# Flip (mathematics)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

In algebraic geometry, flips and flops are codimension-2 surgery operations arising in the minimal model program, given by blowing up along a relative canonical ring. In dimension 3 flips are used to construct minimal models, and any two birationally equivalent minimal models are connected by a sequence of flops. It is conjectured that the same is true in higher dimensions.

## The minimal model program

{{#invoke:main|main}} The minimal model program can be summarised very briefly as follows: given a variety $X$ , we construct a sequence of contractions $X=X_{1}\rightarrow X_{2}\rightarrow \cdots \rightarrow X_{n}$ , each of which contracts some curves on which the canonical divisor $K_{X_{i}}$ is negative. Eventually, $K_{X_{n}}$ should become nef (at least in the case of nonnegative Kodaira dimension), which is the desired result. The major technical problem is that, at some stage, the variety $X_{i}$ may become 'too singular', in the sense that the canonical divisor $K_{X_{i}}$ is no longer Cartier, so the intersection number $K_{X_{i}}\cdot C$ with a curve $C$ is not even defined.

The (conjectural) solution to this problem is the flip. Given a problematic $X_{i}$ as above, the flip of $X_{i}$ is a birational map (in fact an isomorphism in codimension 1) $f:X_{i}\rightarrow X_{i}^{+}$ to a variety whose singularities are 'better' than those of $X_{i}$ . So we can put $X_{i+1}=X_{i}^{+}$ , and continue the process.

Two major problems concerning flips are to show that they exist and to show that one cannot have an infinite sequence of flips. If both of these problems can be solved then the minimal model program can be carried out. The existence of flips for 3-folds was proved by Template:Harvtxt. The existence of log flips, a more general kind of flip, in dimension three and four were proved by Template:Harvs whose work was fundamental to the solution of the existence of log flips and other problems in higher dimension. The existence of log flips in higher dimensions has been settled by Template:Harvs. On the other hand, the problem of termination—proving that there can be no infinite sequence of flips—is still open in dimensions greater than 3.

## Definition

If f:XY is a morphism, and K is the canonical bundle of X, then the relative canonical ring of f is

$\oplus _{m}f_{*}({\mathcal {O}}_{X}(mK))$ and is a sheaf of graded algebras over the sheaf OY of regular functions on Y. The blowup f+

$f^{+}:X^{+}=Proj(\oplus _{m}f_{*}({\mathcal {O}}_{X}(mK)))\to Y$ of Y along the relative canonical ring is a morphism to Y. If the relative canonical ring is finitely generated (as an algebra over OY) then the morphism f+ is called the flip of f if −K is relatively ample, and the flop of f if K is relatively trivial. (Sometimes the induced birational morphism from X to X+ is called a flip or flop.)

In applications, f is often a small contraction of an extremal ray, which implies several extra properties:

• The exceptional sets of both maps f and f+ have codimension at least 2,
• X and X+ only have mild singularities, such as terminal singularities.
• f and f+ are birational morphisms onto Y, which is normal and projective.
• All curves in the fibers of f and f+ are numerically proportional.

## Examples

The first example of a flop, known as the Atiyah flop, was found in Template:Harv. Let Y be the zeros of xy = zw in A4, and let V be the blowup of Y at the origin. The exceptional locus of this blowup is isomorphic to P1×P1, and can be blown down to P1 in 2 different ways, giving varieties X1 and X2. The natural birational map from X1 to X2 is the Atiyah flop.

Template:Harvtxt introduced Reid's pagoda, a generalization of Atiyah's flop replacing Y by the zeros of xy = (z+wk)(zwk).