Textbook chaos theory identifies chaos through spectral instability. Compute the Jacobian of the dynamical system. Find its eigenvalues. If eigenvalues leave the unit circle (discrete time) or cross the imaginary axis (continuous time), the system becomes unstable, and the route to chaos begins: period-doubling, quasi-periodicity, intermittency.
Sornette, Saiprasad, and Troude (arXiv:2603.08191) construct a bounded three-dimensional system where all eigenvalues of the instantaneous Jacobian remain strictly inside the stability region at all times — and the system is nonetheless chaotic, with positive Lyapunov exponents and sensitive dependence on initial conditions.
The mechanism is non-normality. A matrix is non-normal when its eigenvectors are not orthogonal. In a non-normal system, even with all eigenvalues decaying, perturbations can grow transiently because non-orthogonal eigenvectors amplify components before the eventual decay takes hold. The amplification is not visible in the eigenvalues — it lives in the geometry of the eigenvector basis, not in the spectrum.
In a static non-normal system, the transient growth eventually subsides as the decaying eigenvalues win. But in this system, the non-normal amplification is continually regenerated by switching between configurations. Each configuration is individually stable. The transitions between configurations reinject energy into the non-orthogonal directions before the previous amplification has fully decayed. The result is sustained, bounded chaos — not through spectral instability but through perpetual transient growth.
This is structurally different from every known route to chaos. Period-doubling requires bifurcation. Quasi-periodicity requires incommensurate frequencies. Intermittency requires saddle-node collision. All require eigenvalues crossing a stability boundary. Non-normal chaos requires none of these. The eigenvalues never become unstable. The chaos comes entirely from the geometry of the directions, not the magnitudes of the modes.
The implication is diagnostic: eigenvalue analysis — the standard tool for assessing stability — is blind to this route. A system can pass every spectral stability test and still be chaotic. The stability criterion that the entire field relies on has a blind spot, and the blind spot is geometric, not spectral.
Sornette, Saiprasad, & Troude, "Non-Normal Route to Chaos," arXiv:2603.08191 (2026). Note (2026-11-06): the abstract of arXiv:2603.08191 (now v2; first and current abstracts share about 33 percent of their wording) has changed substantially since its first version. I wrote this essay in early 2026 and may have worked from the earlier version, so it can be accurate about the version I read and still differ from the paper as it now stands. I have not compared this essay with either version in this pass.