Some remarks on first passage of Lévy processes, the American put and pasting principles (Q2572401)

Let \(X= (X_t)\) be a Lévy process defined on a filtered probability space \((\Omega,{\mathcal F},({\mathcal F}_t),P)\). For given constants \(K> 0\) and \(r\geq 0\) consider the following American put optimal stopping problem: \[ v(x)= \sup_\tau\,E_x[e^{-r\tau}(K- e^{X_\tau})^+]\tag{*} \] (the supremum taken over all \(({\mathcal F}_t)\)-stopping times \(\tau\)). Various authors [e.g. \textit{E. Mordecki}, Finance Stoch. 6, No. 4, 473--493 (2002; Zbl 1035.60038)] have shown that for many Lévy processes (*) is solved by a strategy of the form \(\tau^*= \text{inf}\{t\geq 0: X_t< x^*\}\) for a specific value \(x^*<\log K\) so that \[ v(x)= KE_x[e^{-r\tau^*}]- E_x[e^{-r\tau^*+ X_{\tau^*}}]. \] The aim of the present paper is to comprehensively link a variety of identities for first passage problems of different Lévy processes and their connection to the American put optimal stopping problem and to explain precisely when smooth pasting is absent. In particular, it is thus possible to give alternative proofs of results obtained in Mordecki's paper.