Deformation of a generically finite map to a hypersurface embedding (Q1739226)

The paper concerns the question whether a smooth hypersurface \(W_0\) in a smooth projective variety \(Z_0\) can be deformed in a family of such hypersurface embeddings. The author obtains a partial answer to this, through an iterated blowup of special type: \begin{align*} \sigma (z)&=w_0t_0,\\ Q_1(w_0,z)&=w_1t_1,\\ & \ldots \\ Q_k(w_0,\ldots ,w_{k-1},z)&=w_kt_k,\\ Q_{k+1}(w_0,\ldots ,w_k,z)&=0. \end{align*} When we fix a value \(t_0,\ldots ,t_k\not= 0\), by eliminating the \(t_k\)'s, we obtain a hypersurface embedding \[ Q_{k+1}(\sigma (z)/t_0,Q_1(\sigma (z)/t_0,z)/t_1,\ldots ,Q_k(\sigma (z)/t_0, Q_1(\sigma (z)/t_0,z)/t_1\ldots ))=0. \] In the setting of the paper, the polynomial \(Q_j\)'s is to define a hypersurface on a line bundle over the variety defined by the first \(j\)'s blowups. The main result of the paper (Theorem 2.4) includes 2 parts, which are partial converses to each other. Part 1, which is more complicated to prove, is the following: Assume that one has a \(1\)-parameter family of maps \(\phi _t: \mathcal{W}_t\subset \mathcal{Z}_t\) between smooth \(n\) and \(n+1\) dimensional varieties. Assume that for \(t\not= 0\) then \(\phi _t\) is an embedding, while \(\phi _0\) is a generically finite morphism of degree \(m\) to its image \(\Sigma _0=mX\) where \(X\) is reduced. Importantly, assume also that \(\mathcal{W}_0\) has ample canonical divisor. Then \(X_0\) is smooth and the map \(\mathcal{W}_0\rightarrow X_0\) is an iterated blowup of the above type. The assumption that \(K_{\mathcal{W}_0}\) is ample is used in Lemma 3.1, which says that under the mentioned assumptions, provided the image of \(\mathcal{W}\) by the map \(\Phi\) is normal and Gorenstein, then \(\mathcal{W}_0\) is isomorphic to \(X\). Besides this, a main technique is to reduce to dimension 1, such as in Lemma 3.5. The iterated blowup of the mentioned type appears as resolution of singularities of a monomial polynomial in a variable y and with coefficients depending on another variable \(t\), and an explicit procedure is given in Lemma 3.8. As applications, the authors consider hypersurfaces of projective spaces and Abelian varieties, with ample canonical divisor. In these cases, the partial converses in Theorem 2.4 are indeed perfect converses. However, it leaves open the question how to make Theorem 2.4 to be more complete. For example, if \(\mathcal{W}_0\) does not have an ample canonical divisor, can we still say something interesting?