Flips of semi-stable 4-folds whose degenerate fibers are unions of Cartier divisors which are terminal factorial 3-folds (Q1359507)

3-dimensional flips are now well-understood by virtue of the works of Reid, Kawamata, Mori, Kollár, Shokurov, and others. So we are interested in flips in dimension 4. In this article the author deals with flips from semi-stable 4-folds, with a certain additional assumption on singularities ((3) below). This should be thought as the very first step toward the understanding of the minimal model problem in dimension 4. Let \(f:X\to\Delta\) be a projective surjective morphism from a 4-fold \(X\) with terminal singularities to the 1-dimensional disc \(\Delta\), satisfying the following conditions: (1) Each fiber of \(f\) over \(\Delta-\{0\}\) is a 3-fold with terminal singularities, (2) The central fiber \(D\) is reduced, and \(D\) is normal crossing outside the singular locus of \(X\), (3) Each irreducible component of \(D\) is a Cartier divisor which is a terminal factorial 3-fold. Let \(g:X\to Y\) be a flipping contraction over \(\Delta\), i.e. a contraction determined by an extremal ray of \(\overline{NE} (X/\Delta)\), with \(\text{dim Exc } g\leq 2\). Then the author completely classifies the local structure of \(g\) along \(\text{Exc }g\). Let \(E\) be any connected component of \(\text{Exc }g\), and let \(D=\bigcup D_k\) be the irreducible decomposition of \(D\), then his result says \(E\simeq\mathbb{P}^2\), and that \(g\) fits exactly into one of the following three types: (A) \(E\subset D_1\cap D_2\), \(E\cap D_3\), \(E\cap D_4\) are both lines, and \(E\cap D_k=\emptyset\) \((k\geq 5)\), (B) \(E\subset D_1\cap D_2\), \(E\cap D_3\) is a smooth conic, and \(E\cap D_k=\emptyset\) \((k\geq 4)\), or (C) \(E\subset D_1\), \(E\cap D_2\) is a line, and \(E\cap D_k=\emptyset\) \((k\geq 3)\), after a suitable renumbering of the \(D_k\)'s. In each case \(X\) is found to be smooth along \(E\), so by combining Kawamata's flip theorem from smooth 4-folds, one concludes that the flip \(g^+\) of \(g\) certainly exists. The author also completely describes the local structures of the flips \(g^+:X^+\to Y\) along their exceptional loci (called flipped curves), for each type of \(g\)'s. Especially he remarkes that once we do the flip \(g^+: X^+\to Y\) of \(g:X\to Y\) of type (B), then the resulting \(X^+\) no longer satisfies the assumption (2) above. Actually in this case the proper transform of \(D_3\) has singularities entirely along the flipped curve, and this in fact gives the anti-flip of the 3-fold \(D_3\).