Virasoro constraints for complete intersections and projective-bundle towers

The explanation

Start with curves mapped to the simplest target, a single point. Fix a genus g, the number of handles of a surface, and n marked points on it. For each point pick a whole number k. These choices give an intersection number on the space of all such curves, written ⟨τ_k1 … τ_kn⟩_g. It can be nonzero only when the k's add up to 3g − 3 + n. Two standard values are ⟨τ_0 τ_0 τ_0⟩_0 = 1 and ⟨τ_1⟩_1 = 1/24.

These numbers obey linear rules. The dilaton rule says that adding a τ_1 multiplies the number by 2g − 2 + n. Here n counts the points before the new one. So ⟨τ_1 τ_0 τ_0 τ_0⟩_0 = (0 − 2 + 3) × 1 = 1. Likewise ⟨τ_1 τ_1⟩_1 = (2 − 2 + 1) × 1/24 = 1/24. The string rule says that adding a τ_0 lowers one other k by one, summed over the choices. So ⟨τ_0 τ_2⟩_1 = ⟨τ_1⟩_1 = 1/24. Witten's conjecture and Kontsevich's theorem fix the whole generating function of these numbersSource 1. Its equations include these two rules.

Gromov–Witten invariants make the same kind of count for curves inside a space X. Descendant invariants also record powers of a natural class at each marked point, as k does above. The Virasoro conjecture says their generating function Z obeys equations L_k Z = 0 for every k ≥ −1. The operators L_k depend on X through its pairing of classes, a Hodge grading and its first Chern classSource 2. The Hodge grading sorts the cohomology classes of X into types. The first Chern class is a basic curvature class of X.

Eguchi, Hori and Xiong proposed such equations for Fano varieties in 1997Source 3. Liu and Tian proved them in genus zeroSource 4. Givental's methods with Teleman's theorem give all genera when quantum cohomology is generically semisimpleSource 5. Semisimple roughly means the curve-count multiplication splits into independent one-dimensional pieces. Okounkov and Pandharipande proved the equations for smooth target curvesSource 6. Coates, Givental and Tseng passed them between a base and toric bundles built from sums of line bundlesSource 7.

The September paper claims the full conjecture for every smooth complete intersection in projective spaceSource 8. A complete intersection is the common zero set of the right number of polynomials, like a smooth cubic surface. The claim covers every genus and curve class. Primitive and odd classes are included, with no semisimplicity hypothesisSource 9. The October paper claims a transfer theorem. The equations pass from a base B to the projectivization of any vector bundle of rank at least twoSource 10. Projectivization replaces each point of B by all lines through the origin of its fiber. The bundle need not split, and it needs no positivity conditionSource 11. So the result applies again and again to towers of projective bundlesSource 12. With Okounkov and Pandharipande's theorem, it covers every projective bundle over a smooth curveSource 13.

The September paper breaks X into simpler pieces glued along a seam, a step called degeneration. Its main obstacle is primitive cohomology. Primitive classes need not extend one at a time through a degenerationSource 14. So the paper treats the error in each equation as a single object and probes it with allowed testsSource 15. Pairing an error with its complex conjugate, in a second copy of the theory, gives its squared sizeSource 16. That size is positive unless the error is zero, so the error must vanishSource 17.

How sure is this? There is no Lean scope note for either paper, so no computer proof checker has tested them. Each paper claims its theorem in full, using earlier published results as inputs. No specialist review has been published.

What it doesn't show. This does not prove the Virasoro conjecture for every smooth projective variety, and nothing here has a Lean check.