Twists and braids for general 3-fold flops

Abstract: Given a quasi-projective 3-fold X with only Gorenstein terminal singularities, we prove that the flop functors beginning at X satisfy higher degree braid relations, with the combinatorics controlled by a real hyperplane arrangement H. This leads to a general theory, incorporating known special cases with degree 3 braid relations, in which we show that higher degree relations can occur even for two smooth rational curves meeting at a point. This theory yields an action of the fundamental group of the complexified complement of H on the derived category of X, for any such 3-fold that admits individually floppable curves. We also construct such an action in the more general case where individual curves may flop analytically, but not algebraically, and furthermore we lift the action to a form of affine pure braid group under the additional assumption that X is Q-factorial. Along the way, we produce two new types of derived autoequivalences. One uses commutative deformations of the scheme-theoretic fibre of a flopping contraction, and the other uses noncommutative deformations of the fibre with reduced scheme structure, generalising constructions of Toda and the authors which considered only the case when the flopping locus is irreducible. For type A flops of irreducible curves, we show that the two autoequivalences are related, but that in other cases they are very different, with the noncommutative twist being linked to birational geometry via the Bridgeland-Chen flop-flop functor.

Review text

Let \(f: X \to X_{\text{con}}\) be a algebraic 3-fold flopping contraction with \(X\) having Gorenstein terminal singularities. By [\textit{T. Bridgeland}, Invent. Math. 147, No. 3, 613--632 (2002; Zbl 1085.14017)] and [\textit{J.-C. Chen}, J. Differ. Geom. 61, No. 2, 227--261 (2002; Zbl 1090.14003)], \(X\) and its flop is derived equivalent, and one can construct braid group action on the bounded derived category \(D(X)\) of coherent sheaves \(X\) by braid relations of flop functors in some specific situation. This paper establishes a general construction of higher degree braid group action on \(D(X)\). As explained in the introduction of the paper, the main results are divided intro three parts. As for the first part, assume that in the contraction \(f\) each of the \(n\) irreducible exceptional curves is individually floppable. Then by [\textit{M. Wemyss}, Invent. Math. 211, No. 2, 435--521 (2018; Zbl 1390.14012)], the flop functors are controlled by a real hyperplane arrangement \(\mathcal{H}\) which need not be Coxeter. In Theorem 3.23, it is shown that \(D(X)\) has an action of the fundamental group of the complexified complement of \(\mathcal{H}\) by the flop functors. Thus one finds actions of higher degree braid group on \(D(X)\). The second part introduces the notion of \(J\)-twists, which is used to give a geometric description of generators of the action constructed, and is also used to study the case when the exceptional curves of \(f\) are not individually floppable. For each subset \(J\) of the curves in the fibers of \(f\), the \(J\)-twist is an autoequivalence of \(D(X)\) corresponding to monodromy around a wall of codimension \(|J|\). This autoequivalence is constructed by the noncommutative algebra \(A_J\) representing the functor of simultaneous noncommutative deformations of the reduced fibers in \(J\). Explicitly, \(A_J\) is the quotient of the noncommutative algebra \(A\) by idempotents corresponding to the curves in \(J\), where \(A\) is the one derived equivalent to the formal fiber of \(f\) [\textit{M. Van den Bergh}, Duke Math. J. 122, No. 3, 423--455 (2004; Zbl 1074.14013)]. See Theorem 5.23 for the precise statement. The third part gives another new construction of autoequivalences, called fiber twists, using commutative deformations of exceptional fibers. Similarly as \(J\)-twists, fiber twists are also constructed from the quotient \(A_0\) of \(A\) by all the idempotents. See Theorem 5.23 for the precise statement. Fiber twists does not belong to the subgroup of autoequivalences generated by \(J\)-twists. In the type A case, these two twists are conjugate by line bundle twist, as shown in Theorem 6.5. Both the construction of \(J\)-twists and fiber twists and the proof that they are twist autoequivalences are based on the theory of mutations. In particular, the construction of fiber twists is based on the maximal mutation algebras developed in [\textit{O. Iyama} and \textit{M. Wemyss}, Invent. Math. 197, No. 3, 521--586 (2014; Zbl 1308.14007)]. This article gives a brief recollection and an excellent application of these topics.

Recommendations

Cited in

(26)

This page was built for publication: Twists and braids for general 3-fold flops

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2418883)