\(\mathcal{A}_\infty\)-functors and homotopy theory of dg-categories (Q1676076)

Given two nonunitary \(A_{\infty}\)-categories \(A\) and \(B\), there exists a nonunitary \(A_{\infty}\)-category \(\mathscr{A}_{\infty}(A,B)\) whose objects are nonunitary \(A_{\infty}\)-functors from \(A\) to \(B\) and the graded space of morphisms from an \(A_{\infty}\)-functor \(f\) to another \(A_{\infty}\)-functor \(g\) is given by the graded functors from the bar construction of \(A\) to (the pullback by \((f,g)\) of) \(B\). The \(A_{\infty}\)-category structure is induced by the noncounitary dg cocategory structure of the cobar construction of \(A\) and the \(A_{\infty}\)-category structure of \(B\) (see [\textit{K. Lefèvre-Hasegawa}, Sur les \(A_{\infty}\)-catégories. Paris: Université Denis Diderot (PhD Thesis) (2003)], 8.1.1), in the same fashion as the convolution algebra. In particular, if \(A\) and \(B\) are nonunitary dg categories, then \(\mathscr{A}_{\infty}(A,B)\) is a nonunitary dg category. The previous results can also be extended to the case where the \(A_{\infty}\)-categories are strictly unitary: \(\mathscr{A}_{\infty}(A,B)\) is now formed by the strictly unitary \(A_{\infty}\)-functors provided with the restricted \(A_{\infty}\)-structure (see [loc. cit.], 8.1.3). The first main result of the article under review states that there exists in natural isomorphism in the homotopy category of dg categories (for the model structure given in [\textit{G. Tabuada}, C. R., Math., Acad. Sci. Paris 340, No. 1, 15--19 (2005; Zbl 1060.18010)]) between \(\mathscr{A}_{\infty}(A,B)\) and the dg category of strictly unitary \(A_{\infty}\)-bimodules over \(A\) and \(B\), for any two (unitary) dg categories \(A\) and \(B\) (see Theorem 2.6). Even though the author does not state it, the proof presented in Sections 2.3--2.5 only holds for augmented dg categories. Indeed, the author needs to use there the associated enveloping dg category of a dg category, which makes sense in principle only in the augmented case. To wit, (the first statements of) Lemmas A.20, A.21 and A.22 do not even make sense, neither do Def. A.23 or A.24. Indeed, in Def. A.14, A.16, and A.18, the author claims that the corresponding reductions give dg categories, \(A_{\infty}\)-categories and dg cocategories, respectively. However, this is not true in general: if a dg category is given by a plain algebra \(A\) (i.e., a dg category with one object, with morphisms in zero degree), then the product of \(A\) does not necessarily induce a well-defined product on \(A/k\) (e.g. if \(A\) is the Weyl algebra \(k\langle x,y \rangle/(xy - yx - 1)\)). However, all the statements are right if the dg (co)categories and \(A_{\infty}\)-categories considered are augmented (resp., coaugmented), and all the morphisms spaces between the (co)augmented objects are in the (co)augmented sense. The same applies to several of the following results in Section B.3, even though the author seems to state the opposite in the first paragraph of that section (p. 997). Reviewer's remark:The mentioned Theorem 2.6 could be proved in full generality (i.e. without the augmentation assumption) by modifying the proof as follows. There is a possibility of defining the enveloping dg category and the bar construction of a (not necessarily augmented) dg category, essentially behaving as in the augmented case. Indeed, \textit{L. Positselski} has introduced in [Mem. Am. Math. Soc. 996, i--iii, 133 p. (2011; Zbl 1275.18002)] the notion of curved dg algebra and curved dg coalgebra, and he showed that the bar (resp., cobar) construction of usual augmented dg algebras (resp., cocomplete coaugmented dg coalgebras) can be extended to the curved case. The bar construction of a unitary dg algebra is a curved coaugmented dg coalgebra, whose cobar construction is in turn again a unitary dg algebra. This is called the enveloping algebra of the unitary dg algebra one started with (and they are both quasi-isomorphic, and fulfill various of the expected properties). In the case of this article, one would need the corresponding versions for dg categories, which are straightforward. Concerning the other results in the article, the author then introduces the \((\infty,2)\)-category \(\text{dgCat}_{(\infty,2)}\) whose objects are all the dg categories, and considers the associated \((\infty,1)\)-category \(\text{dgCat}_{(\infty,1)}\), proving that the latter gives a model for the simplicial localization at the model structure of Tabuada (see Theorem 4.2). Finally, by noting that the Hochschild cohomology of an augmented \(A_{\infty}\)-category \(A\) is just the cohomology of the space of endomorphisms in \(\mathscr{A}_{\infty}(A,A)\) of the identity functor -- which follows immediately from the definition of \(\mathscr{A}_{\infty}(A,A)\)--, the author obtains a nice application of his Theorem 4.2: the Hochschild cohomology of \(A\) is the homotopy of the mapping space of endomorphisms at the identity functor in the \((\infty,2)\)-category of \(A_{\infty}\)-categories \(\mathscr{A}_{\infty}\text{Cat}_{(\infty,2)}\) (see Theorem 5.3).