Invariant-Manifold Jets for Explicit Facial Reduction in SOS Lyapunov Programs

Andréa Gourion. Author’s version submitted to the 2027 European Control Conference (ECC), received on 9 November 2026 and under review.

1The paper in brief

How can a computer combine numerical computation and exact verification to prove that a nonlinear system returns to equilibrium when linearization is inconclusive?

For a polynomial system, Sum-of-Squares (SOS) programming turns the search for a Lyapunov function into a Semidefinite Program (SDP). At such an equilibrium, decay is governed by reduced dynamics on an invariant manifold. When these dynamics are sufficiently flat, they can force the feasible Gram matrices of the SOS program onto a proper face of the positive semidefinite cone, which can complicate the numerical solution of the optimization problem.

The paper constructs explicit facial reduction certificates from these dynamics, before the program is solved.

Paper status
Submitted paper, under review. It has not been peer reviewed yet and may differ from the final version.

2Submitted version

Read the PDF8 pages · 0.3 MB

How to cite

A. Gourion, “Invariant-Manifold Jets for Explicit Facial Reduction in SOS Lyapunov Programs,” preprint, submitted to the 2027 European Control Conference (ECC), November 2026.

3Six videos, step by step

This series explains the paper step by step, with its computations and its proofs. Prerequisites: linear algebra, differential equations and the basics of differential calculus.

The videos are silent: each step is carried by an on-screen caption, held long enough to be read. A chapter starts the video at that point.

  1. Video 1. The problem and the tools

    13:39

  2. Video 2. Polynomials and computing the invariant manifold

    11:10

  3. Video 3. Theorem 1: forced kernels

    11:25

  4. Video 4. Theorem 2: the explicit reduction

    13:15

  5. Video 5. Adapted coordinates, non-polynomial fields

    09:09

  6. Video 6. Witnesses, exact certification, results

    11:54

The videos expand some proofs and add teaching examples that are not in the paper. The paper remains the reference.

4Code and data

The public repository gathers the MATLAB programs, the data and the rational certificates of the numerical study. Exact verification of the certificates starts no semidefinite solver. Only new SOS optimizations need YALMIP and MOSEK.

Open the repository (opens in a new tab)

An opportunity, a research collaboration, or a project to develop? Tell me about the context and what you would like to discuss.

get in touch