Family 054 — K3 structures without rationality for cubic fourfolds

← All digests

Family 054 · Algebraic and complex geometry
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-12

1. The question

A polynomial equation can define a space whose points nevertheless admit a rational parametrization. The simplest example is a conic with a chosen point: intersecting it with lines through that point supplies a parameter. In higher dimensions, deciding whether such a parametrization exists becomes much harder than solving the equation locally.

A cubic fourfold is the smooth four-dimensional complex variety cut out by one homogeneous cubic equation in projective five-space. Several of its most informative structures resemble those of a K3 surface. For decades this resemblance has guided proposals for recognizing rational cubic fourfolds.

The manuscript in family 054 claims that this guide has a fundamental limitation. On infinitely many prescribed families of cubics, a very general member is irrational even though its associated category is the ordinary derived category of a K3 surface. It also carries the expected integral K3 cohomological structure. The point is a separation between rational parametrization and powerful structures that had been proposed to characterize it.

2. Background and essential definitions

Rationality and genericity

A complex fourfold $X$ is rational if it is birational to $\mathbb P^4$: dense open subsets are isomorphic, or equivalently its rational-function field is $\mathbb C(t_1,t_2,t_3,t_4)$. A birational map may fail on a lower-dimensional subset; it need not extend everywhere.

A very general point of a parameter space lies outside a countable union of proper closed algebraic subsets. A theorem for very general members allows exceptional subfamilies, including rational examples. Moreover, being very general in a chosen divisor is different from being very general in the whole parameter space: the divisor itself might belong to the exceptional set of a theorem about the larger space.

K3 surfaces, lattices, and categories

A K3 surface is a compact simply connected complex surface with trivial canonical bundle. Its second integral cohomology is a lattice: a free abelian group equipped with an integer-valued intersection pairing. The Hodge decomposition further divides complexified cohomology according to holomorphic and antiholomorphic degree. An integral Hodge isometry preserves both structures.

For a cubic $X$, let $h$ be the hyperplane class. A special cubic has an additional integral Hodge class in degree four. A Hassett divisor $\mathcal C_d$ parametrizes cubics with a specified saturated rank-two lattice containing $h^2$, whose intersection matrix has determinant $d$. Saturated means the lattice has no missing integral points in its rational span. The positive integer $d$ is its discriminant.

The bounded derived category $D^b(\operatorname{Coh}X)$ packages coherent sheaves and complexes, retaining how they map to one another. After removing the three elementary pieces $\mathcal O_X,\mathcal O_X(1),\mathcal O_X(2)$, its remaining component is denoted $\operatorname{Ku}(X)$. Calling it geometric K3 means there is an exact equivalence

$$\operatorname{Ku}(X)\simeq D^b(\operatorname{Coh}S)$$

for an actual K3 surface $S$. This is an ordinary, untwisted category. Allowing additional twisting data would give a different condition.

3. Development history and previous partial results

Classical constructions of Fano and Tregub produced rational cubics containing suitable special surfaces. Beauville and Donagi connected the variety of lines on a cubic with holomorphic symplectic geometry; Pfaffian cubics supplied particularly striking rational examples related to K3 surfaces. Hassett subsequently organized special cubics through the divisors $\mathcal C_d$ and their associated K3 lattices. These constructions explain why K3 geometry became central to the rationality problem. The manuscript’s references include the original papers.

Kuznetsov proposed that rationality should be equivalent to having a geometric K3 component. Addington and Thomas related this categorical condition to Hassett’s Hodge-theoretic condition on a dense open subset of the appropriate divisors. Work on stability conditions by Bayer, Lahoz, Macrì, Stellari and their collaborators extended the categorical correspondence. Thus the K3 side of the new statement builds on substantial earlier theory; the manuscript’s proposed new ingredient is irrationality in that same range. Addington–Thomas.

Several rationality tests must be distinguished. Addington showed that having the variety of lines birational to the Hilbert square of a K3 surface is more restrictive than merely having an associated K3 surface. The Hilbert square parametrizes unordered pairs of points together with their limiting configurations. Addington’s comparison.

The manuscript also places itself after recent irrationality results for very general cubics in the full moduli space and after Fay’s September 2026 obstruction on special divisors. Fay’s obstruction uses a quaternion algebra; it is nonzero on some discriminants but vanishes throughout the admissible range considered here. Consequently that earlier obstruction does not supply the new theorem. Fay’s paper.

There remain rational cubics on special loci. Russo and Staglianò constructed rationality results for particular discriminants, while Yang and Yu found rational subloci inside every Hassett divisor. The manuscript therefore does not assert that a whole large-discriminant divisor consists of irrational cubics. The distinction between its very general points and its exceptional points is essential.

4. What the new manuscripts claim

Irrationality inside every sufficiently large admissible divisor

Call $d$ admissible if

$$d>6,\qquad d\equiv0\text{ or }2\pmod6,$$

and $d$ is divisible by neither $4$, nor $9$, nor any odd prime congruent to $2$ modulo $3$.

Theorem 1.1 of M1 claims that there is an integer $d_0$ such that, for every admissible $d>d_0$, a very general cubic $X\in\mathcal C_d$ is irrational and has a geometric K3 component. No explicit numerical value of $d_0$ is supplied. The statement is uniform over all sufficiently large admissible discriminants, with a separate very-general exceptional set on each divisor.

The associated polarized K3 surface also realizes the expected integral Hodge structure: the orthogonal complement of the cubic’s rank-two labelled lattice matches the primitive degree-two K3 lattice with the appropriate shift of Hodge degrees and reversal of the pairing. Thus replacing the categorical condition by this integral cohomological condition does not restore sufficiency for rationality.

Stronger K3 resemblance and zero-cycles

For every integer $n\ge2$, set

$$d=2n^2+2n+2.$$

These discriminants are admissible and unbounded. When $d>d_0$, the manuscript combines its irrationality theorem with Addington’s result to obtain very general irrational cubics whose variety of lines $F(X)$ is birational to $S^{[2]}$, the Hilbert square of a projective K3 surface. The conclusion is birationality, not an isomorphism; the different K3 surfaces appearing in these descriptions need not be identified with each other.

A further consequence concerns zero-cycles, which are formal integer sums of points modulo relations coming from rational functions on curves. For these irrational cubics, the degree map is an isomorphism after every field extension $E/\mathbb C$:

$$\deg:\operatorname{CH}_0(X_E)\longrightarrow\mathbb Z.$$

This is universal integral triviality of zero-cycles. The source obtains it from Voisin’s earlier theorem, since admissibility includes $4\nmid d$. It is not a new zero-cycle calculation. It means these irrational examples pass a major necessary test for stable rationality. It does not decide whether they are stably rational.

5. Why the advance matters

Importance: 5/5. If correct, the claim would overturn central sufficiency directions in the K3-based picture of cubic-fourfold rationality, across infinitely many specifically identified divisors. It would also demonstrate that several mutually reinforcing indicators can all coexist with irrationality.

The significance lies in the precision of the separation. The cubics have an ordinary geometric K3 category, an integral K3 Hodge description, and universally trivial zero-cycles. An infinite subcollection even has its variety of lines birational to a K3 Hilbert square. These are substantial structures, not merely superficial numerical similarities.

At the same time, the theorem would not make the associated K3 geometry irrelevant. It remains a rich description of the cubic and its moduli. What changes is the inference one may draw from that description: possessing it would no longer guarantee a rational parametrization.

6. The main idea

Count the same exceptional geometry in two ways

The proposed argument assumes a birational map from $\mathbb P^4$ to $X$ and associates an integer to it. Resolving such a map introduces and removes exceptional divisors. The integer counts these with signs, but only when their underlying geometry contains a K3 surface from a carefully chosen collection.

More precisely, one removes the rationally connected directions from a divisor. The resulting maximal rationally connected quotient records the part that cannot be swept out by rational curves. The test is one if that quotient is birational to one of the selected K3 surfaces, and zero otherwise. Adding the tests of divisors acquired at the target and subtracting those lost at the source defines $c_u(f)$.

The selected K3 surfaces have Picard number one, a polarization of degree $d$, and a specified entire integral transcendental lattice. The transcendental lattice is the part of cohomology orthogonal to algebraic classes. Retaining the whole integral lattice is stronger than retaining a rational subspace or only its dimension.

Smooth factorization gives one

Weak factorization expresses a birational map between smooth projective varieties as blowups and blowdowns along smooth centers. In dimension four, surface centers can contribute K3-type cohomology.

The manuscript uses operators constructed from genus-zero curve-counting invariants to distinguish the contributions of whole transcendental lattices. Its proposed calculation says that, under the very-general lattice hypotheses, one selected K3 contribution survives. In its notation,

$$c_u(f)=1.$$

This step does not assume that every surface’s transcendental lattice is indecomposable. Such a general indecomposability proposal has known counterexamples; the source instead introduces additional operators to control the blocks relevant to this particular calculation.

Mori factorization gives zero

The Sarkisov program decomposes the same birational map into links between Mori fiber spaces, the fibered building blocks of birational geometry. Here the manuscript introduces correction terms attached to each intermediate space. A link’s contribution becomes the difference of the corrections at its two ends.

Summing these differences cancels every intermediate term. For sufficiently large $d$, the source’s boundedness arguments make this calculation uniform, independent of the number of links or the chosen birational map. Both endpoint corrections vanish because the endpoints have point bases. Therefore the second calculation gives

$$c_u(f)=0.$$

The contradiction supplies the irrationality criterion. Period and lattice results then put very general members of each sufficiently large admissible divisor within its hypotheses. This is the strategy reported by the manuscript, rather than an independent verification of its factorization calculations.

7. Scope, limitations, and open questions

The field is $\mathbb C$, the cubic is smooth, and the theorem requires both admissibility and an unspecified lower bound on discriminant. It does not determine rationality at every small discriminant or for every member of a covered divisor. The known rational subloci are consistent with its very-general formulation.

The categorical and cohomological associations are supplied by earlier results. The principal claimed advance is the birational obstruction that operates despite them. Likewise, universal zero-cycle triviality comes from Voisin’s theorem and should not be read as evidence that the manuscript proves stable rationality.

An effective threshold, explicit individual irrational equations in the stated range, and a satisfactory replacement for the discarded sufficiency criteria are natural further questions. Another is how much rationality information can survive in structures finer than the category or integral Hodge lattice used here.

Finally, the ordinary irrationality conclusion leaves stable rationality open: multiplying by projective space is a separate operation, and the manuscript does not establish that all such products remain irrational.

8. Sources and further reading

M1, Theorem 1.1 and Corollaries 1.2–1.4, contains the principal statements. Its Sections 2–6 develop the two factorization calculations; Section 7 applies them to Hassett divisors and derives the consequences from existing K3 and zero-cycle theory.

For the earlier structures, see Addington–Thomas on Hodge theory and derived categories, Addington on the two K3 rationality conditions, and Voisin on universal zero-cycles of cubic hypersurfaces. Fay’s 2026 obstruction explains the adjacent discriminant range already treated by a different method. M1’s bibliography supplies the classical rational constructions and the factorization results used in its argument.

Source manuscripts

• M1. OpenAI, Irrational cubic fourfolds with geometric K3 categories (September 24, 2026).

Sources consulted on 12 October 2026 at repository revision fd4aeeb. Download the manuscript citations.