Virasoro constraints for projective bundles and complete intersections

← All digests

Family 065 · Algebraic and complex geometry
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-12

1. The question

Counting curves on a complex algebraic variety produces far more than one sequence of numbers. Curves have different genera and degrees; they can be required to meet chosen cycles; and each marked point carries further geometric data. Gromov–Witten theory packages all these counts into one enormous generating function.

The Virasoro conjecture predicts universal differential equations for that function. Their form depends on familiar cohomological data of the target, but their content constrains curve counts in every genus. Why should the same algebra of equations govern such different geometric spaces?

This family approaches the question through two constructions. M1 claims that the equations pass from a smooth projective base to the projectivization of any vector bundle on it. M2 claims them for every smooth complete intersection in projective space. Both include cohomology classes that the simplest reconstruction methods cannot see.

2. Background and essential definitions

Curve counts with insertions

Let $Y$ be a smooth complex projective variety. A stable map consists of a possibly nodal marked curve $C$, together with an algebraic map $C\to Y$, with only finitely many automorphisms preserving the data. Its genus is $g$; its curve class $\beta\in H_2(Y,\mathbb Z)$ records where it lies homologically.

At a marked point, a cohomology class $\gamma$ imposes an incidence condition. A power of the cotangent-line class $\psi$ at that point gives a descendant insertion, written $\tau_a(\gamma)$. The corresponding invariant is a virtual intersection number on the moduli space of stable maps. “Virtual” matters: these are systematically defined invariants even when curves do not form a finite collection that can be literally counted.

Gather the connected genus-$g$ invariants into a formal series $F_g^Y$, with variables for the insertions and symbols $Q^\beta$ for curve classes. The total potential is

$$Z_Y=\exp\!\left(\sum_{g\geq0}\hbar^{g-1}F_g^Y\right).$$

Here $\hbar$ keeps track of genus. No numerical convergence of this entire series is required; equations are read coefficient by coefficient.

What the Virasoro equations say

There are prescribed differential operators $L_k^Y$, for integers $k\geq-1$, acting on the insertion variables. The conjecture asserts

$$L_k^Y Z_Y=0\qquad(k\geq-1).$$

The first equations include familiar rules for adding an unconstrained marked point and respecting degrees. Higher operators combine terms with one derivative and two derivatives, reflecting both modifications of a curve and its splitting into pieces. Their commutators form a version of the Virasoro algebra.

The operators use the cohomological intersection pairing, multiplication by the first Chern class of the tangent bundle, and Hodge degrees. For a class of type $(p,q)$, the grading here uses $p-\dim(Y)/2$. Replacing this with half the total degree would lose information when $p\neq q$. Odd-degree classes also require the usual sign when two odd variables are exchanged.

The two geometric constructions

For a rank-$r$ vector bundle $E\to B$, the projective bundle $\mathbb P_B(E)$ has fiber $\mathbb P^{r-1}$. M1 uses the convention that its points are lines in the fibers of $E$. The bundle need not decompose globally into line bundles.

A smooth complete intersection is the smooth common zero set of homogeneous equations in projective space, with the expected codimension. Its ambient cohomology comes from projective space. The remaining middle-dimensional classes are called primitive. These extra classes are often precisely where the geometry becomes interesting—and where a result restricted to ambient insertions would be incomplete.

3. Development history and previous partial results

The point target already has a rich theory: its stable-map spaces are moduli spaces of curves. Witten predicted the governing integrable equations, and Kontsevich proved the corresponding intersection-theoretic theorem. Eguchi, Hori and Xiong proposed the extension to general targets; subsequent formulations incorporated the full Hodge grading and odd classes. Getzler’s treatment explains this general framework.

Several major regions were already understood. Liu and Tian established genus-zero constraints. Givental’s quantization formalism, together with Teleman’s classification of semisimple theories, supplies all-genus results when quantum cohomology has a sufficiently simple algebraic decomposition. Semisimplicity means that, generically, the relevant algebra splits into independent field factors; many targets of interest do not satisfy it. Okounkov and Pandharipande proved the constraints for target curves, including their odd cohomology.

For bundles, Coates, Givental and Tseng proved a base–total-space equivalence for toric bundles constructed from sums of line bundles. This already covers split projective bundles. Fan’s reconstruction theorem treats arbitrary projective bundles at the level of Gromov–Witten invariants. Iritani and Koto developed the quantum differential equations of nonsplit projective bundles and their relationship with the base. Reconstruction of invariants, however, does not automatically establish compatibility with every Virasoro operator. M1 addresses that additional issue outside the semisimple setting.

Complete intersections also had substantial prior coverage. Getzler’s theorem already gives all-genus constraints for the Calabi–Yau case with vanishing first cohomology; a smooth quintic threefold should therefore not be advertised as a newly settled example here. More recently, Guo, Zhang and Zhou proved genus-one constraints for ambient insertions on smooth Fano complete intersections. Argüz, Bousseau, Pandharipande and Zvonkine developed reconstruction of complete-intersection invariants through nodal invariants. M2 aims at the full constraints across all genera, with unrestricted primitive insertions, including the nonsemisimple cases left by earlier methods.

4. What the new manuscripts claim

M1: arbitrary projectivization preserves the constraints

Write $\mathcal V(Y)$ for the full ordinary Virasoro constraints. M1 states that for every smooth connected projective complex variety $B$ and every algebraic vector bundle $E$ of rank at least two,

$$\mathcal V(B)\ \Longrightarrow\ \mathcal V(\mathbb P_B(E)).$$

There is no positivity, splitting, or semisimplicity assumption on $E$ or the base theory. The conclusion concerns every genus, each actual integral curve class, and arbitrary cohomology insertions. Remembering the complete class is stronger than aggregating all classes with the same degree against one divisor.

Repeated application gives projective-bundle towers over any base already satisfying the conjecture. Curves supply one unconditional starting class through the earlier target-curve theorem. M1 states the forward implication; the stronger equivalence from the earlier split toric setting should not be silently transferred to arbitrary bundles.

M2: every smooth projective complete intersection

M2 states $\mathcal V(X)$ for every smooth complete intersection $X\subset\mathbb P^N$, without restricting its dimension, multidegree, or canonical positivity. The theorem includes Fano, Calabi–Yau and general-type examples, and all primitive and odd insertions.

Both papers concern the ordinary, unreduced Gromov–Witten potential and ordinary descendant classes at marked points. Modified theories and ancestor classes—related classes pulled back from moduli spaces of curves—enter the machinery but are not substitutes for the stated conclusion.

Combining the two claims also gives the constraints for arbitrary projective-bundle towers over smooth complete intersections. M2’s own induction uses the previously established split-bundle theorem; it does not need M1’s arbitrary-bundle extension as an input.

5. Why the advance matters

Importance: 5/5. These are broad claims about a central organizing principle in enumerative geometry. Complete intersections supply a basic source of varieties with complicated primitive cohomology, while projectivization is among the most common ways to build new varieties. An all-genus theorem stable under these constructions would enlarge the conjecture’s reach substantially.

The significance is structural. The equations relate different curve counts, and their survival under geometry would make them usable within larger reconstruction arguments. They do not, by themselves, provide a short closed formula for every invariant or imply that every target has semisimple quantum cohomology.

This rating reflects the scope of the claims and their methods, conditional on correctness. It is not an assessment of proof reliability.

6. The main idea

A master space connects the base and bundle

M1 places the two targets inside

$$W=\mathbb P_B(E\oplus\mathcal O_B).$$

Scaling the trivial summand gives a circle-type algebraic action with fixed pieces $B$ and $\mathbb P_B(E)$. Localization expresses invariants of the larger space through these pieces, together with contributions from curves moving between them. Those connecting curves occur in families and must retain their full correspondence data.

A shift operator packages the action into a system of differential equations. Its leading spectral branches satisfy an equation of the form

$$h^r(h+\lambda)=y,$$

where $\lambda$ records the action and $y$ records fiber degree. Near $y=0$, one branch corresponds to the base and the other cluster to the projective bundle. Analytic continuation around branch points permutes the branches. The proposed mechanism transports the base’s identities through this connected system and collects the bundle branches at the end.

This is applied to fixed coefficients, where suitable ancestor expressions are polynomial in the fiber variable. It is not an assumption that the entire all-genus potential converges. Finally, quantization introduces possible scalar errors; the Virasoro commutators are used to recover the prescribed normalization and remove them.

Degeneration handles complete intersections

M2 splits a defining equation into lower-degree factors. The smooth target degenerates into two simpler pieces meeting along a smooth common divisor; a blowup accounts for the intersection geometry. Induction lowers either dimension or the complexity of the defining degrees.

The difficulty is the primitive cohomology. Such classes need not extend coherently across the degeneration. Merely proving equations for ambient classes would leave a large part of the theorem untouched.

The manuscript instead treats each possible failure of a Virasoro equation as a multilinear tensor. It tests that tensor using compatible classes, inverse intersection pairings, and additional copies of curve-counting data. Chains of ruled spaces allow cutting and gluing operations to be compared. Since a Virasoro operator has at most two derivatives, a sufficiently long chain has an unused position at which a cancellation can be arranged.

These cancellations reduce the tested equations to smaller targets. A final Hodge-theoretic pairing turns the remaining primitive error into a nonnegative squared norm; its vanishing is the proposed means of detecting all primitive components. Extra divisor labels are retained where a single projective degree would fail to distinguish curve classes. The strategy therefore separates transporting equations from recovering information hidden by degeneration.

7. Scope, limitations, and open questions

The targets are smooth and projective over $\mathbb C$. The statements here do not automatically extend to arbitrary singular targets, noncompact varieties, orbifolds, or reduced curve-counting theories.

M1 starts with a base already satisfying the full conjecture. It is a transfer theorem, not a proof for every base. M2 supplies a large additional class of bases, but complete intersections and their bundle towers still do not exhaust smooth projective geometry.

“All genera” and “all insertions” are essential parts of the assertions. They include primitive classes and odd signs, not merely the even ambient sector. Equally, the potential is formal: coefficientwise identities do not assert an analytic sum over all curves and genera.

A natural further direction is to understand which other surgeries preserve the full constraints without semisimplicity, and how broadly the primitive-class detection mechanism can be applied. The manuscripts report specific transport arguments for their stated constructions; a universal surgery principle would require more.

8. Sources and further reading

For the foundational formulation, see Getzler’s Virasoro conjecture paper. For the prior all-genus bundle theorem, see Coates–Givental–Tseng; for the nonsplit quantum-cohomology framework, see Iritani–Koto. The recent Guo–Zhang–Zhou paper helps locate the genus-one ambient frontier. These works provide context for the two source manuscripts below.

Source manuscripts

• M1. OpenAI, Virasoro Constraints under Projectivization (October 5, 2026).
• M2. OpenAI, Virasoro Constraints for Projective Complete Intersections (September 24, 2026).

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