The GW/PT conjectures for toric pairs

This paper proves the conjectural correspondence between logarithmic Gromov-Witten and Donaldson/Pandharipande-Thomas theories for toric threefold pairs with singular boundaries, while also establishing new results such as the Laurent polynomial nature of capped vertices and verifying the logarithmic DT/PT conjecture.

The Gist

Imagine you are trying to count the number of ways a rubber band can stretch and twist around a complex, multi-faceted sculpture. In the world of mathematics, this is a problem of counting curves.

This paper by Davesh Maulik and Dhruv Ranganathan solves a massive, decades-old puzzle about how to count these curves in a specific, highly structured environment called a Toric Threefold.

Here is the breakdown of their achievement using everyday analogies.

1. The Two Different Languages (GW and PT)

Mathematicians have two main ways to count these rubber bands (curves):

• GW (Gromov-Witten): Think of this as counting the paths the rubber band takes. It focuses on the shape of the curve itself.
• PT (Pandharipande-Thomas): Think of this as counting the stuff the rubber band is made of (like a sheaf or a collection of points). It focuses on the "fuzz" or the material distribution.

For a long time, mathematicians suspected that these two methods, though they look completely different, actually give the exact same answer if you translate the numbers correctly. This is the GW/PT Correspondence. It's like having two different recipes for a cake; one uses cups of flour, the other uses grams. If the correspondence is true, they will always result in the same delicious cake.

2. The Problem: The "Smooth" vs. The "Rough"

For 20 years, mathematicians could prove this "recipe equivalence" only when the sculpture (the space) was perfectly smooth, like a polished marble statue.

However, real-world shapes (and many mathematical ones) have corners, edges, and singularities. They are "rough."

• The Old Problem: Previous proofs broke down when the boundary of the space was rough or "singular" (like a crumpled piece of paper or a star-shaped object with sharp points).
• The New Breakthrough: This paper proves that the two recipes still match, even when the sculpture is rough, crumpled, and has sharp corners. They call this the "fully logarithmic" setting.

3. The Strategy: Breaking the Monster into LEGO Bricks

How did they solve this? They didn't try to count the curves on the whole rough sculpture at once. Instead, they used a strategy called Degeneration.

Imagine you have a giant, complex LEGO castle. It's too hard to count the ways a string can wrap around the whole thing.

1. Take it apart: They "degenerate" the castle. They imagine the castle slowly falling apart into smaller, simpler LEGO blocks (elementary geometries).
2. The "Rubber" Trick: When the castle falls apart, the string doesn't just snap; it stretches across the gaps. The authors developed a new way to handle these stretched strings (called "rubber calculus") so they can count the string on the small blocks and then glue the answers back together.
3. The Induction: They proved that if the recipe works for the tiny, simple blocks (like a single cube or a flat sheet), it must work for the giant castle.

4. The "Star" Map

To keep track of all these broken pieces, the authors used a map called a Tropical Geometry map.

• Imagine the sculpture is a city.
• The "curves" are delivery trucks driving through the city.
• The "Tropical Map" is a simplified, stick-figure version of the city where the roads are just lines and the buildings are just points.
• The authors proved that if you can count the trucks on the stick-figure map (the "stars"), you can count them on the real city. They showed that the "stars" (the simplest building blocks) follow the rules perfectly.

5. Why This Matters (The "Laurent Polynomial" Surprise)

One of the most exciting side-results of this paper is about the complexity of the answer.

• Before this, we didn't know if the answer to these counting problems would be a simple, finite formula or an infinite, messy series that goes on forever.
• The authors proved that under certain "positive" conditions (like the sculpture being built in a way that doesn't trap the rubber bands), the answer is actually a Laurent Polynomial.
• Analogy: It's like discovering that a complex financial investment, which you thought required an infinite spreadsheet to calculate, actually has a simple, neat algebraic formula. This makes the math much more predictable and usable.

6. The "Capped Vertex"

They also solved a specific 2008 conjecture about the "capped vertex."

• Imagine a corner of a room where three walls meet.
• The "capped vertex" is a specific mathematical object related to that corner.
• The paper proves that the counting formula for this corner is also a neat, finite polynomial. This was a major open problem in the field.

Summary

In short, Maulik and Ranganathan took a theory that only worked for smooth, perfect shapes and extended it to rough, crumpled, and complex shapes. They did this by:

1. Breaking the complex shape into simple LEGO-like blocks.
2. Creating a new "glue" (rubber calculus) to handle the rough edges.
3. Proving that the two different counting methods (paths vs. material) always agree, even in the messiest environments.

This opens the door to solving many other counting problems in geometry that were previously considered too "rough" to handle.

Technical Summary

Here is a detailed technical summary of the paper "The GW/PT Conjectures for Toric Pairs" by Davesh Maulik and Dhruv Ranganathan.

1. Problem Statement

The paper addresses the Gromov–Witten (GW) / Pandharipande–Thomas (PT) correspondence for toric threefold pairs $(Y|\partial Y)$.

• Context: GW theory counts stable maps to a target $Y$, while PT theory counts stable pairs (sheaves) on $Y$. The correspondence conjectures that their generating functions are related by a change of variables ($-q = e^{iu}$) and that the PT series is the Laurent expansion of a rational function.
• The Gap: Previous work established this correspondence for:
  1. Absolute targets ($\partial Y = \emptyset$).
  2. Pairs where the boundary divisor $\partial Y$ is smooth.
• The Challenge: This paper tackles the case where $\partial Y$ is a singular torus-invariant divisor (specifically, a simple normal crossings or snc divisor). This is the "fully logarithmic" setting. The singularity of $\partial Y$ introduces complex combinatorial structures (logarithmic geometry) and "exotic" insertions that were previously unmanageable.

2. Methodology

The authors employ a sophisticated inductive strategy based on logarithmic degeneration formulas and tropical geometry. The core methodology involves:

A. Logarithmic Degeneration and Elementary Geometries

• Degeneration Formula: They utilize the recently proved logarithmic degeneration formula (from Maulik, Ranganathan, et al.) which relates invariants of a general fiber to invariants of a special fiber (a union of simpler components).
• Reduction to Elementary Geometries: They prove that any toric threefold pair can be degenerated into a union of "elementary geometries." These are specific toric pairs defined by their tropicalizations (cone complexes), including:
  • Full toric boundary ($\mathbb{R}^3$).
  • $\mathbb{P}^1$-bundles over toric surfaces (1-non-boundary).
  • $\mathbb{P}^2$-bundles over $\mathbb{P}^1$ (2-non-boundary).
  • Products of projective spaces (3-non-boundary).

B. The "Star" Induction and Partial Ordering

• Stars: They define a "star" as a combinatorial object (a point in the cone complex with outgoing rays and multiplicities) that encodes the curve class and tangency conditions.
• Partial Ordering: A partial order is defined on stars based on degenerations. A star $w$ is "smaller" than $v$ if $w$ arises as a vertex in a degeneration of $v$.
• Induction: The proof proceeds by induction on this partial order. The goal is to show that if the correspondence holds for all "smaller" stars, it holds for the current star.

C. Handling Exotic Insertions and Rubber Calculus

• Exotic vs. Non-Exotic: In the logarithmic setting, degenerations introduce "exotic" insertions (classes from the cohomology of the stack of expansions) that do not appear in the smooth case.
• Rubber Calculus: The authors develop a "rubber calculus" technique to convert invariants with exotic insertions into invariants with non-exotic (standard) insertions. This involves:
  • Rigidifying the moduli spaces (removing automorphisms).
  • Using the degeneration formula to trade boundary insertions for deeper strata.
  • Proving that the correspondence is preserved under these conversions.

D. Equivariant PT Moduli Spaces

• To match the GW and PT sides precisely, they modify the PT moduli spaces using Negut's work on flag Hilbert schemes and Oberdieck's marked PT spaces. This ensures that the evaluation maps on the PT side align perfectly with the GW side, allowing for a direct comparison of invariants.

3. Key Contributions and Results

Main Theorems

• Theorem A (Main Correspondence): The logarithmic GW/PT correspondence holds equivariantly for all toric threefold pairs $(Y|\partial Y)$, regardless of whether $\partial Y$ is smooth or singular. This is the first verification of the conjecture in the fully logarithmic setting.
• Theorem B (Polynomiality): If every toric divisor not in $\partial Y$ is nef (numerically effective), then the equivariant PT series is a Laurent polynomial in $q$ (rather than just a rational function).
  • Significance: This proves a 2008 conjecture by Oblomkov, Okounkov, Pandharipande, and Maulik regarding the "capped vertex" (a specific local invariant).
• Theorem C (DT/PT Correspondence): The logarithmic Donaldson–Thomas (DT) / PT correspondence holds equivariantly for toric pairs. This follows by adapting the GW/PT proof.

Specific Technical Breakthroughs

1. Strata Conjectures: They formulate and prove a generalized correspondence for "strata invariants" (invariants associated with specific loci in the moduli space), which is necessary for the inductive step.
2. Visibility and Rigidity: They introduce a combinatorial condition called "visibility" for stars and Chow 1-complexes. This ensures that the isotropy groups of the moduli spaces are rigid enough to allow for well-defined inductive arguments.
3. Tropical Combinatorics: They utilize Parker's tropical arguments (adapted from Bousseau) to handle the "semi-positive" (nef) cases, showing that consistency relations between different degeneration paths constrain the series to be rational/polynomials.
4. Negative Cases: For non-nef geometries (where standard positivity arguments fail), they reduce the problem to local curve theories (specifically $Tot(\mathcal{O}(-m) \oplus \mathcal{O})$), where the correspondence is already known, using a novel induction on the degree and equivariant parameters.

4. Significance and Impact

• Resolution of the "Fully Logarithmic" Case: This paper closes a major gap in enumerative geometry by handling singular boundaries, which are ubiquitous in degeneration techniques used to study absolute invariants (e.g., for the quintic threefold).
• Foundation for Descendent Conjectures: The authors state that this result is the necessary foundation for proving the GW/PT descendent conjectures (involving $\psi$-classes) for Fano threefolds and beyond. Previous methods failed in the logarithmic setting due to the complexity of singular boundaries.
• New Proof of Known Results: Even for the case where $\partial Y = \emptyset$ (absolute toric threefolds), their methods provide a new proof of the correspondence that is distinct from previous approaches (like those by Okounkov-Pandharipande) and yields stronger results (Laurent polynomiality).
• Connection to Refined Counting: The results link logarithmic GW/PT theory to refined tropical curve counting and Hodge integrals, suggesting deep connections between sheaf theory, tropical geometry, and higher DR (double ramification) cycles.

Summary

Maulik and Ranganathan successfully extend the GW/PT correspondence to the most general toric setting (singular boundaries) by developing a robust inductive framework based on logarithmic degenerations. They overcome the technical hurdle of "exotic" insertions through a refined "rubber calculus" and establish the rationality and polynomiality of the generating series, thereby solving a long-standing conjecture and opening the door to further generalizations in enumerative geometry.