On the irrationality of cubic fourfolds

Abstract: Following the work of Katzarkov--Kontsevich--Pantev--Yu concerning the irrationality of the very general complex cubic fourfold, we prove the following: for every rational smooth complex cubic fourfold, the primitive cohomology is isomorphic as a Hodge structure to the (twisted) middle cohomology of a projective K3 surface.

• The paper strengthens the KKPY invariant by adding the rank of evaluated quantum multiplication, proving birational stability under blow-ups through Iritani’s quantum cohomology formula.
• Every smooth cubic fourfold fails the refined property, while projective space satisfies it, forcing any rationality factorization to involve a surface with K3-type minimal model.
• The paper proves that rationality would imply an isomorphism of Hodge structures between a cubic fourfold’s primitive fourth cohomology and the twisted second cohomology of a projective K3 surface, without proving any new individual irrationality result.

This paper, "On the irrationality of cubic fourfolds" by Jérémy Guéré, refines and extends the framework of Katzarkov–Kontsevich–Pantev–Yu (KKPY) (Katzarkov et al., 7 Aug 2025) for detecting irrationality via Gromov–Witten theory. Its main theorem resolves a Hodge-theoretic consequence of Kuznetsov's conjecture: if a smooth complex cubic fourfold $X$ is rational, then its primitive cohomology is isomorphic as a Hodge structure to the twisted middle cohomology of a projective K3 surface, $$H^4(X,\mathbb{Q})_{\mathrm{primitive}} \simeq H^2(S,\mathbb{Q})(-1)$$ (2603.04518).

Background and strategy

KKPY introduced a birational invariant built from genus-zero Gromov–Witten invariants combined with Hodge theory. Although GW invariants are deformation invariant, their interplay with the Hodge filtration yields quantities stable under blow-ups at smooth centers in dimensions up to four, by Iritani's quantum cohomology blow-up theorem (Iritani, 2023). Guéré's contribution is to replace the KKPY invariant (denoted $\clubsuit$) by a sharper one (denoted $\diamondsuit$), which tracks the rank $\gamma$ of the evaluated quantum multiplication operator on generalized eigenspaces, in addition to the invariants $\nu$, $\nu'$, and $\rho$ measuring intersections with Hochschild-degree pieces and Hodge classes.

The key object is the endomorphism $\kappa_\tau = Eu_\tau \star_\tau$ on cohomology, where $\star_\tau$ is the quantum product extended by the divisor equation to a formal parameter $$H^4(X,\mathbb{Q})_{\mathrm{primitive}} \simeq H^2(S,\mathbb{Q})(-1)$$0 in the Hodge classes of degree two. A central structural fact is that $$H^4(X,\mathbb{Q})_{\mathrm{primitive}} \simeq H^2(S,\mathbb{Q})(-1)$$1 raises degree by two and is a morphism of Hodge structures $$H^4(X,\mathbb{Q})_{\mathrm{primitive}} \simeq H^2(S,\mathbb{Q})(-1)$$2, hence preserves Hodge classes and Hochschild degree. Evaluation maps send formal variables (Novikov variables, divisor parameters) into a non-Archimedean graded field $$H^4(X,\mathbb{Q})_{\mathrm{primitive}} \simeq H^2(S,\mathbb{Q})(-1)$$3, the rational Levi-Civita field, where convergence of formal series is controlled by the valuation; a lemma shows that a series converges in this field exactly when its terms tend to zero, and evaluation maps are precisely the degree-zero homogeneous morphisms of graded algebras.

Birational invariance

The technical core is the behavior under blow-ups $$H^4(X,\mathbb{Q})_{\mathrm{primitive}} \simeq H^2(S,\mathbb{Q})(-1)$$4. Working over the ring $$H^4(X,\mathbb{Q})_{\mathrm{primitive}} \simeq H^2(S,\mathbb{Q})(-1)$$5 of Iritani, with $$H^4(X,\mathbb{Q})_{\mathrm{primitive}} \simeq H^2(S,\mathbb{Q})(-1)$$6 for codimension $$H^4(X,\mathbb{Q})_{\mathrm{primitive}} \simeq H^2(S,\mathbb{Q})(-1)$$7, the paper proves that the evaluated endomorphisms on $$H^4(X,\mathbb{Q})_{\mathrm{primitive}} \simeq H^2(S,\mathbb{Q})(-1)$$8, $$H^4(X,\mathbb{Q})_{\mathrm{primitive}} \simeq H^2(S,\mathbb{Q})(-1)$$9, and $\clubsuit$0 are simultaneously conjugate, with conjugating isomorphisms lying in the span (over a suitable number field) of morphisms of Hodge structures. Consequently, spectra and the invariants $\clubsuit$1 decompose additively along a blow-up:

$\clubsuit$2

A subtle point in the proof concerns the field of definition of Iritani's Fourier transforms. The continuous Fourier transform requires adjoining $\clubsuit$3-th roots of unity, $\clubsuit$4, $\clubsuit$5, and — when the center has complex dimension at least three — the transcendental number $\clubsuit$6. The paper gives a careful degree argument showing that only finitely many powers of $\clubsuit$7 occur, and that $\clubsuit$8 is unnecessary when $\clubsuit$9; Iritani is quoted as conjecturing that $\diamondsuit$0 is never needed. This field bookkeeping is what allows the invariants to propagate along weak factorizations (Włodarczyk), yielding: if all blow-up centers in a weak factorization between $\diamondsuit$1 and $\diamondsuit$2 satisfy Property $\diamondsuit$3, then $\diamondsuit$4 satisfies it if and only if $\diamondsuit$5 does.

The invariant $\diamondsuit$6 on surfaces and the cubic fourfold

For a surface $\diamondsuit$7 with $\diamondsuit$8, an explicit computation of the matrix of $\diamondsuit$9 in blocks by cohomological degree shows it is nilpotent; for $\gamma$0, $\gamma$1, $\gamma$2 (i.e., a K3 or elliptic K3 surface), the evaluated matrix has rank at most one, $\gamma$3, $\gamma$4, and $\gamma$5, so such surfaces fail Property $\gamma$6. This is the decisive feature: surfaces failing $\gamma$7 are exactly those whose minimal model is a K3 (or elliptic K3) surface.

For a cubic fourfold $\gamma$8, Givental's computation gives the ambient-part matrix of $\gamma$9; over $\nu$0 its eigenvalues are $\nu$1 (multiplicity 2) and $\nu$2, $\nu$3, while it vanishes on the primitive part. The eigenvalue $\nu$4 then has $\nu$5, $\nu$6, $\nu$7, so every cubic fourfold fails Property $\nu$8. Since $\nu$9 satisfies $\nu'$0, any weak factorization from a rational cubic fourfold to $\nu'$1 must contain a center failing $\nu'$2 — necessarily a surface with K3 minimal model. The commutation relation $\nu'$3, combined with a dimension count and a non-Archimedean argument extracting a $\nu'$4-linear combination of Hodge morphisms with invertible matrix, upgrades the induced embedding to an isomorphism of Hodge structures $\nu'$5, with $\nu'$6 projective (its $\nu'$7 follows from $\nu'$8).

The implication is substantial: combined with the known Hodge-theoretic criterion, rationality of $\nu'$9 would force its primitive cohomology to carry a K3-type Hodge structure — precisely the conclusion of Kuznetsov's conjecture at the level of Hodge structures, and consistent with the KKPY theorem that the very general cubic fourfold is irrational.

Limitations and open questions

The result is a conditional statement about rational cubic fourfolds; no individual rational cubic fourfold is known, so the theorem constrains a hypothetical class. The paper does not prove irrationality of any specific cubic fourfold beyond what KKPY established for the very general member. The reliance on the $\rho$0-free version of Iritani's blow-up formula for centers of dimension at most two is essential to the argument, and the general case ($\rho$1, requiring $\rho$2) is handled only via the span-of-Hodge-morphisms formalism; whether the blow-up theorem holds over a number field in full generality, as Iritani expects, remains open. Whether the Hodge-structure isomorphism can be strengthened to a categorical statement (Kuznetsov's full conjecture via derived categories) is outside the scope of the paper's GW-theoretic methods.

Conclusion

The paper strengthens the KKPY birational invariant from $\rho$3 to $\rho$4 by incorporating the rank of evaluated quantum multiplication on generalized eigenspaces, proves its invariance under blow-ups at smooth centers via a detailed field-of-definition analysis of Iritani's blow-up formula, and derives that any rational smooth complex cubic fourfold has K3-type primitive cohomology. The argument is a concrete instance of Gromov–Witten theory detecting birational geometry through Hodge-theoretic obstructions, and it reduces Kuznetsov's conjecture for cubic fourfolds to a purely Hodge-theoretic statement.