Dissipative quadratizations of polynomial ODE systems

This paper establishes the existence of, develops an algorithm for computing, and demonstrates the application of quadratizations that preserve the dissipativity (stability) properties of polynomial ODE systems, thereby enabling more reliable model analysis and control in fields such as systems theory and synthetic biology.

The Gist

Imagine you are trying to navigate a complex, winding mountain road in a heavy fog. The road represents a mathematical model of a real-world system (like a chemical reaction, a population of animals, or an electrical circuit). The equations describing this road are polynomials, which can get very twisty and turny (cubic, quartic, etc.), making them incredibly hard to simulate on a computer.

Sometimes, to make the journey easier, mathematicians try to "flatten" the road. They transform the complex, curvy equations into simpler, quadratic ones (equations where the highest power is just 2, like $x^2$). This process is called Quadratization.

Think of quadratization like taking a jagged, rocky path and paving it with smooth, straight segments. It makes the computer's job much easier, allowing for faster simulations and better predictions.

The Problem: The "Smooth" Road Might Be a Cliff

Here is the catch: In the past, when people paved this road, they only cared about making it smooth. They didn't check if the new road still went downhill toward a safe valley.

In math terms, many real-world systems are dissipative. This means they naturally lose energy and settle down into a stable state (like a ball rolling to the bottom of a bowl). If you transform the equations but accidentally turn the "bowl" into a "hill," your computer simulation will crash. The ball will roll off the edge, and the model will become unstable and useless, even though the math looks correct.

The authors of this paper asked a simple but crucial question: "Can we pave the road to make it smooth (quadratic) without turning our safe valley into a dangerous cliff?"

The Solution: The "Stabilizer" Tool

The authors say, "Yes, we can!" and they built a tool to do it. Here is how they did it, using a few analogies:

1. The "Inner-Quadratic" Blueprint
First, they found a specific way to build the new road. They call this an "inner-quadratic" transformation.

• Analogy: Imagine you are building a new house. Instead of just throwing random bricks together, you use a specific architectural rule where every new room is built using materials from the existing rooms. This ensures the new structure is tightly connected to the old one.
• In the paper: They create new variables (new rooms) that are mathematically linked to the old ones in a way that guarantees the new system is still "connected" to the original reality.

2. The "Stabilizer" (The Safety Net)
This is the paper's biggest innovation. Once they have the new quadratic road, they add a special "safety net" called a stabilizer.

• Analogy: Imagine you are balancing a broom on your hand. If it starts to tip, you can add a tiny, invisible weight to the other side to keep it upright.
• In the paper: They take the new equations and add a special term (the stabilizer) that is mathematically zero when the system is behaving normally. However, if the system starts to wobble (become unstable), this term kicks in and pushes it back toward the safe valley. It's like a self-correcting steering wheel that only turns when you are drifting off course.

3. The Algorithm (The GPS)
They wrote a computer program (an algorithm) that:

1. Takes your messy, complex equations.
2. Builds the "inner-quadratic" road.
3. Calculates exactly how much "weight" (the stabilizer) to add to keep the system stable at specific points (like a ball resting at the bottom of a bowl).
4. Outputs a new set of equations that are simple (quadratic) but safe (dissipative).

Why Does This Matter?

The paper shows this works in three real-world scenarios:

• Reachability Analysis (The "Where can we go?" test): Imagine you are a safety engineer for a drone. You want to know: "If the wind blows this way, will the drone crash?" By using their method, engineers can simulate the drone's path much faster and with more confidence that the simulation won't glitch out.
• Synthetic Biology (The "Switch" test): In biology, cells often act like switches (on/off). If you model a cell's behavior with unstable math, you might think a cell will stay "on" forever when it actually shuts down. Their method ensures the model respects the cell's natural "off" switch.
• Coupled Oscillators (The "Swing" test): Imagine a group of swings connected by springs. If you model them poorly, the simulation might show them swinging infinitely high and breaking. Their method keeps the simulation grounded in reality.

The Bottom Line

This paper is like a new set of instructions for renovating a house. Before, builders would knock down a complex, old structure and build a simple, modern one, but they often forgot to check if the new foundation was strong enough.

Cai and Pogudin have provided a blueprint that ensures:

1. The new house is simple and easy to live in (Quadratic).
2. The new house is just as safe and stable as the old one (Dissipative).

They proved that such a renovation is always possible and gave us the tools to do it automatically. This means scientists and engineers can now simulate complex systems faster and with much greater trust that their results are real.

Technical Summary

Here is a detailed technical summary of the paper "Dissipative quadratizations of polynomial ODE systems" by Yubo Cai and Gleb Pogudin.

1. Problem Statement

The paper addresses the problem of quadratization in systems of polynomial Ordinary Differential Equations (ODEs). Quadratization is a transformation that converts a system with polynomial right-hand sides of arbitrary degree into an equivalent system where the right-hand sides are polynomials of degree at most two (quadratic).

While quadratization is a well-established technique used in model order reduction, synthetic biology, and reachability analysis, a critical gap exists: standard quadratization does not guarantee the preservation of stability properties.

• The Issue: A transformed quadratic system may be mathematically equivalent to the original system on the invariant manifold defined by the new variables, but it can exhibit different numerical behaviors off this manifold. Specifically, a standard quadratization might destroy the dissipativity (asymptotic stability) of the original system at its equilibrium points.
• The Goal: The authors aim to develop a method to construct quadratizations that preserve dissipativity at specified equilibrium points. This is crucial for applications like reachability analysis and control, where stability guarantees are required.

2. Methodology

The authors propose a two-step methodology based on the concept of inner-quadratic quadratizations and the use of stabilizers.

A. Inner-Quadratic Quadratization

The authors define an inner-quadratic set of new variables. A set of new variables $y = g(x)$ is inner-quadratic if every new variable $y_i$ can be expressed as a product of two elements from the set of original variables and previously defined new variables (i.e., $y_i = a \cdot b$, where $a, b \in \{x_1, \dots, x_n, y_1, \dots, y_{i-1}\}$).

• Significance: This structure ensures that the relationships between new variables are at most quadratic, providing the necessary algebraic flexibility to modify the system's dynamics without breaking the equivalence to the original system.
• Algorithm 1: The authors adapt an existing Branch-and-Bound algorithm (from QBee) to search for an optimal (minimal dimension) inner-quadratic quadratization.

B. Stabilizers and Dissipativity Preservation

Once an inner-quadratic quadratization is found, the authors introduce stabilizers.

• Definition: For a new variable $y_i = a_i b_i$, the stabilizer is defined as $h_i(x, y) = y_i - a_i b_i$. Note that $h_i$ vanishes when $y = g(x)$, meaning adding multiples of $h_i$ to the system's right-hand side does not change the trajectory on the invariant manifold.
• Theoretical Insight: The authors prove that by adding a linear combination of stabilizers to the right-hand side of the new variables ($y'$), one can arbitrarily shift the eigenvalues of the Jacobian matrix at the equilibrium points.
• Lemma 1 (Key Technical Tool): They utilize a linear algebra lemma stating that for a matrix $A$ and an upper triangular matrix $B$ with ones on the diagonal, the eigenvalues of $A - \lambda B$ will have negative real parts for sufficiently large $\lambda$.
• Algorithm 2: The algorithm iteratively increases a parameter $\lambda$ in the term $-\lambda h(x, y)$ added to the $y'$ equations. It checks the eigenvalues of the resulting Jacobian at the target equilibria until dissipativity (all eigenvalues having negative real parts) is achieved.

3. Key Contributions

1. Existence Theorem (Theorem 1): The paper proves that for any polynomial ODE system and any finite set of its dissipative equilibria, there exists a quadratization that preserves dissipativity at all those points.
2. Constructive Proof: The proof is constructive, directly leading to the design of the algorithms. It demonstrates that inner-quadratic quadratizations are sufficient to achieve this preservation.
3. Algorithmic Framework:
   • Algorithm 1: Computes an optimal inner-quadratic quadratization.
   • Algorithm 2: Takes an inner-quadratic quadratization and modifies its right-hand side using stabilizers to ensure dissipativity at user-specified equilibria.
4. Handling Numerical Instability: The paper highlights that different quadratizations of the same system can lead to drastically different numerical stability (demonstrated via a toy example where one quadratization is stable and another is unstable despite mathematical equivalence). The proposed method specifically targets the stable variant.

4. Results and Case Studies

The authors implemented their algorithms and validated them through several case studies:

• Toy Example: Demonstrated that adding stabilizers can transform an unstable quadratization into a stable one, matching the dynamics of the original equation accurately.
• Reachability Analysis (Duffing Equation): Applied the method to a Duffing oscillator. By preserving dissipativity at the origin, they successfully applied Carleman linearization techniques (from Forets and Schilling) to compute reachability sets, a task previously restricted to systems that were naturally quadratic or weakly nonlinear.
• Bistability Preservation: Applied to a chemical reaction network model. The algorithm successfully preserved the bistability (two stable equilibria) of the original system in the quadratized form.
• Scalability (Coupled Duffing Oscillators): Tested on systems of $n$ coupled oscillators ($n=1$ to $8$).
  • Findings: The number of new variables grows linearly with $n$.
  • Performance: Numerical eigenvalue checking (using NumPy) scaled well, handling up to 256 equilibria efficiently. Symbolic checking (Routh-Hurwitz) became computationally prohibitive as dimension increased, highlighting the importance of numerical verification for large systems.

5. Significance

• Bridging Theory and Practice: The work connects the algebraic transformation of ODEs with dynamical systems theory (stability). It ensures that the "convenience" of quadratic models does not come at the cost of losing physical or dynamical fidelity.
• Enabling Advanced Analysis: By guaranteeing dissipativity, the method unlocks the use of powerful analysis tools (like Carleman linearization for reachability) on a broader class of non-quadratic polynomial systems.
• Generalizability: The concept of "stabilizers" and "inner-quadratic" structures provides a framework that the authors suggest can be extended to other stability properties (e.g., limit cycles, Lyapunov functions) and potentially to non-polynomial systems via polynomialization.

In summary, this paper provides the first systematic approach to ensuring that the process of quadratizing polynomial ODEs does not inadvertently destroy the stability properties of the original model, offering both a theoretical guarantee and a practical algorithm for doing so.