Skip to main navigation Skip to search Skip to main content

A Clarke Matrix-Measure Contraction Certificate for Nonsmooth ODEs with Finite Spectral Verification and Mixed-Signal Circuit Applications

  • Changcheng Shi
  • , Arvind R Singh
  • , Michaelraj Kingston Roberts
  • , S Mohandass
  • , Yichuang Sun
  • , Jamie Lloret

Research output: Contribution to journal › Article › peer-review

Abstract

We study autonomous ordinary differential equations with locally Lipschitzvector fields that are not everywhere differentiable, as happens at switchingmanifolds and other non-smooth interfaces. For this setting, we introduce aClarke matrix measure obtained by taking the largest Euclidean matrix measureover the Clarke generalized Jacobian. On a forward-invariant set satisfyingthe segment condition for pairs of solutions, negativity of this measureyields strict Euclidean contraction with an explicit decay estimate. For continuouspiecewise-affine systems on convex domains, the same requirement isshown to be not merely sufficient but also necessary for strict Euclidean con-traction, and the verification reduces to a finite spectral test: one eigenvalueevaluation of the symmetric part of each mode matrix, with no decision variablesand no numerical solver. We further derive settling-time bounds anddemonstrate the analysis on two mixed-signal circuit models: a continuous-timesigma–delta modulator and a bang-bang phase-locked loop, illustratedthrough Monte Carlo trajectory studies and parameter sweeps of the certifiedrate. A numerical comparison with common quadratic Lyapunov methods,based on an exactly verifiable example, indicates the simplicity of the proposedtest and the conservatism it may entail.
Original languageEnglish
Article number119085
JournalChaos, Solitons and Fractals
Volume212
Early online date4 Sept 2026
DOIs
Publication statusE-pub ahead of print - 4 Sept 2026

Fingerprint

Dive into the research topics of 'A Clarke Matrix-Measure Contraction Certificate for Nonsmooth ODEs with Finite Spectral Verification and Mixed-Signal Circuit Applications'. Together they form a unique fingerprint.

Cite this