WorksMathematics
Quantum Cohomology and irrationality of Küchle fourfolds
Abstract
We prove that very general Küchle fourfolds of types (c1), (c6), (b8), and (b11) are irrational. The key ingredients are a new rationality criterion, based on the theory of Hodge atoms due to Katzarkov–Kontsevich–Pantev–Yu and its integral extension by Guéré, and a computation of the action of quantum multiplication on the vanishing cohomology of these fourfolds, which uses a version of the Quantum Lefschetz Hyperplane Principle. Quantum multiplication acts as a scalar on the vanishing cohomology, and we give an enumerative interpretation of this scalar.
Introduction
In [Küc95], Küchle classified all smooth index-one Fano fourfolds that arise as the zero locus of a regular section of a globally generated, completely reducible homogeneous vector bundle on a Grassmannian. Our main result concerns their rationality.
Theorem 1.1. Very general Küchle fourfolds of types (b8), (b11), (c1), and (c6) are irrational.
What makes the birational geometry of these fourfolds particularly compelling is that their middle cohomology carries a large surface-type Hodge structure that is not of K3 type. Whereas the obstructions to rationality for cubic, Gushel–Mukai, and Verra fourfolds [Gué26b, BMP26, Gué26a] involve K3 surfaces, rationality of a very general fourfold of type (c1) or (c6) would imply the existence of a projective surface of geometric genus 5 or 2, respectively.
The proof of Theorem 1.1 rests on a rationality criterion for Fano fourfolds, which we now state. Let be a smooth projective complex fourfold and define
where orthogonality is taken with respect to the intersection pairing ; we write . For a smooth projective surface , write
We use the following notation for a nondegenerate lattice , as recalled in Section 5:
| the discriminant group | |
| the minimal number of generators of |
As we will describe in Section 4.1 below, the cohomology group , where is a suitable coefficient field, carries an endomorphism given by quantum multiplication by the Euler vector field, depending on parameters in the direction of rational Hodge classes. If satisfies (1.1) below, then acts on the transcendental cohomology by a scalar, as we show in the proof of Proposition 1.2: let us call this scalar . Put , and define, for general ,
Here is the dimension of the generalized eigenspace of containing , while is the specialization of this quantity to the small quantum locus . Since multiplicities can only jump up at special points, .
Proposition 1.2 (Quantum Noether integral obstruction). Let be a smooth projective complex fourfold satisfying
Let . If is rational, there is a smooth minimal projective surface and an integral Hodge isometry such that:
(1) We have
and in particular, .
(2) The Picard-rank of satisfies
(3) If , then
the group is cyclic, and
The criterion also recovers the recent result of Guéré:
Corollary 1.3 (Guéré, [Gué26a, Corollary 28]). A very general Verra fourfold is irrational.
| Kŭchle Type | Rationality Criterion | ||||
| (b8) | 3 | 42 | 45 | ||
| (b11) | 2 | 32 | 35 | ||
| (c1) | 5 | 60 | 63 | ||
| (c6) | 2 | 32 | 35 |
Table 1. Rationality obstructions for very general Kŭchle fourfolds obtained from Proposition 1.2.
Related work
Related work. Amongst the twenty-one families of Küchle varieties [Küc95], only those of types (c5), (c7), and (d3) are of K3 type [4 Section 2.3]. Kuznetsov studied the geometry of the varieties of types (b4), (b9), (c7), and (d3) in [30], where he proved type (b9) to be rational [30 Theorem 2.5].
Kuznetsov also studied the birational geometry of the Küchle fivefold of type (c5) in [31]. Taking its smooth hyperplane section gives the Küchle fourfold of type (c5), whose very general member was proven to be irrational by Benedetti–Fay–Guéré–Manivel–Perrin [3 Proposition 5.1]. This result uses Noether’s formula to bound the rank of the transcendental part of a surface atom [3 Theorem 4.1]; here we combine it with a dimension bound from small quantum multiplication and integral lattice data. The type (c5) irrationality result was recovered by Alexeev–Cavenaghi–Galindo–Georgiev–Katzarkov–Martins in [2], who also prove irrationality of very general Küchle fourfolds of types (c2) and (b10) [2 Theorem 6.13 and 6.16]. Both papers have related monodromy arguments concerning the scalar action on vanishing cohomolology; see [3 proof of Theorem 4.1] and [2 Theorem 5.35].
For the fourfold of type (c7), Kuznetsov identified it with the blowup of a cubic fourfold along a Veronese surface in [30 Corollary 4.11]. Using this description, the author proved a very general Küchle fourfold of type (c7) to be irrational [15 Corollary 1.4].
The motivating thread throughout this paper’s work on birational geometry is the work of Katzarkov–Kontsevich–Pantev–Yu in [27]. To apply it to the fourfolds considered here, we use Guéré’s recent extension to integral Hodge structures [Gué26a, Proposition 26]. Our approach for understanding and recovering the relevant quantum cohomology is driven by the Quantum Lefschetz and Riemann–Roch methods of Coates–Givental [10, 7]. Our motivation for applying it within the context of Fano fourfolds finds itself in the large number of quantum periods computed by Coates–Galkin–Kasprzyk–Strangeway in [8].
Outline
Outline. In Section 2 we show that quantum multiplication by the hyperplane class acts by a scalar on vanishing cohomology, and that this scalar is a repeated eigenvalue on the inherited part of cohomology with a nonzero isotropic eigenvector. In Section 3, we express this scalar as a line count and construct embeddings of vanishing and transcendental cohomology using the Fano surface of lines. We recall the Quantum Lefschetz theorem in Section 4 and apply it to the Küchle fourfolds of type (c1), (c6), (b8), and (b11). In Section 5 we prove our rationality criterion, which we apply to these Küchle fourfolds in Section 6. We also apply it to Verra fourfolds in Section 7. All varieties throughout this paper are smooth and projective over .
Monodromy and Vanishing Cohomology
Let be a smooth hyperplane section of a smooth projective fivefold, and let denote the inclusion. Assume that is an index-one Fano fourfold of Picard-rank one with . Write and set .
Define the inherited and vanishing cohomology by
where orthogonal complements are taken with respect to the Poincaré pairing. By weak and hard Lefschetz, the pairing on is nondegenerate and we have the orthogonal decomposition
Let be the locus of smooth hyperplane sections and . The local system of vanishing cohomology over defines the monodromy representation
where is the monodromy group; see [11 §5].
Gromov–Witten invariants
We briefly recall stable maps and Gromov–Witten invariants; see [28, 35]. For a smooth projective variety and a class , let be the moduli space of stable maps
Here is a connected projective nodal curve of arithmetic genus-zero with distinct smooth marked points , and every irreducible component of contracted by contains at least three special points (marked points or nodes).
For each , there is an evaluation map
The space carries a virtual fundamental class of expected dimension .
The genus-zero Gromov–Witten invariants of are then the rational virtual intersection numbers
where is the first Chern class of the line bundle whose fibre at is the cotangent line of at ; until Section 4 all are zero. Note that these invariants do not change when is deformed, and they satisfy the divisor equation
for and . When carries a hyperplane class , we write an integer in place of for the disjoint union of moduli spaces, or the sum of the invariants, over all with .
The quantum product
For the fourfold with , the small quantum product on is the -bilinear product defined by
Here the Novikov variable records the degree and the term with is . By the dimension formula, the coefficient of in has degree , so the sum is finite. The product is graded commutative, associative, and has unit .
Write
The small quantum product makes
a Frobenius algebra: that is, a finite-dimensional unital commutative associative -algebra equipped with a nondegenerate symmetric bilinear pairing satisfying .
This compatibility, together with the monodromy equivariance of proved in Proposition 2.1, leads to Proposition 2.2: an abstract Frobenius algebra version of Hu’s eigenvector construction for projective complete intersections; see [20 Proposition 5.1].
Proposition 2.1. If the pairing on is indefinite, then is Zariski dense in the orthogonal group . Consequently, the small quantum product on is -equivariant—with acting trivially on and by its standard representation on .
Proof. By standard theory, is generated by Picard–Lefschetz reflections in the vanishing cycles of a Lefschetz pencil through . These cycles span and form a single orbit up to sign. Thus is an absolutely irreducible -module.
By Deligne’s reflection-group alternative [12 Lemme 4.4.2s] (see also [26 Part II] for a statement in English), the Zariski closure of is either a finite reflection group or . It cannot be finite, as assuming so contradicts the stated indefiniteness: a finite group preserves a positive definite form on , which by absolute irreducibility and Schur’s Lemma is proportional to the pairing, which would then be definite.
Finally, deformation invariance of Gromov–Witten invariants makes the small quantum product equivariant under . This is a Zariski-closed condition, and hence extends to and, after extending scalars, to .
Proposition 2.2. Let be an orthogonal decomposition of a Frobenius algebra over a field of characteristic zero, with and , whose product is -equivariant as in Proposition 2.1.
Then is a subalgebra, and there are a unique character and a unique class such that
for all and ; in particular . Moreover,
so that is isotropic and . Finally, for , writing and ,
Proof. Equivariance of the quantum product under gives
The standard representation of is absolutely irreducible, so Schur’s lemma gives . Since is associative and unital, is an algebra homomorphism. Nondegeneracy of the pairing on provides a unique with for all . For and , the Frobenius property then gives
so , again by nondegeneracy. This proves (2.1).
Now fix with , so that . For , associativity gives . Since , there is a nonzero orthogonal to ; then , so
and therefore . This proves (2.2), whence and $b_\chi \star b_\chi = \chi(b_\chi)b_\chi = 0.
Finally, because , so is a line in the hyperplane . Both are preserved by , which acts by on and on , since and for . This gives (2.3).
When has more than one repeated root, as for types (b11) and (c1) below, (2.1) and (2.2) decide between them, as the eigenspace of contains the isotropic class . For an example of this, see the proof of Proposition 4.6.
Lines and the Cylinder Correspondence
Keeping the notation of the previous section, we identify the scalar as times the number of lines of through a general point. Although we do not use this result in the proof of Theorem 1.1, where is found from (2.2) and (2.3), it provides a geometric interpretation of its value. By a line we mean a curve of degree one in , where degrees of stable maps to and are taken with respect to .
Proposition 3.1 (Lines and the vanishing eigenvalue). Let be a smooth hyperplane section of a smooth projective fivefold with , where is the hyperplane class. Suppose that the Fano schemes of lines and are smooth of dimensions 4 and 2, and let be the number of lines of through a general point. Then
In particular, if acts on by a scalar, then this scalar is .
Proof. Since is very ample, a degree one stable map with two marked points is a line with . When these points collide, they move into a new component or “bubble”, attached to the line at a node, and contracted by the map to a single point of . In either case,
where denotes the universal line. The stable maps with a bubble form the diagonal of , which is a copy of . By assumption , and are smooth of dimensions 6, 4 and 5, which are the expected dimensions since and . In particular, the evaluation map has degree .
Let denote the evaluation maps and put
We claim that
where is the diagonal.
A line in either lies in or meets in a single point. Hence
and both pieces have dimension . Each occurs with multiplicity one. Indeed, if defines , then is cut out by . Where , these equations say that vanishes identically on , so they cut out , which is smooth of codimension two.
Near a line , write for nearby lines . Then and , which again cut out a smooth subvariety of codimension two. The remaining locus consists of the bubbles on lines contained in and has dimension three. Thus
by [16 Proposition 7.1]. Pushing forward by [16 Theorem 6.2(a)] gives (7): the map factors through the diagonal and the induced map satisfies
by compatibility of proper pushforward with the refined pullback along .
Now let and . Since is orthogonal to , we have , and so
By (7) this reads , which proves (6): on vanishing classes, the contribution of the lines of cancels that of the bubbles.
Finally, the coefficient of in lies in , and by the divisor equation the coefficient of is
where and are dual bases of . So if , then .
Gromov–Witten invariants and the decomposition are invariant under deformations of the pair . So the conclusions of Proposition 3.1 hold as soon as can be deformed to a pair for which and are smooth of the expected dimensions.
Corollary 3.2 (The cylinder correspondence). In the situation of Proposition 3.1, let be the universal line over and its evaluation map, and let
be the Abel–Jacobi map.
In particular, if , then restricts to embeddings
the second provided that consists of algebraic classes, so that .
Proof. Since , the projection formula gives for all , so (8) is (6). It follows that is injective on when , and is a morphism of Hodge structures of bidegree because it is induced by an algebraic correspondence. Finally, if and , then , because is an algebraic class: so maps to .
The map is the transpose of the cylinder map studied by Lewis in [33, 34] for hypersurfaces of degree in ; see also [38]. For these hypersurfaces, where , the eigenvalue is due to Collino–Jinzenji [9 Proposition 2].
Quantum Cohomology of Küchle fourfolds
In this section, we compute the small quantum Euler operator of the Küchle fourfolds of types (c1), (c6), (b8), and (b11). In each case, Quantum Lefschetz gives the restriction of to the inherited cohomology , and the results of Section 2 then determine its action on the vanishing cohomology . We briefly recall Quantum Lefschetz in §4.1 and outline the method in §4.2.
Quantum Lefschetz
We use Quantum Lefschetz in the form of [10 Theorem 1.1], which relates Givental’s Lagrangian cone of a zero locus to the Euler-twisted cone of the ambient variety. In this subsection only, the inclusion denotes the zero locus of a regular section of a convex vector bundle on a smooth projective variety .
The genus-zero Gromov–Witten invariants of are encoded by a Lagrangian cone in the symplectic space
Here is the Novikov ring whose monomials record degrees of effective curve classes .
A general point of has the form
where with , and and are Poincaré dual bases of . The -function is the unique point of of the form .
Replacing the invariants of by the invariants of twisted by the -equivariant Euler class of , with equivariant parameter , gives a cone with a general point . It lives in the corresponding equivariant symplectic space for with the twisted pairing
Theorem 4.1 (Quantum Lefschetz, [10 Theorem 1.1]). Let be the smooth zero locus of a regular section of a convex vector bundle . The non-equivariant limit of exists and satisfies
In particular, . Throughout the homomorphism is applied to .
Let be a homogeneous basis of , with , and write . We recover the big quantum product from the -function by Coates–Givental [7 Equation 16]:
This formulation is particularly useful when inherits classes from that are not powers of the hyperplane class, as in our examples. It provides the input we need to compute the Hodge atom birational invariants of Katzarkov–Kontsevich–Pantev–Yu in [27].
As discussed in the introduction, to use these invariants we study the generalized eigenspaces of quantum multiplication by the Euler vector field
At , the product is the small quantum product of § 2. For the index-one, Picard-rank one fourfolds considered in this paper, we write and obtain
Method
Let be the Grassmannian of -dimensional subspaces of , with tautological subbundle , quotient bundle , and Plücker line bundle . We write for the zero locus of a general regular section of the vector bundle on a variety . Each fourfold considered below has the form
where is a globally generated homogeneous vector bundle on and is a direct sum of copies of ; see Table 2.
| Küchle Type | ||||||
| (b8) | 3 | 38 | 44 | |||
| (b11) | 2 | 31 | 35 | |||
| (c1) | — | 5 | 52 | 62 | ||
| (c6) | 2 | 30 | 34 |
Table 2. Küchle fourfolds from [Küc95, Theorem 3] and their middle Hodge numbers.
By Küchle [Küc95, Theorem 3], is a Fano fourfold of index one, so that
It satisfies the standing assumptions of § 2, namely, it is Picard-rank one with .
Now let be a general five-dimensional linear section of containing . It is a smooth fivefold with hyperplane section , and we apply § 2 to . By the Lefschetz hyperplane theorem is an isomorphism for , while and are spanned by and . Hence the inherited cohomology of § 2 is the image of . We put , so that is the orthogonal complement of in .
Lemma 4.2. Suppose that is spanned by algebraic classes. Then
and the Poincaré pairing on has signature .
Proof. Since is Fano, . The subspace is the orthogonal complement of , which is of type ; hence it is a sub-Hodge structure of with the stated Hodge numbers. Let . Since we have , and therefore . Thus consists of primitive classes, and Hodge–Riemann gives the stated signature relations. □
Corollary 4.3. Suppose that the middle inherited cohomology is spanned by algebraic classes and that and . Then is stable under , and there is with the following properties.
for all , and is a repeated root of .
There is a unique class such that for all . It is isotropic and satisfies and .
.
Proof. By Lemma 4.2, the Poincaré pairing on is indefinite. Hence Proposition 2.1 applies to , and the small quantum product is -equivariant. Proposition 2.2 with shows that is invariant under , and also gives (1) and (2): because the pairing on is nonzero, the class is unique. Part (3) follows from (1). □
Inherited matrix computation. Assume that Corollary 4.3 applies, and let be a homogeneous basis of , with .
As has index one, has degree one, and the degree axiom gives
with whenever . We call
the inherited matrix of in the basis ; it is the matrix of .
Writing , we have , so that
Note that the entries of this matrix, for , are the structure constants of the cup product with . Moreover, is self-adjoint for the Poincaré pairing, so is symmetric, where is the Gram matrix of .
To compute , we apply Theorem 4.1 to a family of points of the twisted cone from § 4.1. This gives
and in each case for some . By the string equation and the characterization of in §4.1,
Writing and applying (11) to our normalization of the family at gives . Equivalently, if is the matrix of , then
Remark 4.4 (String shift and the vanishing eigenvalue). Let be the coefficient of the unit-coordinate correction in (4.8). We show that it is the number of lines of through a general point. Indeed, the hypergeometric modification for the last hyperplane section multiplies the degree- term of by . Since has index two, the coefficient of in is with , and multiplying by gives
Thus the unit-coordinate correction is , which agrees with Iritani’s mirror-map formula for a hyperplane section of an index-two Fano variety [24 Section 0.2].
Because (18) is deformation invariant, this holds for every smooth pair of the given type. For types (b8), (b11), (c1), (c6) we have , which agree with the vanishing eigenvalues computed in §4.
Consequently, the translation in (17) gives an operator annihilating the vanishing summand:
This gives a cohomological interpretation of what Golyshev refers to as the Givental constant in [17 Section 1.2]. There, he explains that the Quantum Lefschetz comparison of spectra requires an additional translation when the hyperplane section has index one. After this translation, the vanishing cohomology contributes only zero eigenvalues to the shifted operator, which is the behavior suggested by Iritani’s question on extending the Quantum Lefschetz comparison of characteristic polynomials to include primitive cohomology [24 Question 15].
Applications to Küchle fourfolds
Küchle fourfold of type (b8)
Let and
Put . The inherited cohomology has the basis
and its Poincaré pairing is determined by the intersection numbers
Since the classes and are algebraic, Lemma 4.2 gives and . In particular, and Corollary 4.3 applies.
Proposition 4.5 (Type (b8)). Let be a Küchle fourfold of type (b8).
In the basis , the matrix of is
Consequently,
For all ,
where and the class
spans the -eigenspace of .
(3) We have
In particular .
Proof. (1) Apply the method of § 4.2 to the Euler-twisted Grassmannian -function of , twisted by , constructed using the abelian/non-abelian correspondence of Inoue–Ito–Miura [22 Definition 3.3, Theorem 3.4]. This gives and the characteristic polynomial follows.
(2) By Proposition 2.2, acts on by a scalar which is a repeated root of the polynomial from the previous step. Hence . A direct computation with shows that the -eigenspace of is the line spanned by and . By Proposition 2.2, this line contains , and . Hence .
(3) This is Corollary 4.3 since . □
Küchle fourfold of type (b11)
Let and
Let be Meng’s universal bundles on , of ranks two and three [36], and put
Define
The inherited cohomology has the basis
and its Poincaré pairing is determined by , the relation and the Gram matrix of , which is
Since the classes are algebraic, Lemma 4.2 gives and . In particular, and Corollary 4.3 applies.
Proposition 4.6 (Type (b11)). Let be a Küchle fourfold of type (b11).
(1) In the basis , the matrix of is
Consequently,
(2) For all ,
where and the class
spans the -eigenspace of .
(3) We have
In particular .
Proof. (1) We obtain from Meng’s quantum Chevalley formula [36 Corollary 5.26], and apply Coates–Givental hypergeometric modification [7 Theorem 2] for . The method of §4.2 then gives and the characteristic polynomial follows.
(2) By Corollary 4.3, acts on by a scalar , which is a repeated root of the polynomial in (1). The repeated roots are and : the factor has discriminant , and does not vanish at nor at .
We exclude because a computation with shows that the -eigenspace of is the plane spanned by
and that
Hence, for ,
which vanishes only for (because is not a square in ). Therefore the eigenspace contains no nonzero isotropic vector, whereas is isotropic. Hence .
A computation with shows that the -eigenspace of is the line spanned by , and . As this line contains and , we get .
(3) This is Corollary 4.3 since . □
Küchle fourfold of type (c1)
Let and
Put . The inherited cohomology has the basis
and its Poincaré pairing is determined by the intersection numbers
Since and are algebraic, Lemma 4.2 gives and . In particular , and Corollary 4.3 applies.
Proposition 4.7 (Type (c1)). Let be a Küchle fourfold of type (c1).
(1) In the basis , the matrix of is
Consequently
(2) For all ,
where and the class
spans the -eigenspace of .
(3) We have
In particular .
Proof. (1) Apply the method of § 4.2 to the Euler-twisted -function of , twisted by , using the abelian/non-abelian correspondence of Inoue–Ito–Miura [22 Definition 3.3, Theorem 3.4]. This gives , and the characteristic polynomial follows.
(2) By Corollary 4.3, acts on by a scalar which is a repeated root of the polynomial in (1), so . We exclude as in the proof of Proposition 4.6: a computation with shows that the -eigenspace of is the plane spanned by
and that , and . Since 3 is not a square in , this plane contains no nonzero isotropic vector, whereas is isotropic. Hence , in agreement with Remark 4.4, where . The -eigenspace of is the line spanned by and . As this line contains and , we get .
(3) This is Corollary 4.3 since .
Küchle fourfold of type (c6)
Let and
Put . The inherited cohomology has the basis
and its Poincaré pairing is determined by the intersection numbers
Since and are algebraic, Lemma 4.2 gives and . In particular , and Corollary 4.3 applies.
Proposition 4.8 (Type (c6)). Let be a Küchle fourfold of type (c6).
(1) In the basis , the matrix of is
Consequently
(2) For all ,
where and the class
spans the -eigenspace of .
(3) We have
In particular .
Proof. (1) Apply the method of § 4.2 to the Euler-twisted Grassmannian -function of , twisted by , using the abelian/non-abelian correspondence of Inoue–Ito–Miura [22 Definition 3.3, Theorem 3.4]. This gives , and the characteristic polynomial follows.
(2) By Corollary 4.3, acts on by a scalar , which is a repeated root of the polynomial in (1). Its cubic factor is squarefree over , and gives the nonzero value at . Thus is the unique repeated root and .
A computation with shows that the -eigenspace of is the line spanned by and . As this line contains and , we get .
(3) This is Corollary 4.3 since . □
The following statement is used in Section 6.
Proposition 4.9. Let be a Küchle fourfold of type (c6) and . Then the classes and generate a saturated sublattice of .
Proof. Let be the alternating 2-forms defining the smooth sixfold , and let be a common 4-dimensional isotropic subspace for and (any pair of alternating 2-forms on admits a common -dimensional isotropic subspace).
Choose , a line , and a 2-plane . Then the Schubert planes
lie in . Writing , standard Schubert calculus gives
Because this pairing matrix is unimodular, and span a saturated rank-2 sublattice of . Since is cut out inside by two hyperplane sections, integral Weak Lefschetz shows that is injective with torsion-free cokernel, so saturation passes to .
Quantum Noether Rationality criterion
In [Gué26a, Proposition 26], Guéré proves an integral Hodge isometry between the transcendental part of a Hodge atom of a rational fourfold and the transcendental lattice of an associated minimal surface. Here we combine it with the Hodge atoms of [27] to prove Proposition 1.2. To do this, we consider the generalized eigenspaces of at general points of the non-archimedean Hodge locus of [27 Section 5.4].
We first recall the elementary lattice theory used below; see Nikulin [Nik80, Section 1].
Lattices
From now on, lattice means a free abelian group of finite rank equipped with a symmetric bilinear pairing that is nondegenerate over . Write
and let denote the minimum number of generators of the discriminant group . A lattice is unimodular if ; the lattices and are unimodular by integral Poincaré duality. A sublattice is primitive (or saturated) if is torsion-free, equivalently if .
If is the Gram matrix of in an integral basis, then
Thus the discriminant group can be computed from the Smith normal form of .
If are lattices of the same rank, then is a finite group. Its number of elements, , is called the index of in . Since their Gram matrices satisfy
a sublattice is saturated whenever is squarefree.
If is primitive and nondegenerate and is unimodular, then , as finite abelian groups [Nik80, Proposition 1.6.1].
Proof of Proposition 1.2. Suppose that is rational. By Guéré’s integral Hodge isometry [Gué26a, Propositions 25–26], applied to a weak factorization of a birational map , there is a surface centre whose smooth minimal model satisfies
as integral Hodge structures. The isometry gives . Thus is not ruled and minimality implies that is nef.
Since is polarizable and , it is a simple Hodge structure and it is not of Tate type (as ), where its orthogonal complement in is entirely Tate by (1.1). Given that Gromov–Witten classes are algebraic, it follows that preserves and acts on it by a scalar . In particular, lives in a single Hodge atom of , which corresponds (under weak factorization) to an atom of and all of ; see [27 Lemma 5.24]. After specialization this gives
The lattices and are primitive orthogonal complements in the unimodular lattice . Hence
Together with and (19), this proves (1) and the lower bound in (2). Noether’s formula gives
Since , the upper bound in (2) follows.
Finally, suppose . Equation (19) requires
Consequently, and
so is cyclic. Writing modulo torsion, with , Noether’s formula becomes
This proves (3). □
Irrationality of Küchle fourfolds
Proposition 6.1. For a very general Küchle fourfold of type (b8), (b11), (c1) or (c6),
Moreover, with as in Section 4, a -basis of is given by in type (b8), in type (b11), and by in types (c1) and (c6), with Gram matrices
In particular is isomorphic to and , , , respectively.
Proof. For very general , the Hodge group of contains the connected algebraic monodromy group ; see [13 Proposition 7.5]. Its standard representation is absolutely irreducible and contains no nonzero invariant vectors. Thus contains no nonzero rational Hodge classes and Schur’s lemma gives:
Since is algebraic, it follows that
Let be the lattice generated by these classes, whose rational span is by § 4.3.
For (c1), is squarefree, so is saturated.
For (b8), standard integral Schubert calculus on gives
Thus is also an integral basis.
For (b11), the classes , , form an integral basis by Meng’s cellular calculation [36 Sections 3.3–4, Proposition 4.4].
For (c6), saturation follows from Proposition 4.9.
In types (b8) and (b11), integral weak Lefschetz through the successive smooth hyperplane sections shows that is primitive in as well. Since its rational span is , this gives .
The displayed Gram matrices are the same as those from §4, except for (b8), where with and we use the basis . The stated discriminant groups follow from their Smith normal forms, and unimodularity of identifies with .
Proof of Theorem 1.1. Let be very general of one of the four types. By Proposition 6.1, satisfies (1.1) with . The computations in § 4.3 show that we have in each case.
Suppose that is rational. Proposition 1.2(3) then requires to be cyclic. Hence types (b8) and (b11) cannot be rational by Proposition 6.1. Moreover, for some . For types (c1) and (c6) this is:
respectively, and neither has an integer solution.
Applications to Verra fourfolds
We now apply Proposition 1.2 to recover Guéré’s result that the very general Verra fourfold is irrational [Gué26a, Corollary 28]. Recall that a Verra fourfold is a smooth double cover
branched over a divisor of bidegree . It has two associated quadric surface fibrations which we denote for .
Proposition 7.1. For a very general Verra fourfold one has
is primitive in and has . Furthermore .
Proof. is primitive and nondegenerate by definition as is nondegenerate. The rank and discriminant group assertions follow from the stated lattice identification on the Hodge structure of , which is provided in [IKKR17, Section 4].
Let be the degree two K3 surface associated with either quadric surface fibration . As rational Hodge structures we have
As a consequence of [32 Proposition 2.10(ii)], the associated degree two K3 surfaces vary dominantly in their moduli space. Hence is very general when is very general. For such an , the Hodge group of is ; see [21 Chapter 6, Corollary 4.7 and Theorem 4.9]. Thus .
Proof of Corollary 1.3. By Proposition 7.1 a very general Verra fourfold satisfies (1.1) with , and . If were rational, Proposition 1.2(2) would give .
Remark 7.2. Various works have compared Küchle fourfolds with Verra fourfolds. In particular, Benedetti–Fay–Guéré–Manivel–Perrin proved that Hodge-general Küchle fourfolds of type (c5) are not birational to any Verra fourfold [3 Corollary 5.3] and the author proved the same for very general Küchle fourfolds of type (c7) [15 Corollary 1.6]. Benedetti–Guéré–Manivel–Perrin studied the birational geometry of Verra fourfolds in [5]. For very general symmetric Verra fourfolds, equivariant irrationality with respect to the factor-swap involution was proved in [14 Theorem 6.4]. More recently, Jacovskis–Meinsma–Moschetti gave a construction of cubic fourfolds birational to Verra fourfolds in [25].
Code Availability
The code and data associated with the Quantum Lefschetz calculations in Section 4 can be found at https://github.com/aideenfay/kuchle-quantum-lefschetz. Our SageMath implementation of Atiyah–Bott equivariant localization for the line counts in Table 3 is available at https://github.com/aideenfay/grassmannian-lines.
| Küchle type | |||
| (b1) | 18 | ||
| (b5) | 7 | ||
| (b6) | 8 | ||
| (b7) | 5 | ||
| (b8) | 4 | ||
| (b11) | 3 | ||
| (c1) | 6 | ||
| (c2) | 8 | ||
| (c4) | 0 | ||
| (c5) | 4 | ||
| (c6) | 5 | ||
| (c7) | 0 | ||
| (d1) | 0 | ||
| (d3) | 5 |
Table 3. The numbers of lines through a general point of the fivefold defined by , containing the fourfold as a hyperplane section.
Acknowledgments
The author thanks Tom Coates for his assistance and many valuable discussions, and gratefully acknowledges EPSRC grant EP/Y028872/1. Parts of this work were assisted by GPT-5.5 and GPT-6 Astra in the form of finding gaps in proofs and checking the final paper for mistakes.
References
Line counts by Atiyah–Bott Equivariant Localization
Let be a fourfold from Küchle’s list [Küc95] defined by the bundle , and let be the fivefold cut out by . Thus is a hyperplane section of in the Plücker embedding . For a general point , every line through is parameterized by , and those contained in are the zeros of a section of the bundle with fibres
Since has rank , the number of lines through is
We compute this integral using Atiyah–Bott equivariant localization [1 Equation 3.8] and record the results in Table 3. See § 8 for information on code availability.
Department of Mathematics, Imperial College London, London SW7 2AZ, United Kingdom
Email address: [email protected]
References
- [1][AB84] M.F. Atiyah and R. Bott. The moment map and equivariant cohomology. Topology, 23(1):1–28, 1984.
- [2][ACG+26] Boris Alexeev, Leonardo F. Cavenaghi, Giovane Galindo, Bogdan Georgiev, Ludmil Katzarkov, and Pedro Antonio Muniz Martins. On the atomic decomposition of complete intersection in flag varieties, 2026.arxiv.org/abs/2609.21914
- [3][BFG+26] Vladimiro Benedetti, Aideen Fay, Jérémy Guéré, Laurent Manivel, and Nicolas Perrin. An atomic criterion for irrationality without quantum computations, 2026.arxiv.org/abs/2607.26718
- [4][BFMT26] Marcello Bernardara, Enrico Fatighenti, Laurent Manivel, and Fabio Tanturri. Fano fourfolds of K3 type, 2026.arxiv.org/abs/2111.13030
- [5][BGMP26] Vladimiro Benedetti, Jérémy Guéré, Laurent Manivel, and Nicolas Perrin. Quantum cohomology and birational geometry of Verra fourfolds, 2026.arxiv.org/abs/2605.30450
- [7][CG07] Tom Coates and Alexander B. Givental. Quantum Riemann–Roch, Lefschetz and Serre. Annals of Mathematics, 165:15–53, 2007.DOI
- [8][CGKS20] Tom Coates, Sergey Galkin, Alexander Kasprzyk, and Andrew Strangeway. Quantum periods for certain four-dimensional Fano manifolds. Experimental Mathematics, 29(2):183–221, 2020.
- [9][CJ99] Alberto Collino and Masao Jinzenji. On the structure of the small quantum cohomology rings of projective hypersurfaces. Communications in Mathematical Physics, 206(1):157–183, September 1999.
- [10][Coa14] Tom Coates. The Quantum Lefschetz principle for vector bundles as a map between Givental cones, 2014.arxiv.org/abs/1405.2893
- [11]Pierre Deligne. La conjecture de Weil. I. Publications Mathématiques de l’Institut des Hautes Études Scientifiques, 43(1):273–307, 1974.DOI
- [12]Pierre Deligne. La conjecture de Weil : II. Publications Mathématiques de l’IHÉS, 52:137–252, 1980.DOI
- [13]Pierre Deligne. La conjecture de Weil pour les surfaces K3. Inventiones mathematicae, 15:206–226, 1971/72.DOI
- [14]Aideen Fay. Equivariant irrationality of very general symmetric Verra fourfolds, 2026.arxiv.org/abs/2605.30439
- [15]Aideen Fay. Hasse obstructions to rationality for special fourfolds, 2026.arxiv.org/abs/2609.06759
- [16]William Fulton. Intersection Theory, volume 2 of Ergebnisse der Mathematik und ihrer Grenzgebiete. 3. Folge / A Series of Modern Surveys in Mathematics. Springer-Verlag, Berlin, Heidelberg, 1984.
- [17]V. Golyshev. Spectra and strains, 2008.arxiv.org/abs/0801.0432
- [20]Xiaowen Hu. Big quantum cohomology of Fano complete intersections, 2025.
- [21]Daniel Huybrechts. Lectures on K3 Surfaces. Cambridge Studies in Advanced Mathematics. Cambridge University Press, 2016.DOI
- [22]Daisuke Inoue, Atsushi Ito, and Makoto Miura. I-functions of Calabi–Yau 3-folds in Grassmannians. Communications in Number Theory and Physics, 11(2):273–309, 2017.DOI
- [24]Hiroshi Iritani. Quantum ring and Quantum Lefschetz, 2017.
- [25]Augustinas Jacovskis, Reinder Meinsma, and Riccardo Moschetti. The Verra involution on special cubic fourfolds, 2026.arxiv.org/abs/2610.32952
- [26]Nicholas M. Katz. Larsen’s alternative, moments, and the monodromy of Lefschetz pencils. In Contributions to automorphic forms, geometry, and number theory, pages 521–560. Johns Hopkins University Press, Baltimore, MD, 2004.
- [27]Ludmil Katzarkov, Maxim Kontsevich, Tony Pantev, and Tony Yue Yu. Birational invariants from Hodge structures and quantum multiplication, 2026.arxiv.org/abs/2508.05105
- [28]M. Kontsevich and Yu. Manin. Gromov–Witten classes, quantum cohomology, and enumerative geometry. Communications in Mathematical Physics, 164(3):525–562, August 1994.
- [30]A G Kuznetsov. On Küchle varieties with Picard number greater than 1. Izvestiya: Mathematics, 79(4):698–709, August 2015.DOI
- [31]Alexander Kuznetsov. Küchle fivefolds of type c5. Mathematische Zeitschrift, 284(3–4):1245–1278, 2016.DOI
- [32]Robert Laterveer. Algebraic cycles and Verra fourfolds. Tohoku Mathematical Journal, Second Series, 72(3):451–485, 2020.
- [33]James D. Lewis. The cylinder homomorphism associated to quintic fourfolds. Compositio Mathematica, 56:315–329, 1985.
- [34]James D. Lewis. The cylinder correspondence for hypersurfaces of degree n in ℙⁿ. American Journal of Mathematics, 110(1):77–114, 1988.DOI
- [35]Jun Li and Gang Tian. Virtual moduli cycles and Gromov–Witten invariants of algebraic varieties, 1998.arxiv.org/abs/alg-geom/9602007
- [36]Junyu Meng. Quantum cohomology of a Fano quiver moduli space, 2024.arxiv.org/abs/2412.15987
- [38]Ichiro Shimada. On the cylinder homomorphisms of Fano complete intersections. Journal of the Mathematical Society of Japan, 42(4):719 – 738, 1990.DOI