Universal Length 2 Flop: Geometry & Algebra

Updated 8 July 2026

Universal Length 2 Flop is the canonical model for simple threefold flops with length 2, defined via a D4 partial resolution that isolates the central vertex.
It admits compatible geometric, noncommutative, and hypersurface descriptions, including simultaneous partial resolutions and a quiver presentation with explicit relations.
The construction supports advanced techniques such as GIT window methods, contraction algebras, and derived monodromy, serving as a critical test case in modern algebraic geometry.

The universal length $2$ flop is the universal model for simple threefold flops whose Katz–Morrison length invariant is $2$. In the formulation of Karmazyn, it is obtained from the $D_4$ partial resolution that retains only the exceptional component corresponding to the central $D_4$ -vertex, and every simple flop of length $2$ is, locally and after shrinking, a pullback of this family. The model admits mutually compatible geometric, noncommutative, and hypersurface descriptions: a pair of simultaneous partial resolutions $Y_2 \to X_2 \leftarrow Y_2^+$ over a $4$-dimensional base, a universal flopping algebra presented by a quiver with relations, and the Curto–Morrison hypersurface

$x^2+uy^2+2vyz+wz^2+(uw-v^2)t^2=0.$

It is the first genuinely nontrivial case beyond length $1$, and recent work treats it as a concrete test case for GIT window techniques, contraction algebras, and derived monodromy (Karmazyn, 2017).

1. Length invariant and the place of length $2$

A simple threefold flop is a flop of smooth threefolds whose exceptional locus is an irreducible curve, necessarily $2$0. Equivalently, one considers a simple threefold flopping contraction

$2$1

contracting an irreducible rational curve $2$2 to a point. Kollár’s length invariant is defined as the generic rank of

$2$3

on $2$4. Katz–Morrison proved that every simple threefold flopping contraction has length $2$5 or $2$6, that the length is uniquely determined by the generic hyperplane section, and that this generic hyperplane section is a partial resolution of a Kleinian surface singularity of one of six types. This invariant is finer than the normal bundle invariant: length $2$7 corresponds to the $2$8 and $2$9 cases, whereas the $D_4$ 0 flops split into lengths $D_4$ 1. Accordingly, length $D_4$ 2 is the first genuinely nontrivial $D_4$ 3-type case (Karmazyn, 2017).

2. Geometric construction from the $D_4$ 4 partial resolution

For length $D_4$ 5, the relevant generic hyperplane section is a partial resolution of the Kleinian singularity of type $D_4$ 6. In Karmazyn’s notation, one takes the idempotent $D_4$ 7, so the distinguished non-extended vertex is the central $D_4$ 8-vertex $D_4$ 9; equivalently, one contracts all curves except the one corresponding to that central node. Brieskorn–Tjurina theory supplies the versal deformation of the $D_4$ 0 singularity, the versal deformation of the partial resolution, and the quotient descriptions

$D_4$ 1

In the length $D_4$ 2 case there are exactly two simultaneous partial resolutions over

$D_4$ 3

producing the universal flop

$D_4$ 4

The subgroup is

$D_4$ 5

generated by $D_4$ 6, and the invariant generators may be chosen as

$D_4$ 7

so the base is

$D_4$ 8

A complementary geometric description, used in later GIT work, presents the universal length $D_4$ 9 flop as a $2$0-dimensional resolution $2$1 of

$2$2

with reducible singular locus consisting of the components

$2$3

and a second component $2$4 whose closure meets $2$5 in the quadric cone $2$6. Its exceptional locus has two components: $2$7, a family of conics, and $2$8, generically a $2$9-family (Varshney, 7 Aug 2025).

3. Universal flopping algebra and moduli reconstruction

The universal flopping algebra of length $Y_2 \to X_2 \leftarrow Y_2^+$ 0 is

$Y_2 \to X_2 \leftarrow Y_2^+$ 1

with distinguished idempotent $Y_2 \to X_2 \leftarrow Y_2^+$ 2 corresponding to $Y_2 \to X_2 \leftarrow Y_2^+$ 3. For $Y_2 \to X_2 \leftarrow Y_2^+$ 4, Karmazyn gives an explicit presentation over

$Y_2 \to X_2 \leftarrow Y_2^+$ 5

by a quiver with two vertices $Y_2 \to X_2 \leftarrow Y_2^+$ 6 and $Y_2 \to X_2 \leftarrow Y_2^+$ 7, arrows

$Y_2 \to X_2 \leftarrow Y_2^+$ 8

and loops

$Y_2 \to X_2 \leftarrow Y_2^+$ 9

subject to the relations

$4$0

$4$1

$4$2

Thus

$4$3

This algebra is obtained from the $4$4 simultaneous-resolution algebra by passing to invariants and performing the change of variables

$4$5

It carries a positive grading in which $4$6 have degree $4$7, while $4$8 have degree $4$9.

The geometric spaces are recovered by a King–GIT type moduli construction. One has

$x^2+uy^2+2vyz+wz^2+(uw-v^2)t^2=0.$ 0

and

$x^2+uy^2+2vyz+wz^2+(uw-v^2)t^2=0.$ 1

where $x^2+uy^2+2vyz+wz^2+(uw-v^2)t^2=0.$ 2 is the moduli of $x^2+uy^2+2vyz+wz^2+(uw-v^2)t^2=0.$ 3-generated $x^2+uy^2+2vyz+wz^2+(uw-v^2)t^2=0.$ 4-modules of fixed dimension vector. Since the non-free indecomposable MCM module has rank $x^2+uy^2+2vyz+wz^2+(uw-v^2)t^2=0.$ 5, the relevant dimension vector is

$x^2+uy^2+2vyz+wz^2+(uw-v^2)t^2=0.$ 6

Accordingly, the vertex $x^2+uy^2+2vyz+wz^2+(uw-v^2)t^2=0.$ 7 carries a $x^2+uy^2+2vyz+wz^2+(uw-v^2)t^2=0.$ 8-dimensional space, the vertex $x^2+uy^2+2vyz+wz^2+(uw-v^2)t^2=0.$ 9 carries a $1$0-dimensional space, and the $1$1-generated condition means that the image of the $1$2-vertex generates the whole module. Example $1$3 computes this moduli space explicitly by two affine charts, according to whether $1$4 or $1$5, together with transition maps on the overlap (Karmazyn, 2017).

4. Hypersurface model, maximal Cohen–Macaulay modules, and specialisations

A central commutative algebra attached to the universal length $1$6 flop is

$1$7

Karmazyn exhibits generators

$1$8

together with

$1$9

satisfying an initial degree-$2$0 relation. After the change of variables

$2$1

$2$2

this relation becomes

$2$3

so that

$2$4

This is exactly the Curto–Morrison hypersurface form of the universal flop of length $2$5.

The algebra also determines rank-$2$6 maximal Cohen–Macaulay modules

$2$7

and these are encoded by an explicit $2$8 matrix factorisation of $2$9. In Karmazyn’s computation, $2$00 is the cokernel of a matrix $2$01, $2$02 is the cokernel of a matrix $2$03, and $2$04 up to basis and sign conventions. This realizes noncommutatively the Curto–Morrison matrix factorisation description.

The length $2$05 case is also the paper’s principal worked specialisation. Under

$2$06

one obtains

$2$07

which is the Laufer flop. The specialised algebra has a quiver presentation with loops $2$08, generated by the superpotential

$2$09

and its contraction algebra is

$2$10

The dimensions recorded in the paper are

$2$11

yielding GV invariants $2$12, $2$13, and confirming that the length is $2$14 (Karmazyn, 2017).

5. GIT, derived autoequivalences, and monodromy

Recent work places the universal length $2$15 flop in a non-abelian linear GIT framework. The universal flop is realised as a quotient of

$2$16

with $2$17 $2$18-dimensional, and equivalently from a quiver GIT problem with dimension vector $2$19. Two stability chambers produce two birational models $2$20 and $2$21, and the resulting birational map is the universal length $2$22 flop. Applying Halpern-Leistner–Sam window theory, the associated stringy Kähler moduli space is computed to be a sphere with north and south pole punctures and two equatorial punctures. The two equatorial monodromies are identified with the fibre twist and the contraction algebra twist, so the GIT/windows description agrees with the Hirano–Wemyss picture via Bridgeland stability and recovers the Donovan–Wemyss monodromy action (Varshney, 7 Aug 2025).

This categorical picture fits into a broader theory of flop autoequivalences. For flops over canonical hypersurface singularities of multiplicity $2$23, Bodzenta and Bondal prove that the standard flop functors are derived equivalences and that the composite flop–flop functor is a nontrivial autoequivalence. When the base is affine, they identify this autoequivalence as the spherical cotwist associated to the derived inclusion

$2$24

of the null-category

$2$25

and they show that the endomorphism algebra of a projective generator of $2$26 is the contraction algebra (Bodzenta et al., 2015).

A complementary formalism due to Barbacovi treats any roof

$2$27

for which the pull–push functors are equivalences. In that setting, the flop–flop autoequivalence is the inverse of the spherical twist around a conservative spherical functor from the quotient source category $2$28 or $2$29. This gives a uniform mechanism for realising flop–flop as spherical monodromy and provides a natural source category to study in concrete examples (Barbacovi, 2020).

A string-theoretic reconstruction of the universal length $2$30 flop appears in work on D2-branes probing deformed ADE surfaces. There the universal length $2$31 flop is a family of $2$32 surfaces over

$2$33

with partial simultaneous resolution that blows up only the central $2$34 root. The paper states that the length equals the dual Coxeter label of the Dynkin node corresponding to the simple root that is blown up, so the central $2$35 node gives length $2$36. Using an explicit $2$37-valued Higgs field, it derives the hypersurface normal form

$2$38

and matches the effective quiver relations with Karmazyn’s universal flopping algebra of length $2$39. The same analysis distinguishes loci where the fibre is singular from loci where the total space is singular, and interprets the surviving monopole operators as charge-$2$40 and charge-$2$41 wrapped-brane states (Moleti et al., 2024).

A different, tropical viewpoint studies the discrete family

$2$42

which is presented in the supplied summary as the basic local model for $2$43-flops. The paper does not construct a universal deformation family, but it does build an integral affine manifold with singularities $2$44, a cone $2$45, and exactly two subdivisions of $2$46 corresponding to the two sides of the flop. This suggests a tropical replacement for toric combinatorics in the length-$2$47 setting: the Atiyah-flop picture of a cone with two triangulations survives, but only after replacing an ordinary toric lattice by an affine manifold with monodromy (Ducat, 2021).

The terminology should also be separated from an unrelated usage of “flip-flop” in quantum information. The paper "Universal set of quantum gates for the flip-flop qubit in the presence of $2$48 noise" studies silicon flip-flop qubits and proposes the gate set

$2$49

but it is not about threefold flops or the algebro-geometric notion of a universal length $2$50 flop (Ferraro et al., 2021).