Exit times for a class of piecewise exponential Markov processes with two-sided jumps (Q2642039)

Let \(X\) be a right-continuous piecewise deterministic Markov process with iid jumps, described by the stochastic differential equation of the form \[ dX_t =\kappa X_t dt+dU_t, \] where \(U_t=\sum_{n=1}^{N_t} V_n\) is a compound Poisson process, and the jump sizes are iid with distribution \(G\), independent of the location of the jump. For an arbitrary level \(l\) let \(\tau:=\{\inf t>0:\) \(X_t\leq l\}\) and \(Z=l-X_\tau\) be the first hitting time and the undershoot. The main purpose of the paper is to derive the joint Laplace transform \[ E_x \exp(-\theta \tau -\zeta Z),\quad \theta,\;\zeta\geq 0, \quad x\geq l. \] To solve this problem the authors describe the partial (i.e. with respect to \(\theta\)) eigenfunctions of the generator \(\mathcal{A}\), corresponding to the process \(X\): \[ \mathcal{A} f_{\theta,\zeta}(x)=\theta f_{\theta,\zeta}(x),\quad x\geq l, \] and \(f_{\theta,\zeta}(x)=C_{\theta,\zeta}\exp(-\zeta(l-x))\) for \(x<l\). Then, by martingale arguments, \(E_x \exp(-\theta \tau -\zeta Z)\) can be expressed in terms of \(f_{\theta,\zeta}(x)\). Further, the extension for the case when the driving Lévy process is the sum of compound Poisson and an independent Brownian motion is given. The results are illustrated by many interesting examples.