Derived algebraic geometry, determinants of perfect complexes, and applications to obstruction theories for maps and complexes (Q2344038)

The article under review applies techniques from derived algebraic geometry to the formalism of perfect obstruction theories [\textit{K. Behrend} and \textit{B. Fantechi}, Invent. Math. 128, No. 1, 45--88 (1997; Zbl 0909.14006)]. Recall that a perfect obstruction theory a morphism \(E\to{\mathbf L}_{\mathcal X}\) in the derived category which is an isomorphism in degree zero cohomology, surjective in degree minus one cohomology and where \(E\) is of perfect amplitude in degrees \([-1,0]\). It is used to define virtual fundamental classes in the Chow group of the stack, and in the case of the stack of pointed stable maps one obtains all Gromov-Witten invariants by integrating against this class. In section 1 it is explained how a derived enhancement \(\mathcal X^{\text{der}}\) of a Deligne-Mumford stack \(\mathcal X\) induces via the canonical closed immersion \(j:\text{t}_0({\mathcal X}^{\text{der}})\hookrightarrow{\mathcal X}^{\text{der}}\) a perfect obstruction \(j^*({\mathbf L}_{\mathcal X^{\text{der}}})\to{\mathbf L}_{\mathcal X}\), provided the stack is quasi-smooth (i.e.\ the cotangent complex is of perfect amplitude contained in \([-1,0]\)). The functoriality properties in derived algebraic geometry then improve the limited and ad hoc functoriality properties for perfect obstruction theories. In section~2 derived enhancements of well-known moduli spaces are constructed: the Picard stack resp.\ the stack of (\(n\)-pointed) (pre)stable maps of genus \(g\), and open substacks of these fixing a class \(c_1\in\text{H}^2(X,{\mathbb Z})\) resp.\ \(\beta\in\text{H}_2(X,{\mathbb Z})\). It is explained how the derived enhancement \({\mathbb R}\overline{M}_g(X;\beta)\) induces the \textit{standard} perfect obstruction theory on the underived moduli stack \(\overline{M}_g(X;\beta)\) of the moduli stack of stable maps of type \((g,\beta)\) (and likewise for a pointed version) inducing Gromov-Witten invariants. This is not the only perfect obstruction theory in the literature: it can be shown that for a \(K3\) surface this standard perfect obstruction theory gives trivial Gromov-Witten invariants, hence the need for the \textit{reduced} perfect obstruction theory. To construct a derived enhancement inducing this reduced perfect obstruction theory a version of the determinant map in derived algebraic geometry is constructed in section 3. This is a morphism \(\text{det}_{\text{perf}}:{\mathbb R}\text{Perf}\to\text{Pic}\) generalising \(\text{Vect}\to\text{Pic}\). Using this determinant map a morphism \(\text{A}_X:{\mathbb R}\overline{M}_g(X)\to{\mathbb R}\text{Perf}(X)\) is constructed using the universal family on \({\mathbb R}\overline{M}_g(X)\), and it is explained how its tangent morphism is related to the Atiyah class of its associated perfect complexes. Section 4 is devoted to the special case of \(\overline{M}_g(S;\beta)\) where \(S\) is a \(K3\) surface, to obtain the reduced perfect obstruction theory which gives non-trivial Gromov-Witten invariants. First the properties of this obstruction theory are recalled. To construct the correct enhancement the derived Picard stack for a \(K3\) surface is shown to exhibit a special structure, which allows to use the morphism \(\text{A}_S\) from the previous section to define a new derived enhancement \({\mathbb R}\overline{M}_g^{\text{red}}(S;\beta)\). The main technical part of the paper is in the proof that the induced perfect obstruction theory is indeed the reduced obstruction theory. In section 5 the moduli stack of perfect complexes and various substacks are studied. In the case of a \(K3\) surface it is shown using the perfect determinant map that the reduced enhancement from section 4 gives rise to a smooth \textit{underived} substack of simple perfect complexes, whose coarse moduli space was shown to be a smooth algebraic space by \textit{M. Inaba} [Adv. Math. 227, No. 4, 1399--1412 (2011; Zbl 1220.14010)]. In the case of a smooth projective Calabi-Yau 3-fold the standard enhancement from section~3 gives rise to a (non-proper) quasi-smooth derived stack of simple perfect complexes with fixed determinant. Finally a comparison morphism between this stack and an appropriate open substack of \({\mathbb R}\overline{M}_{g,n}(X)\) from section~2 is constructed, which allows for an open version of the comparison between Donaldson-Thomas and Gromov-Witten invariants.