Singularities of diagonals of Laurent series for rational functions

The paper establishes that the complete diagonal of the Laurent series expansion of a nondegenerate rational function in $n$ complex variables admits analytic continuation throughout the complex torus, excluding only a specific Landau variety constructed from the discriminants of the denominator's face truncations.

The Gist

The Big Picture: Finding the "Hidden Map"

Imagine you have a complex, multi-layered cake (a rational function). This cake is made of many different ingredients (variables) mixed together. Mathematicians love to slice this cake in specific ways to understand its flavor.

One specific way of slicing is called taking a "diagonal."

• The Cake: A rational function $g(z)/f(z)$ involving $n$ variables (like $z_1, z_2, z_3...$).
• The Slice: Instead of looking at the whole cake, we look at a specific line running through it where the variables are linked together (e.g., $z_1 = t, z_2 = t^2, z_3 = t^3$).
• The Result: This slice gives us a new, simpler function (a Laurent series) that depends on fewer variables. This new function is the "diagonal."

The Problem:
This new diagonal function works perfectly in a safe, calm neighborhood (a "logarithmically convex domain"). But what happens if we try to walk further out into the wilderness? Does the function break? Does it explode? Where are the "cliffs" or "quicksand" that stop us from going further?

These cliffs are called singularities. The paper's goal is to draw a map that tells us exactly where these cliffs are located.

---

The Key Players

To understand the map, we need to know the characters in this story:

1. The Newton Polyhedron (The Shape of the Cake):
   Every polynomial has a shape made of its ingredients. If you plot the powers of the variables on a graph, they form a geometric shape (a polyhedron). Think of this as the blueprint of the cake.

   • The Paper's Rule: The cake must be "nondegenerate." This just means the cake is well-constructed and doesn't have weird, flat spots where the ingredients cancel each other out perfectly. Most cakes are like this.

2. The Landau Variety (The Danger Zone):
   This is the main discovery of the paper. It is a specific set of points (a "complex analytic set") where the diagonal function stops behaving nicely.

   • Analogy: Imagine you are walking on a frozen lake (the safe domain). The Landau variety is the thin ice or the cracks. If you step on it, you fall in. The paper tells us exactly where these cracks are.

3. The Amoeba (The Shadow):
   Mathematicians use a tool called an "amoeba" to visualize where the function goes wrong. If you take the cake and project its "shadow" onto a lower-dimensional floor, the shadow has holes and boundaries. The Landau variety is constructed by looking at the shadows of the cake's edges (faces of the Newton polyhedron).

---

How the Paper Solves the Puzzle

The author, Dmitriy Pochekutov, uses a clever trick to find the Landau variety. Here is the step-by-step process in plain English:

1. The "Shadow" Trick (Truncations)

The cake (the polynomial) has many faces (like the sides of a pyramid). The author looks at each face individually.

• He takes a "truncation," which is like cutting off the top of the cake and only looking at the bottom layer of that specific face.
• He asks: "If I only look at this specific layer, where does the function break?"

2. The "Critical Points" (The Tipping Point)

For each layer, he looks for a special condition where the function becomes unstable.

• Analogy: Imagine balancing a pencil on its tip. There is a specific point where it tips over. The paper calculates exactly where that tipping point is for every layer of the cake.
• Mathematically, this involves checking where the "gradient" (the slope) of the function is zero while the function itself is also zero.

3. Building the Map (The Landau Variety)

He takes all these tipping points from all the different layers and combines them.

• The result is a giant, complex shape called the Landau Variety ($L$).
• The Main Theorem: The paper proves that as long as you stay away from this Landau variety, you can walk (analytically continue) along any path you want, and your diagonal function will remain smooth and well-behaved. If you hit the Landau variety, that's where the function gets singular (breaks).

---

Why Does This Matter?

You might ask, "Who cares about these mathematical cliffs?"

1. Predicting Behavior: In physics and combinatorics (counting things), these diagonal functions often represent real-world probabilities or counts. Knowing where they break helps scientists understand the limits of their models.
2. Algebra vs. Transcendence: The paper hints at a bigger mystery: Is this diagonal function a simple algebraic equation (like a circle or parabola) or a wild, transcendental one (like $e^x$ or $\sin x$)?
   • Fun Fact: If you only have 2 variables, the diagonal is always a simple algebraic equation. But if you have 3 or more variables (which this paper studies), it gets complicated. Knowing the "cliffs" (Landau variety) is the first step to solving this mystery.

Summary Analogy

Imagine you are an explorer trying to map a new continent (the mathematical function).

• You start in a safe village (the domain of convergence).
• You want to know how far you can travel before hitting a wall or a swamp (singularities).
• The author looks at the geology of the continent (the Newton polyhedron).
• He examines the fault lines in the rock layers (the truncations of the faces).
• He draws a map of all the quicksand pits (the Landau variety).
• The Conclusion: "You can walk anywhere you want, as long as you stay off the quicksand. Here is exactly where the quicksand is."

This paper provides the ultimate "Do Not Enter" sign for these complex mathematical landscapes, allowing researchers to navigate them safely.

Technical Summary

1. Problem Statement

The paper addresses the problem of determining the singularities of the complete diagonal of a Laurent series expansion for a rational function $g(z)/f(z)$ in $n$ complex variables.

• Context: Given a rational function where $f(z)$ is a Laurent polynomial and $g(z)$ is coprime to $f$, one can expand it into a Laurent series centered at the origin. The "complete diagonal" refers to a specific sub-series obtained by restricting the multi-index of the coefficients to a sublattice generated by an $r$-tuple of vectors $Q = (q_1, \dots, q_r)$.
• Goal: To describe the domain of analytic continuation of this diagonal function $d_Q(t)$ and explicitly identify the set of singularities (the Landau variety) in terms of the denominator $f(z)$.
• Motivation: Understanding these singularities is a crucial step in determining whether such diagonals are algebraic or transcendental functions, particularly for dimensions $n \geq 3$. While the case for $n=2$ is known to yield algebraic functions, the higher-dimensional case remains complex.

2. Methodology

The author employs a combination of combinatorial geometry (Newton polyhedra and amoebas), complex analysis (integral representations), and algebraic geometry (stratification and Landau varieties).

A. Geometric Setup

• Newton Polyhedron ($\Delta_f$): The convex hull of the exponents of the Laurent polynomial $f$.
• Nondegeneracy: The paper assumes $f$ is nondegenerate with respect to its Newton polyhedron. This means that for any face $\delta$ of $\Delta_f$, the truncation $f_\delta$ has no critical points in the complex torus $(\mathbb{C}^*)^n$.
• Amoebas: The mapping $\Lambda(z) = (\log|z_1|, \dots, \log|z_n|)$ maps the zero set of $f$ to the "amoeba" $A_f$ in $\mathbb{R}^n$. The complement of the amoeba consists of convex components, each corresponding to a specific Laurent expansion.

B. Coordinate Transformation

To simplify the diagonal extraction, the author utilizes a monomial transformation $z = w^A$ (where $A$ is derived from the lattice basis containing $Q$). This transforms the problem from variables $z$ to variables $(t, u)$, where:

• $t \in (\mathbb{C}^*)^r$ are the variables of the diagonal.
• $u \in (\mathbb{C}^*)^s$ (with $s = n-r$) are the remaining integration variables.
  The rational function becomes $\tilde{f}(t, u)$, a Laurent polynomial in $u$ with coefficients depending on $t$.

C. Integral Representation

The diagonal $d_Q(t)$ is represented as an $s$-dimensional contour integral:
$$d_Q(t) = \int_{\tilde{\Gamma}_t} \omega(t)$$
where $\omega(t)$ is a rational differential form involving $\tilde{f}(t, u)$, and $\tilde{\Gamma}_t$ is a cycle in the homology group of the complement of the zero set of $\tilde{f}$.

D. Landau Variety Construction

The singularities of the integral are identified using the Landau variety theory. The author constructs a stratified set $S$ in the product space of parameters and integration variables. The Landau variety $L$ is defined as the image under projection of the critical points of this stratification.
Specifically, $L$ is the union of sets $L_\sigma$ associated with faces $\sigma$ of the Newton polyhedron of $\tilde{f}$:
$$L = \bigcup_{\sigma \in \Sigma} L_\sigma$$
where $L_\sigma$ is the discriminant set of the truncation $\tilde{f}_\sigma(t, u) = 0$ with respect to $u$.

3. Key Contributions and Results

Main Theorem (Theorem 1)

The central result states that if the denominator $f$ is nondegenerate for its Newton polyhedron, the complete $Q$-diagonal $d_Q(t)$ can be analytically continued along any path in the complex torus $T^r_t$ that avoids the explicitly defined Landau variety $L$.

• Construction of $L$: The variety $L$ is the union of discriminants associated with specific truncations of the transformed polynomial $\tilde{f}(t, u)$ to the faces of its Newton polyhedron.
• Explicit Definition: For a face $\sigma$ where the truncation depends on $t$, $L_\sigma$ is the set of $t$ such that the system:
  $$\tilde{f}_\sigma(t, u) = 0 \quad \text{and} \quad \nabla_u \tilde{f}_\sigma(t, u) = 0$$
  has a solution $u$.

Proof Strategy

The proof relies on the Thom Isotopy Theorem:

1. The domain $T = T^r_t \setminus L$ is shown to be linearly connected.
2. The integral representation defines a locally trivial homological fibration over $T$.
3. Because the fibration is locally trivial, the integration cycle can be continuously deformed as the parameter $t$ moves along a path in $T$.
4. This deformation ensures that the integral remains analytic, proving that $L$ contains all singularities.

4. Examples and Applications

The paper validates the theory with two examples:

1. Generalized Hypergeometric Function: The diagonal of $1/(1 - z_1 - (z_2 z_3)^l - z_3)$ is shown to have a singularity at a specific point $t_0$, which matches known results for the generalized hypergeometric function ${}_lF_{l-1}$.
2. Appell Function $F_4$: The diagonal of $1/(1 - z_1 - z_2 - z_3 - z_4)$ yields the Appell function $F_4$. The calculated Landau variety $L_\Delta$ is an explicit algebraic curve in $\mathbb{C}^2$, defining the domain of analyticity for this function.

5. Significance

• Multidimensional Generalization: This work extends the classical results on the diagonals of rational functions (previously well-understood for $n=2$) to arbitrary dimensions $n$.
• Explicit Singularity Locus: It provides a constructive, algorithmic method to find the exact location of singularities (the Landau variety) based solely on the geometry of the denominator's Newton polyhedron.
• Transcendence Theory: By characterizing the singularities, the paper lays the groundwork for investigating the algebraic or transcendental nature of these diagonals, a major open problem in enumerative combinatorics and mathematical physics.
• Interdisciplinary Impact: The results are relevant to statistical physics (Ising models), theoretical physics, and the theory of difference equations, where such diagonal series frequently appear.

In summary, Pochekutov establishes a rigorous bridge between the combinatorial geometry of the denominator's Newton polyhedron and the analytic properties of the resulting diagonal series, proving that the singularities are entirely determined by the discriminants of the polynomial's face truncations.