English

Random Switching between Vector Fields Having a Common Zero

Probability 2018-07-03 v3

Abstract

Let EE be a finite set, {Fi}iE\{F^i\}_{i \in E} a family of vector fields on Rd\mathbb{R}^d leaving positively invariant a compact set MM and having a common zero pM.p \in M. We consider a piecewise deterministic Markov process (X,I)(X,I) on M×EM \times E defined by X˙t=FIt(Xt)\dot{X}_t = F^{I_t}(X_t) where II is a jump process controlled by X:X: Pr(It+s=j(Xu,Iu)ut)=aij(Xt)s+o(s)\Pr(I_{t+s} = j | (X_u, I_u)_{u \leq t}) = a_{i j}(X_t) s + o(s) for iji \neq j on {It=i}.\{I_t = i \}. We show that the behavior of (X,I)(X,I) is mainly determined by the behavior of the linearized process (Y,J)(Y,J) where Y˙t=AJtYt,\dot{Y}_t = A^{J_t} Y_t, AiA^i is the Jacobian matrix of FiF^i at pp and JJ is the jump process with rates (aij(p)).(a_{ij}(p)). We introduce two quantities Λ\Lambda^- and Λ+\Lambda^+ respectively %called the {\em minimal} and {\em maximal average growth rate.} Λ\Lambda^- (respectively Λ+\Lambda^+) is defined as the {\em minimal} (respectively {\em maximal}) {\em growth rate} of Yt,\|Y_t\|, where the minimum (respectively maximum) is taken over all the ergodic measures of the angular process (Θ,J)(\Theta, J) with Θt=YtYt.\Theta_t = \frac{Y_t}{\|Y_t\|}. It is shown that Λ+\Lambda^+ coincides with the top Lyapunov exponent (in the sense of ergodic theory) of (Y,J)(Y,J) and that under general assumptions Λ=Λ+.\Lambda^- = \Lambda^+. We then prove that, under certain irreducibility conditions, XtpX_t \to p exponentially fast when Λ+<0\Lambda^+ < 0 and (X,I)(X,I) converges in distribution at an exponential rate toward a (unique) invariant measure supported by M{p}×EM \setminus \{p\} \times E when Λ>0.\Lambda^- > 0. Some applications to certain epidemic models in a fluctuating environment are discussed and illustrate our results.

Keywords

Cite

@article{arxiv.1702.03089,
  title  = {Random Switching between Vector Fields Having a Common Zero},
  author = {Michel Benaïm and Edouard Strickler},
  journal= {arXiv preprint arXiv:1702.03089},
  year   = {2018}
}

Comments

42 pages, 10 figures Proof of proposition 2.4 has been changed A section on the noncompact case has been added