Conifold transitions and Mori theory (Q812523)

Conifold transitions of sixdimensional symplectic real manifolds \(X\) were defined and intensively studied in \textit{I. Smith, R. P. Thomas} and \textit{S.-T. Yau} [J. Differ. Geom. 62, No. 2, 209--242 (2002; Zbl 1071.53541)] in connection with mirror symmetry and non-Kählerian symplectic sixfolds with \(c_1=0\). A symplectic conifold transition on \(X\) is a surgery which replaces a Lagrangian three-sphere with a Lagrangian two-sphere, cutting out \(S^3\times D^3\) and replacing it by \(D^4\times S^2\). As was shown in the above cited paper the simultaneous surgery of disjoint Lagrangian three-spheres \(L_i\) can be performed compatibly with the symplectic structure if the \(L_i\) satisfy a homology relation \(\sum_i \lambda_i[L_i]=0\in H_3(X;\mathbb Z)\) with \(\prod_i\lambda_i\not=0\). The main theorem of the paper under review says that there are symplectic conifold transitions of projective threefolds which are not deformation equivalent to Kähler threefolds. This implies in particular that there exist projective algebraic varieties containing a Lagrangian sphere which is not the vanishing cycle of any Kähler degeneration. For the proof of the theorem the authors consider a projective Enriques surface \(E\) defined over \(\mathbb R\) and construct a nullhomologous Lagrangian three-sphere \(L\) in the product \(E\times\mathbb P^1\). The conifold transition along \(L\) gives a symplectic manifold \(Z\) which is projective or non-Kählerian. If \(Z\) is projective, then the theorem of \textit{S. Mori} [Ann. Math. (2) 116, 133--176 (1982; Zbl 0557.14021)] would apply to \(Z\): either \(Z\) is a Fano manifold or \(K_Z\) is nef or \(Z\) admits an extremal ray which can be contracted by a fibring or divisorial contraction. But the authors show by cohomological calculations that Mori's theorem does not apply to \(Z\). Editorial remark: According to the erratum [the authors, ibid. 23, No. 3, 733--734 (2016; Zbl 1441.14054)], the first statement of Corollary 2 of this paper is incorrect. This error unfortunately propagates through the paper and invalidates the proofs of the main results, Theorem 1 and Corollary 1.