain is given by the indecomposable injective objects of the category \(\mathsf{QCoh}(\mathbb{X})\). The equivalence \(\Phi\) gives rise to a bijection \(\mathrm{Sp}(\mathbb{X}) \overset{\sim}{\longrightarrow} \mathrm{Sp}(\mathbb{Y})\), which is then shown to induce an isomorphism of schemes \(\varphi\). In view of the results above, it seems natural to fix the isomorphism class of the underlying scheme \(X\) and to ask for additional, tangible data which would determine the Morita equivalence class of \(\mathbb{X}\) completely. This is the theme for two types of non-commutative noetherian schemes \(\mathbb{X}\) in the second half of the paper. In the first setup, \(\mathcal{A}\) is assumed to be a \textit{sheaf of Azumaya algebras}. The Morita equivalence class of \(\mathbb{X}\) turns out to be essentially determined by the isomorphism class of its underlying scheme \(X\) together with the equivalence class of the sheaf \(\mathcal{A}\) in the Brauer group of the scheme \(X\). A rigorous form of this statement was conjectured by \textit{A. Căldăraru} [J. Reine Angew. Math. 544, 161--179 (2002; Zbl 0995.14012)], and proven first in the above generality by \textit{B. Antieau} [J. Reine Angew. Math. 712, 175--188 (2016; Zbl 1375.14072)]. The authors give a new, short and self-contained proof of Căldăraru's conjecture (Theorem 5.3) based mainly on their Morita-theoretic results. In the second setup, the schemes \(\mathbb{X}\) and \(\mathbb{Y}\) are specialized to be \textit{reduced non-commutative curves} which include all of the aforementioned examples. A rough version of the last main result (Theorem 7.8) states that the curves \(\mathbb{X}\) and \(\mathbb{Y}\) are Morita equivalent if and only if the following conditions are satisfied: \begin{itemize} \item[--] There is an isomorphism \(\varphi\colon X \overset{\sim}{\longrightarrow} Y\) of the underlying schemes. \item[--] The isomorphism \(\varphi\) restricts to a bijection on sets of certain \textit{non-regular points} \(\mathfrak{S}_\mathbb{X} \overset{\sim}{\longrightarrow} \mathfrak{S}_{\mathbb{Y}}\). \item[--] For each non-regular point \(x \in \mathfrak{S}_\mathbb{X}\) the stalks \(\widehat{A}_x\) and \(\widehat{B}_{\varphi(x)}\) of the sheaves \(\mathcal{A}\) respectively \(\mathcal{B}\) as well as certain \textit{rings of rational functions} \(\Lambda_\mathbb{X}\) and \(\Lambda_\mathbb{Y}\) are Morita equivalent, where the equivalence functors have to be compatible with the isomorphism \(\varphi\) in a certain sense. \end{itemize} Finally, the authors demonstrate that the compatibility constraints may not be removed (Example 7.12). In particular, there are non-commutative curves which are `locally' but not `globally' Morita equivalent.