Line bundles on complex tori and a conjecture of Kodaira (Q2483780)

In 1963 Kodaira proved that every smooth compact Kähler surface is almost algebraic in the sense that it can be realized as a deformation of a projective surface. Until recently it was an open problem whether it is true in general that a compact Kähler manifold is almost algebraic. An affirmative answer would have implied that every compact Kähler manifold has the homotopy type of a projective algebraic manifold and is projective if it is rigid. But in 2004 \textit{C. Voisin} [Invent. Math. 157, No. 2, 329--343 (2004; Zbl 1065.32010)] constructed for every dimension greater than three compact Kähler manifolds with homotopy type different from the homotopy type of a projective manifold. Her examples are built from compact complex tori by blowing up processes. In the paper under review the authors are equally interested in giving an affirmative answer to the above question for a subclass of compact Kähler manifolds and in finding new counterexamples. They discuss different strategies for \(\mathbb P(V)\)-bundles on complex tori. These considerations are based on the fact that the structure of \(\mathbb P(V)\)-bundles or \(\mathbb P_r\)-bundles over a compact complex manifold survives under deformation (Theorem 8). The authors consider the following situation: Let \(A\) be a three-dimensional compact complex torus and \(L_1, L_2, L_3\) holomorphic line bundles on \(A\) representing three linear independent elements in the Néron-Severi group NS\((A)\). The manifold \(Y=\mathbb P({\mathcal O}_A\oplus L_1)\times_A\mathbb P({\mathcal O}_A\oplus L_2)\times_A\mathbb P({\mathcal O}_A\oplus L_3)\) is a holomorphic \( \mathbb P^3_1\)-bundle over \(A\) with a natural holomorphic section \(Z\) given by the direct summand \({\mathcal O}_A\) in every factor. The main result of the paper (Theorem 4) asserts that \(Y\) is algebraically approximable by projective Albanese bundles \(Y_n\rightarrow A_n\) with \(Y_n=\mathbb P({\mathcal O}_{A_n}\oplus L_1)\times_{A_n}\mathbb P({\mathcal O}_{A_n}\oplus L_2)\times_{A_n}\mathbb P({\mathcal O}_{A_n}\oplus L_3)\) and \(\lim_{n\rightarrow\infty}A_n=A\) in the sense of deformation theory. The proof uses an explicit description of NS\((A)\) in terms of skew-symmetric integer \(6\times 6\) matrices and calculations with support of the computer algebra program Macauley 2 [see \textit{D. Eisenbud} et al., Algorithms and Computation in Mathematics. 8. (Berlin: Springer) (2002; Zbl 0973.00017)]. Blowing up in the bundle \(Y\) every fiber \(F\) in the point \(F\cap Z\) gives a compact Kähler manifold \(X\), a holomorphic fiber bundle over \(A\) with projective rational fiber. Under the additional assumption that not all of the line bundles \(L_1, L_2, L_3\) remain holomorphic under small deformations of \(A\), the manifold \(X\) is rigid (Proposition 3) and could a priori be a counterexample. But the assumption on the \(L_i\) forces \(A_n=A\) in Theorem 4, hence \(X\) is already projective. The authors explain their ideas how modifications of their construction and more general settings could eventually lead to new counter-examples.