Semiregularity and obstructions of complete intersections (Q1941144)

Let \(X\) be a smooth algebraic variety over an algebraically closed field \(k\) of characteristic \(0\), \(Z\subset X\) a locally complete intersection closed subvariety of codimension \(p\). There is a \textit{semiregularity} map \(\pi:H^1(Z,N_{Z|X})\rightarrow H^{p+1}(X,\Omega^{p-1}_X)\), described by local cohomology by the existence of the \textit{canonical} cycle class \(\{Z\}^\prime\in H^1(Z,\mathcal H^p_Z(\Omega^p_X))\); the contraction with this class gives a morphism of sheaves \(N_{Z|X}\overset{\lrcorner\{Z\}^\prime}\longrightarrow\mathcal H^p_Z(\Omega_X^{p-1}),\) and passing to cohomology, the result is a map \(H^1(Z,N_{Z|X})\overset{\lrcorner\{Z\}^\prime}\longrightarrow H^1(Z,\mathcal H^p_Z(\Omega^{p-1}_X))=H^{p+1}_Z(X,\Omega^{p-1}_X)\). The last equality follows from the spectral sequence of local cohomology. Using Hodge theory and de Rham cohomology, Bloch proved that if \(X\) is projective, certain obstructions to embedded deformations of \(Z\) in \(X\) are annihilated by the semiregularity map. The annihilated obstructions contains the curvilinear ones, and so, if the semiregular map is injective, the Hilbert scheme of subschemes of \(X\) is smooth at \(Z\). Bloch did not prove that all obstructions are annihilated, and the authors intend to prove this mainly for two reasons: the first is to test the power of \textit{derived deformation theory} in a classical problem, the second is related to the theory of reduced Gromow-Witten invariants. In the last case one needs the virtual fundamental class, defined by obstruction theory. If the obstruction theory is not chosen correctly, the virtual fundamental class is zero, and then the standard GW theory is trivial. This means that a reduced obstruction theory must be used, this is obtained by considering the kernel of a suitable map annihilating obstructions. Derived deformation theory is summarized as follows (\(\text{char} k=0\)): Every deformation problem is the classical truncation of an extended deformation problem controlled by a differential graded Lie algebra (DGLA) via the Maurer-Cartan equation and gauge equivalence. The DGLA is defined up to quasi-isomorphism and its first cohomology group is equal to the Zariski tangent space of the local moduli space. A morphism of deformation theories is a morphism in the derived category of DGLA's. In fact, the induced morphisms in cohomology gives the tangent- and obstruction spaces. A morphism from a defomation theory into an unobstructed one provides an obstruction map annihilating every obstruction. The authors prove that the semiregularity map annihilates every obstruction when \(Z\) is closed of codimension \(p\) in \(X\) and there exists a Zariski open subset \(U\subset X\) and a vector bundle \(E\rightarrow U\supset Z\) of rank \(p\) such that \(Z\) is the zero locus of a section \(f\in\Gamma(U,E)\). Now, we state the main result of the article word by word, recall that hypercohomology is ``cohomology of cohomology'' (in the derived category): Main theorem. Let \(X\) be a smooth algebraic variety over \(k=\overline{k}\), \(\text{char} k=0\), and let \(Z\subset X\) be a codimension \(p\) closed subvariety as above. Then the composition of the semiregularity map composed with the truncation \[ H^1(Z,N_{Z|X})\overset\pi\rightarrow H^{p+1}(X,\Omega^{p-1}_X)\overset{\tau}\rightarrow\mathbb H^{2p}(X,\Omega^0_X\overset {d}\rightarrow\cdots\overset {d}\rightarrow\Omega^{p-1}_X) \] annihilates every obstruction to infinitesimal embedded deformations of \(Z\) in \(X\). The intention of the authors is to give a purely algebraic proof for smooth manifolds, and extend to the case of DG schemes, and to replace the embedding \(Z\subset X\) with a quasi quasi-isomorphic embedding of smooth DG-schemes. It is for the last idea that the rank \(p\)-bundle above is needed. The authors review the essential things needed for the proof, such as \textit{semifree modules over DG rings}, and the \textit{homotopy fiber} of a morphism of DGLA's \(\chi:L\rightarrow M\), the differential graded Lie algebra \[ TW(\chi):=\{(l,m(t,dt))\in L\times M[t,dt]|m(0,0)=0,\:m(1,0)=\chi(l)\}. \] Two DGLA's controlling the functor \(\text{Hilb}_{Z|X}:\mathbf{Art}\rightarrow \mathbf{Set}\) of infinitesimal embedded deformations of \(Z\) in \(X\) is given. The first is simpler with a clear geometric interpretation, the second, quasi-isomorphic to the first, is the one used in the final proof of the main theorem. The proof also uses \(L_\infty\)-algebras and calculus on the de Rham complex. The article is a nice survey of deformation theory, and illustrates common methods for treating the subject. Also, these methods are used in its most advanced way to prove a nice result for this actual subject.