The relative Clemens Conjectures for $\frac{1}{2}$-log Calabi-Yau threefolds

This paper formulates and verifies a relative analogue of the Clemens conjecture for $\frac{1}{2}$-log Calabi-Yau threefolds by establishing a perfect deformation/obstruction duality that suppresses virtual complications, thereby reducing relative Gromov-Witten invariants to classical enumerative counts of rational curves with balanced normal bundles.

The Gist

Imagine you are an architect trying to count the number of unique, straight paths a hiker can take through a specific, magical forest.

This paper is about solving a very difficult counting problem in the world of algebraic geometry (which is essentially the study of shapes defined by equations). The author, Rodolfo Aguilar, is trying to figure out exactly how many "rational curves" (think of them as perfect, smooth, one-dimensional loops or lines) exist on a specific type of 3D shape, and what their properties are.

Here is the breakdown of the paper using simple analogies:

1. The Old Problem: The "Quintic" Forest

For a long time, mathematicians had a famous set of rules called the Clemens Conjectures. These rules predicted that in a specific type of 3D forest (called a Quintic Threefold), there are only a finite number of these perfect paths for any given length. Furthermore, these paths are "balanced"—meaning they are perfectly centered and stable, not wobbling or leaning to one side.

• The Catch: In the real world (and in math), things get messy. Sometimes, you can find infinite families of paths, or the paths might be "unstable" (like a wobbly table). The old rules worked for the "perfect" forest, but they broke down when the forest had a bit of a twist or a boundary.

2. The New Idea: The "Half-Log" Forest with a Fence

Aguilar asks: What if we look at a forest that has a specific type of fence around it?

In math terms, he looks at a pair $(X, Y)$, where $X$ is the 3D forest and $Y$ is a 2D surface (the fence) inside it. He focuses on a special case where the geometry is "half-Calabi-Yau" (a fancy way of saying the math balances out perfectly only if you count the fence twice).

The New Rule (The Conjecture):
If you fix a specific set of points on the fence (the boundary $Y$) and ask, "How many perfect paths go through exactly these points?", the answer should be:

1. Finite: There is a specific, countable number of paths.
2. Balanced: Every single one of these paths is perfectly stable (mathematically, they have a "balanced normal bundle," which we can think of as the path being perfectly centered in the forest, not leaning left or right).

3. The Difficulty: The "Virtual" Fog

Why hasn't anyone proven this before?
In modern math, when we try to count these paths, we often get stuck in a "fog" of virtual complications.

• The Analogy: Imagine trying to count cars on a highway, but some cars are driving in circles, some are double-parked, and some are "ghost cars" that exist only in a simulation. Standard counting methods get confused by these ghosts and give you the wrong number.
• In the "absolute" case (the old Quintic forest), this fog is so thick that we can't easily separate the real cars from the ghosts.

4. The Solution: The "Specialization" Trick

This is the paper's big breakthrough. Aguilar (with help from Adrian Zahariuc in the appendix) shows that in this specific "Half-Log" forest with the fence, the fog disappears.

They use a technique called Specialization:

• The Analogy: Imagine you are trying to count how many ways you can arrange furniture in a room. It's hard if the room is full of random obstacles. But, if you slowly move the furniture into a corner (specializing the problem), the obstacles vanish, and the arrangement becomes obvious.
• By moving the problem to a specific, simpler version of the forest (specifically, Prime Fano threefolds, which are like very symmetrical, high-quality forests), they prove that the "ghost cars" and "wobbly paths" simply don't exist in this setting.

5. The Result: A Clean Count

Because the fog clears up, the complex, modern mathematical tools (called Gromov-Witten invariants) stop giving "virtual" guesses and start giving honest, classical counts.

The Conclusion:
For these specific types of forests (Prime Fano threefolds of index two), the conjecture is TRUE.

• If you pick a set of points on the boundary fence, there is a finite number of perfect paths connecting them.
• Every single one of those paths is perfectly balanced and stable.
• We can count them exactly, without any mathematical "ghosts" interfering.

Summary in One Sentence

The author proves that if you look at a specific type of 3D shape with a boundary fence, the messy, confusing math that usually makes counting paths impossible suddenly becomes clean and simple, confirming that there are exactly the right number of perfect, stable paths passing through any chosen set of points on the fence.

Technical Summary

1. Problem Statement and Context

The Classical Context:
The paper addresses the Clemens Conjectures (1987), which predict the enumerative geometry of rational curves on a generic quintic Calabi-Yau threefold ($X \subset \mathbb{P}^4$). The conjectures rely on two pillars:

1. A connection between Hodge theory (Abel-Jacobi maps) and deformation theory.
2. A perfect duality between deformations and obstructions ($H^0(N_{C/X}) \cong H^1(N_{C/X})^*$).

Under these conditions, Clemens predicted that a generic quintic contains a finite number of rational curves of each degree, all with a "balanced" normal bundle $\mathcal{O}(-1) \oplus \mathcal{O}(-1)$.

The Obstacle:
In general non-rigid Calabi-Yau threefolds, or in the standard Log Calabi-Yau setting (pairs $(X, Y)$ where $K_X + Y \sim 0$), this perfect duality breaks down. Consequently, naive enumerative expectations often fail (e.g., Voisin showed examples with infinite families of lines), and modern Virtual Gromov-Witten (GW) invariants are difficult to relate to honest classical counts due to degenerate maps and multiple covers.

The Specific Problem:
The author seeks to formulate a relative analogue of the Clemens conjectures for 1/2-Log Calabi-Yau threefold pairs $(X, Y)$, defined by the condition:
$$K_X + 2Y \cong \mathcal{O}_X$$
The goal is to determine if, for a generic configuration of points $Z = C \cap Y$ on the boundary divisor $Y$, the number of rational curves anchored to $Z$ is finite and if they possess the balanced normal bundle property.

2. Methodology

The paper employs a synthesis of classical deformation theory, log-Hodge theory, and modern specialization techniques.

A. Restoration of Duality (Theoretical Framework):

• The author establishes that for the specific case of 1/2-Log CY ($K_X + 2Y \sim 0$), the deformation/obstruction duality is restored.
• Theorem 3.1: For a smooth curve $C \subset X$ intersecting $Y$ transversally, there is an isomorphism:
  $$H^0(N_{C/X}(-Y)) \cong H^1(N_{C/X}(-Y))^*$$
  This is derived via Serre duality and the adjunction formula, utilizing the specific canonical bundle relation $K_X|_C \cong \mathcal{O}_C(-2Y)$.
• Theorem 3.2: The map $\delta$ in the long exact sequence of cohomology is shown to be the dual of the infinitesimal log Abel-Jacobi map ($dAJ$). This implies that if the infinitesimal map is injective (Question 3.3), the deformations are unobstructed.

B. Relative Gromov-Witten Formalism:

• The problem is framed using Jun Li's relative GW theory, where rational curves are viewed as stable maps to a target expansion (a chain of rubber components) relative to the divisor $Y$.
• The evaluation map $ev: \mathcal{M}_\Gamma(X, Y) \to Y^e$ sends a curve to its intersection points with $Y$.

C. Specialization Techniques (Computational Engine):

• To bridge the gap between virtual invariants and classical counts, the paper utilizes specialization methods developed by Zahariuc [Zah18].
• Instead of computing virtual classes directly, the authors specialize the configuration of points $\xi \in Y^e$ to a general position. This allows them to prove that the relative moduli space trivializes: the fiber of the evaluation map contains no "rubber" components (non-trivial target expansions) and no reducible domains.

3. Key Contributions

1. Formulation of Relative Clemens Conjectures:
   The paper proposes Conjecture 1.1 (and Conjecture 3.1): For a generic 1/2-Log CY pair $(X, Y)$ and a fixed intersection configuration $Z$ of $e$ points on $Y$, there are only finitely many rational curves of degree $e$ intersecting $Y$ exactly at $Z$. Furthermore, these curves have the relative normal bundle:
   $$N_{C/X}(-Y) \cong \mathcal{O}_C(-1) \oplus \mathcal{O}_C(-1)$$
   (Which implies the absolute normal bundle is $\mathcal{O}(e-1) \oplus \mathcal{O}(e-1)$).

2. Rigorous Verification for Prime Fano Threefolds:
   The conjecture is rigorously proven for Prime Fano threefolds of index two (smooth del Pezzo threefolds of degree $d \in \{2, 3, 4, 5, 8\}$). These are the canonical examples of 1/2-Log CY pairs where $-K_X = 2Y$.

3. Trivialization of Virtual Complications:
   The paper demonstrates that in this specific geometric setting, the "virtual" nature of GW theory collapses. The relative moduli space fiber over general points is reduced, finite, and consists entirely of honest closed immersions with smooth sources. This eliminates the need for virtual fundamental classes to count these curves; classical intersection theory suffices.

4. Main Results

Theorem 1.1 (Main Theorem):
Let $X$ be a smooth del Pezzo threefold of degree $d \in \{2, 3, 4, 5, 8\}$ (general if $d=2$). Let $\xi = (\xi_1, \dots, \xi_e)$ be a configuration of $e$ general points on the hyperplane class $Y$.
The fiber of the relative evaluation morphism $ev: \mathcal{M}_\Gamma(X, Y) \to Y^e$ over $\xi$ satisfies:

1. No Rubber Levels: There are no maps with non-trivial target expansions.
2. No Reducible Domains: The domain curves are irreducible and smooth.
3. Isomorphism: The fiber is isomorphic to the fiber of the absolute evaluation map $M_{0,e}(X, e) \to X^e$ restricted to $\xi$.
4. Enumerative Reality: The fiber is finite, reduced, and consists of closed immersions $f: \mathbb{P}^1 \to X$ that intersect $Y$ transversally at $\xi$ with balanced normal bundle $\mathcal{O}(e-1) \oplus \mathcal{O}(e-1)$.

Proof Strategy (Appendix A):
The proof (joint with Zahariuc) proceeds by induction on the degree $e$.

• Smoothness: It is shown that the domain curve must be smooth; assuming a node leads to a contradiction regarding the dimension of the moduli space and the genericity of the points.
• Transversality: It is proven that the image of the curve cannot be contained in $Y$ (using specialization to a hyperplane section of $Y$).
• Balanced Bundle: The result relies on the irreducibility of the moduli space of rational curves on Fano threefolds (Lehmann-Tanimoto [LT19]) and the fact that the main component contains unramified maps with balanced normal bundles.

5. Significance and Future Directions

• Resolution of Enumerative Ambiguity: This work provides a rare instance where modern virtual enumerative machinery (GW theory) cleanly reduces to classical, honest counting. It validates the intuition that "anchored" rational curves in 1/2-Log CY settings behave much more rigidly and predictably than in the absolute Calabi-Yau case.
• Contrast with Absolute Case: While the Clemens conjectures remain open (and partially false in the naive sense) for general Calabi-Yau threefolds due to Voisin's counterexamples, this paper proves the "Clemens phenomenon" holds rigorously for a large class of Fano threefolds when viewed through the relative lens.
• Geometric Manin's Conjecture: The paper highlights a fascinating interplay between the Geometric Manin's Conjecture (which predicts the asymptotic growth of global moduli components) and the Relative Clemens framework (which predicts the exact finite trivialization of these components under maximal incidence conditions). This suggests a new avenue for research connecting global distribution laws with local enumerative rigidity.

In summary, the paper successfully constructs a robust theoretical framework for counting rational curves on specific Fano threefolds by leveraging the unique duality of 1/2-Log Calabi-Yau geometry, effectively turning a difficult virtual problem into a solvable classical one.