🔢 mathematics

Duality for Landau-Ginzburg models

This article surveys various duality statements associated with a pair comprising a smooth complex quasi-projective variety and a regular function on it, serving as a tribute to Bumsig Kim.

Original authors: Claude Sabbah

Published 2026-04-16

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Claude Sabbah

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Two Ways to Look at a Landscape

Imagine you have a beautiful, complex landscape (a mathematical shape called a variety) and a function that assigns a "height" or "energy" value to every point on that landscape (this is the function $f$ ). In physics and math, this setup is often called a Landau-Ginzburg model. It's like a topographical map where the terrain represents a physical system.

The author, Claude Sabbah, is trying to answer a fundamental question: How do we measure the "shape" of this landscape when we look at it through two different lenses?

Lens A (The Smooth View): We look at the landscape with a standard ruler, but we add a special "twist" to our measurements based on the height function $f$ . This is the Twisted de Rham complex.
Lens B (The Critical View): We zoom in only on the "peaks and valleys" (the critical points where the slope is zero) and measure the shape there. This is related to the Kontsevich complex.

The paper's main goal is to prove that these two ways of measuring the landscape are actually dual to each other. They are like two sides of the same coin. If you know the shape from Lens A, you automatically know the shape from Lens B, and vice versa.

The Problem: The "Infinity" Trap

The author starts by pointing out a tricky problem.

If you try to measure the landscape using Lens A, the math usually works out perfectly. The numbers you get are finite and manageable.
However, if you try to measure using Lens B (focusing only on the peaks and valleys), the math often explodes. You get infinite amounts of data, which is useless for calculation.

The Analogy: Imagine trying to count the number of trees in a forest.

Lens A is like counting trees in a specific, bounded park. You get a nice, finite number.
Lens B is like trying to count trees in the entire universe. Unless the forest is finite, you get an infinite number, and the calculation breaks.

The Solution: Introducing the "Magic Variable" ( $u$ )

To fix the "infinity" problem, Sabbah introduces a new variable, $u$ . Think of $u$ as a zoom knob or a tuning dial.

Instead of looking at the landscape with a fixed ruler, we look at it with a ruler that changes depending on the setting of $u$ .

When $u$ is set to a normal number, we get our standard measurements.
When $u$ is set to zero, we get the "infinity" problem back.
But, by treating $u$ as a variable (like a polynomial), we can organize the infinite mess into a neat, finite structure.

The Analogy: Imagine you are trying to organize a chaotic pile of toys.

If you just look at the pile, it's a mess (infinite complexity).
But if you sort the toys by color (the variable $u$ ), you realize that for every color, there is a finite number of toys. The whole pile becomes a manageable, structured library.

The Main Discoveries (The Theorems)

The paper proves three major things, which the author calls Theorem A, Theorem B, and Theorem C.

1. Theorem A: The "Perfect Pairing"

This theorem says that if you use the "Magic Variable" $u$ to organize your measurements on a "good" version of the landscape (a projective version where the edges are well-behaved), you get a perfect match between the two lenses.

The Analogy: Imagine you have two different maps of the same city. One map shows the roads (Lens A), and the other shows the subway lines (Lens B). Theorem A proves that these two maps are perfectly compatible. If you know the road distance between two points, you can calculate the subway distance, and they fit together like a lock and key. This "lock and key" relationship is called a duality pairing.

2. Theorem B: The "Universal Truth"

Theorem A works well if you are careful about the edges of your map. But what if you want to ignore the edges and just look at the open space?
Theorem B says that if you allow your "zoom knob" $u$ to go into the "negative" range (using fractions like $1/u$ ), the two lenses become identical even without worrying about the edges.

The Analogy: It's like realizing that whether you are looking at a city through a window (Lens A) or a telescope (Lens B), if you zoom out far enough (using $1/u$ ), you see the exact same city layout. The distinction between "inside" and "outside" disappears.

3. Theorem C: The "Formal" View

Sometimes, we don't care about the exact value of $u$ ; we just want to know how the system behaves as $u$ gets very, very small (approaching zero).
Theorem C proves that even in this "formal" limit (looking at the system as a series of approximations), the duality still holds, provided the "peaks and valleys" of the landscape are contained in a finite area.

The Analogy: Imagine listening to a song. You can listen to the whole song (Theorem A), or you can listen to the song played backward (Theorem B). Theorem C says that even if you just listen to the very first second of the song and try to guess the rest, you can still understand the whole melody, as long as the song doesn't go on forever.

Why Does This Matter?

This paper is a bridge between different fields of mathematics and physics:

Singularity Theory: Understanding how shapes break or change at sharp points.
Mirror Symmetry: A concept in string theory where two completely different-looking universes are actually the same.
D-Modules: A way of doing calculus on shapes that is very powerful but hard to visualize.

The "Takeaway" Metaphor:
Think of the landscape as a musical instrument.

Lens A is the sound you hear when you pluck a string.
Lens B is the vibration pattern inside the wood of the instrument.
Usually, calculating the vibration pattern is a nightmare (infinite complexity).
Sabbah's work provides a new way to tune the instrument (the variable $u$ ) so that the sound and the vibration pattern become perfectly synchronized. He proves that you can predict the internal vibration just by listening to the sound, and vice versa, no matter how you tune the instrument.

Summary for the General Audience

Claude Sabbah has solved a puzzle about how to measure complex shapes. He showed that there are two different ways to measure these shapes: one that looks at the whole picture and one that looks at the "special points." Usually, these two methods give different or impossible results.

By introducing a clever mathematical tool (a variable $u$ ), he proved that these two methods are actually dual partners. They are two sides of the same coin. This discovery helps mathematicians and physicists understand the deep connections between geometry, calculus, and the laws of the universe, ensuring that no matter which "lens" they use, they are describing the same underlying reality.

Technical Summary: Duality for Landau-Ginzburg Models

Author: Claude Sabbah
Context: This paper addresses a question posed by Bumsig Kim regarding the comparison of cohomologies associated with a regular function $f$ on a smooth complex quasi-projective variety $U$ . It establishes duality statements for Landau-Ginzburg models, specifically focusing on the relationship between twisted de Rham cohomologies and their compact support counterparts, parameterized by a variable $u$ .

1. Problem Statement

The core problem is to compare two distinct types of cohomologies attached to a pair $(U, f)$ , where $U$ is a smooth connected complex quasi-projective variety of dimension $n$ and $f: U \to \mathbb{A}^1$ is a regular function:

Twisted de Rham Cohomology: The hypercohomology of the complex $(\Omega^\bullet_U, d + df)$ and its compact support variant.
Standard de Rham Cohomology: The cohomology of the complex $(\Omega^\bullet_U, df)$ .

Key Challenges:

Finiteness: While the hypercohomology of the twisted complex is generally finite-dimensional, the cohomology of $(\Omega^\bullet_U, df)$ is infinite-dimensional unless the critical set of $f$ is compact.
Parameter Dependence: Introducing a parameter $u$ to form the complex $(\Omega^\bullet_U[u], u d + df)$ leads to a $\mathbb{C}[u]$ -module. In general, this module is not free or of finite rank (e.g., if $f=0$ ).
Duality: Establishing a perfect pairing between these cohomologies that respects the natural connections and duality structures (Serre duality vs. twisted duality).

2. Methodology

Sabbah employs a geometric and algebraic approach involving several sophisticated tools:

Good Projectivization: To handle the non-compactness of $U$ , the author introduces a "good projectivization" $(Y, f)$ , where $Y$ is a smooth quasi-projective variety containing $U$ as an open subset, and the boundary $H = Y \setminus U$ is a divisor with normal crossings. The function $f$ extends to a projective morphism $Y \to \mathbb{A}^1$ .
Logarithmic Complexes: Instead of working directly on $U$ , the analysis is shifted to the logarithmic de Rham complex $(\Omega^\bullet_Y(\log H)[u], u d + df)$ . This allows for the control of singularities at infinity.
Kontsevich Complexes: The paper utilizes the Kontsevich complex $\Omega^\bullet_f$ , defined as the subcomplex of $\Omega^\bullet_Y(\log H)$ consisting of forms $\omega$ such that $df \wedge \omega$ remains logarithmic.
D-Module Theory: The cohomologies are interpreted as de Rham cohomology of specific $\mathcal{D}$ -modules (specifically, extensions of the module $(\mathcal{O}_U[u, u^{-1}], d + df/u)$ ). This includes the use of Rees modules to formalize the dependence on the parameter $u$ .
Mixed Hodge Modules: For proving freeness properties, the theory of mixed Hodge modules (M. Saito) is applied to analyze the pushforwards of filtered $\mathcal{D}$ -modules.
Formal Power Series: In specific cases (Theorem bC), the analysis is extended to formal power series rings $\mathbb{C}[[u]]$ to handle cases where the critical set is compact but the function is not proper.

3. Key Contributions and Results

Theorem A: Freeness and Duality on the Projective Model

The central result establishes that for a good projectivization $(Y, f)$ :

The hypercohomology modules $\mathbb{H}^k(Y, (\Omega^\bullet_Y(\log H)[u], u d + df))$ and their twisted compact support counterparts are free $\mathbb{C}[u]$ -modules of finite rank.
These modules are equipped with a meromorphic connection having a pole of order at most 2 at $u=0$ and a regular singularity at $u=\infty$ .
There exists a natural perfect pairing:
$\mathbb{H}^{n+k}(Y, (\Omega^\bullet_Y(\log H)[u], u d + df)) \otimes_{\mathbb{C}[u]} \mathbb{H}^{n-k}(Y, (\Omega^\bullet_Y(\log H)(-H)[u], u d - df)) \to \mathbb{C}[u]$
This pairing is compatible with the connections and independent of the choice of projectivization.
Significance: This generalizes the Brieskorn lattice of singularity theory to the global, non-compact setting of Landau-Ginzburg models.

Theorem B: Generic Pairing and D-Module Interpretation

By restricting to the generic point (working over $\mathbb{C}[u, u^{-1}]$ ), the paper proves:

The restriction maps induce isomorphisms between the logarithmic cohomology on $Y$ and the de Rham cohomology of the $\mathcal{D}$ -module extensions on $U$ .
Specifically, $\mathbb{H}^{n+k}(Y, \dots) \cong H^k_{dR}(U, E^f_{U})$ and the twisted version corresponds to cohomology with compact support $H^{-k}_{dR, c}(U, E^{-f}_{U})$ .
This yields a perfect pairing over $\mathbb{C}[u, u^{-1}]$ , clarifying the independence of the projectivization choice.

Theorems bA and bC: Formalization and Tame Functions

Theorem bA: The results of Theorem A hold when replacing polynomials $\mathbb{C}[u]$ with formal power series $\mathbb{C}[[u]]$ .
Theorem bC: Under the condition that the critical set of $f$ is compact (Condition 9.3), the cohomology $\mathbb{H}^k(U, (\Omega^\bullet_U[[u]], u d + df))$ is a free $\mathbb{C}[[u]]$ -module.
Katz-Tameness: For "Katz-tame" functions (where the cone of $Rf_! \mathbb{C}_U \to Rf_* \mathbb{C}_U$ has constant cohomology), the twist by $(-H)$ in the pairing can be omitted, simplifying the duality structure to resemble the proper case.

Geometric Application (Section 9.6)

The paper applies these results to a specific geometric construction involving the quotient of a homogeneous polynomial $f$ by a cyclic group action $\mu_d$ . It demonstrates that the resulting Landau-Ginzburg model satisfies the tameness conditions, allowing the application of the derived duality theorems.

4. Significance

Unification of Dualities: The paper successfully unifies the duality of twisted de Rham complexes (related to quantum cohomology and mirror symmetry) with classical Serre duality and D-module theory.
Resolution of Finiteness Issues: By introducing the parameter $u$ and working with logarithmic complexes on a compactification, the author resolves the issue of infinite-dimensional cohomology in non-proper settings, proving that the relevant structures are free modules of finite rank.
Foundation for Mirror Symmetry: These duality statements are crucial for the mathematical formulation of Landau-Ginzburg models in the context of Homological Mirror Symmetry (HMS), particularly in comparing the A-model (symplectic) and B-model (complex) sides.
Generalization of Brieskorn Lattices: The construction provides a global, non-singular analogue of the local Brieskorn lattice, essential for studying the variation of Hodge structures in non-compact geometries.

In summary, Sabbah provides a rigorous framework for understanding the cohomological duality of Landau-Ginzburg models, bridging the gap between local singularity theory and global algebraic geometry through the lens of $\mathcal{D}$ -modules and mixed Hodge structures.