🔢 mathematics

Root bounds of vertical systems using tropical geometry

This paper establishes that the generic number of complex zeros for augmented vertically parametrized sparse polynomial systems equals the tropical intersection number of a tropical and a classical linear space, while also providing mixed volume expressions, bounds on positive zeros, and a Julia implementation for these results.

Original authors: Elisenda Feliu, Paul Alexander Helminck, Oskar Henriksson, Yue Ren, Benjamin Schröter, Máté L. Telek

Published 2026-05-11

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Elisenda Feliu, Paul Alexander Helminck, Oskar Henriksson, Yue Ren, Benjamin Schröter, Máté L. Telek

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

Imagine you are a detective trying to solve a mystery: How many solutions does a complex mathematical puzzle have?

In the world of chemistry and biology, these puzzles often describe how molecules interact in a reaction network (like a cell's internal machinery). The "solutions" to these puzzles are the steady states—moments where the system settles down and stops changing. Scientists want to know: Can this system settle into one state, or can it flip between many different states?

This paper, titled "Root Bounds of Vertical Systems Using Tropical Geometry," introduces a new, clever way to count these solutions without getting bogged down in the messy algebra that usually makes these problems impossible to solve.

Here is the breakdown of their method using simple analogies:

1. The Problem: Too Many Variables, Too Many Dependencies

Usually, when you have a system of equations, you assume every number (coefficient) can change independently. But in real-world chemistry, numbers are often linked. If you change the amount of Enzyme A, it might automatically change the rate of Reaction B because they share a common ingredient.

The authors call these "Vertical Systems." Imagine a tall building (the system) where the floors (equations) are connected by elevators (linear dependencies). You can't just change one floor without affecting the others. Traditional math tools often fail here because they treat every floor as if it were independent.

2. The Solution: "Tropical Geometry" (The Shadow Method)

The authors use a branch of math called Tropical Geometry.

The Analogy: Imagine shining a bright light on a complex 3D sculpture (your chemical system). The sculpture is hard to measure directly. But the shadow it casts on the wall (the "tropicalization") is much simpler. It turns curved, twisting lines into straight, flat lines and sharp corners.
The Trick: Instead of solving the hard 3D puzzle, they solve the easy 2D shadow puzzle. If the shadow has 3 distinct intersection points, the original 3D sculpture likely has 3 solutions too.

3. The Strategy: Re-embedding the Puzzle

To make the shadow even clearer, the authors perform a "Toric Re-embedding."

The Analogy: Imagine you have a tangled ball of yarn (the equations). It's hard to see where the strands cross. The authors take the yarn and stretch it out onto a flat table, adding a few extra guide rails (linear forms) to keep it organized.
Now, the problem becomes finding where a straight line (the linear part) crosses a grid of curves (the toric part). In the "shadow world" (Tropical Geometry), these become simple geometric shapes intersecting.

4. The Main Discovery: Counting the Crossings

The paper proves that the maximum number of solutions is exactly equal to the number of times these "shadow shapes" cross each other.

The "Cotransversal" Shortcut: Sometimes, the shapes are perfectly aligned in a special way (called "cotransversal"). When this happens, the authors show you don't even need to draw the shadows. You can just calculate a "Mixed Volume" (a fancy way of measuring the size of a shape) to get the answer instantly. It's like knowing a room's volume just by looking at its floor plan, without needing to measure every wall.

5. Real-World Application: The Phosphorylation Network

The authors tested their method on a very common biological system: Multi-site Phosphorylation.

The Scenario: Think of a protein that can be "tagged" with phosphate groups one by one, like adding beads to a string. The system can exist in many different tagged states.
The Result: For a protein with $k$ tagging sites, the authors confirmed that the system can have exactly $2k + 1$ different steady states.
Why it matters: This confirms a long-standing mathematical guess about how complex these biological switches can get. Their method calculated this number much faster than previous methods, especially for larger systems.

6. The "Positive" Twist: Real vs. Imaginary

In math, equations can have "imaginary" solutions that don't exist in the real world. But in chemistry, we only care about positive solutions (you can't have negative amounts of a molecule).

The paper provides a way to count only the real, positive solutions.
They found that while the total number of mathematical solutions might be high, the number of real, physical solutions is often lower. Their method gives a "lower bound" (a guaranteed minimum) for these real solutions by looking at the "positive shadow" of the system.

Summary

The authors built a digital toolkit (written in the Julia programming language) that acts like a high-tech X-ray machine.

It takes a messy chemical equation system.
It flattens it into a simpler "shadow" version using Tropical Geometry.
It counts the intersections in the shadow to tell you the maximum number of steady states.
It uses special shortcuts (Cotransversality and Toric structures) to speed up the calculation when the system has specific patterns.

The Bottom Line: They found a faster, more reliable way to count how many stable states a chemical system can have, proving that for certain biological networks, the number of states follows a predictable pattern ( $2k+1$ ), and they did it by looking at the "shadows" of the equations rather than the equations themselves.

Technical Summary: Root Bounds of Vertical Systems Using Tropical Geometry

Problem Statement
The paper addresses the fundamental problem in computational algebraic geometry of determining the maximal number of (complex, real, or positive) zeros for sparse polynomial systems where coefficients are not independent but exhibit linear dependencies. Specifically, the authors focus on augmented vertically parametrized systems, which take the form:
$F = (Cx^M, Lx - b) \in \mathbb{C}[a, b][x^{\pm}]^n$
Here, $x$ represents variables, while $a$ and $b$ are parameters. The matrix $C$ is "vertical," meaning each parameter appears in exactly one column, allowing for linear dependencies among coefficients of the same monomial. Such systems arise naturally in optimization (critical points) and chemical reaction network theory (steady states). While classical results like Bernstein's theorem provide root bounds for systems with independent coefficients, the case with linear dependencies is less explored. The goal is to compute the generic root count (the number of nondegenerate complex zeros for generic parameters) and bounds on the maximal positive root count (nondegenerate positive real zeros).

Methodology
The authors employ tropical geometry to transform the algebraic problem of counting roots into a combinatorial problem involving polyhedral geometry. The core strategy involves three main steps:

Toric Reembedding: The system $F$ is re-embedded into a higher-dimensional space $(C^*)^{r+n}$ by introducing auxiliary variables $y$ . This transforms the problem into finding the intersection of a toric variety $V(y - x^M)$ and a linear space $V(Cy, Lx - b)$. This decomposition separates the problem into a binomial part (easy to tropicalize) and a linear part.
Tropicalization: The authors compute the tropical varieties of these components.
- The tropicalization of the binomial ideal $\langle y - x^M \rangle$ is a classical linear space (specifically, the row span of $[M \mid I_n]$ ).
- The tropicalization of the linear ideal $\langle Cy, Lx - b \rangle$ is determined by the matroid of the augmented coefficient matrix.
Stable Intersection: The generic root count is calculated as the tropical intersection number of these two tropical varieties. The authors utilize the Transverse Intersection Theorem, which states that if tropical varieties intersect transversely, the number of intersection points (weighted by multiplicity) equals the number of algebraic solutions. For positive roots, they utilize positive tropicalizations and a positive analog of the transverse intersection theorem to derive lower bounds.

Key Contributions and Results

Exact Generic Root Count (Theorem A): The paper proves that the generic number of nondegenerate complex zeros for an augmented vertically parametrized system is exactly the tropical intersection number of a tropical linear space and a classical linear space in $\mathbb{R}^{r+n}$ $R^{r + n}$ .
- Formula: $grc(F) = \text{rowspan}([M \mid I_n]) \cdot \text{Trop}(\langle Cy, Lx - b \rangle)$ .
- In the purely vertical case ( $d=0$ ), this simplifies to a formula involving the degree of the monomial map and an intersection in $\mathbb{R}^r$ .
Lower Bounds for Positive Roots (Theorem B): For systems with real coefficients, the maximal number of nondegenerate positive zeros is bounded from below by the number of intersection points between the positive tropicalizations of the binomial and linear parts.
Specialized Structural Bounds:
- Cotransversality: If the matroids associated with the coefficient matrices are cotransversal, the generic root count simplifies to a mixed volume computation (Theorem 4.2). This connects the tropical approach to classical Bernstein-type bounds but applies to systems with linear dependencies.
- Parametrically Toric Systems: If the vertical part admits a monomial parametrization (common in reaction networks), the authors provide upper and lower bounds for positive roots that are computed in $\mathbb{R}^n$ rather than the higher-dimensional reembedded space, often yielding sharper bounds (Theorem 4.7).
Application to Reaction Networks: The framework is applied to multi-site phosphorylation networks. The authors demonstrate that these networks possess the necessary cotransversal and toric structures, allowing their steady state degrees to be computed via mixed volumes. They confirm the known result that the steady state degree for a $k$ -site network is $2k+1$ .

Implementation
The authors have implemented these algorithms in a Julia package named VerticalRootCounts.jl, built on the OSCAR computer algebra system. The package supports:

Generic root count computation via stable tropical intersection (Algorithm 1).
Mixed volume computation when cotransversality is detected.
Toric bounds for systems with known exponent matrices.
Lower bound estimation for positive roots via parameter specialization and tropical perturbation.

Significance and Claims
The paper claims that its approach provides a general-purpose upper bound on the number of nondegenerate zeros for any polynomial system (since any system is a specialization of a vertically parametrized one). This bound is often tighter (lower) than the standard BKK (Bernstein-Kushnirenko-Khovanskii) bound because it accounts for the specific linear dependencies in the coefficients.

The authors emphasize that their method offers:

Exactness: It computes the exact generic root count, not just an upper bound, for the class of vertically parametrized systems.
Certificates: The use of tropical transversality provides a certificate that the parameters are generic enough to achieve the root count.
Computational Efficiency: By exploiting special structures (cotransversality and toricity) found in biochemical networks, the method can be significantly faster than general tropical intersection or Gröbner basis methods.
Positive Root Analysis: It provides a systematic way to derive lower bounds for positive real roots, a quantity of critical importance in applications like chemical reaction network theory where multistationarity (multiple positive steady states) is a key phenomenon.

The paper concludes that while the general tropical intersection approach is powerful, recognizing and exploiting the specific matroidal and toric structures inherent in application-driven systems (like phosphorylation networks) leads to both computational speedups and sharper theoretical bounds.