Local time-space stochastic calculus for Lévy processes (Q2495381)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Local time-space stochastic calculus for Lévy processes
scientific article

    Statements

    Local time-space stochastic calculus for Lévy processes (English)
    0 references
    0 references
    30 June 2006
    0 references
    Let \(X_{t}\) be a Lévy process, with local times \(L_{t}^{x}\), \(x\in {\mathbb R}\), \(t\geq 0\), \(X\) having a Brownian component \(\alpha B_{t}\) and with \(\sum_{0\leq s\leq t}| X_{s}-X_{s- }| <\infty\). Let \(X^{\ast}_{t}=X_{(1-t)-}\), \(X^{\ast}_{1}=0\). For \(f:{\mathbb R}\times [0,1]\to {\mathbb R}\), let \(\| f\| =2E(\int_{_{0}}^{^{1}} f^{2}(X_{s},s)ds)^{1/2}+E(\int_{_{0}}^{^{1}}| s^{- 1}f(X_{s},s) B_{s}| ds)\). Then the integral \[ \int_{_{0}}^{^{1}}\int_{_{R}}1_{(x,y]\times(s,t]}dL_{s}^{x}=L_{t}^{y}-L_{s} ^{y}-L_{t}^{x}+L_{s}^{x} \] can be extended linearly to the set \(I\) of all \(f\) with \(\| f\| <\infty\), and it satisfies \(E(| \int fdL| )\leq\| f\|\) and \(\int_0^t \int_R fdL=\alpha \int_0^t f(X_{s-},s)dB_s +\alpha \int_{1-t}^1 f(X^{\ast}_{s-},1-s)dB^{\ast}_s\), \(t\in [0,1]\). If, for \(F\) defined on \({\mathbb R}\times [0,\infty )\), \(\partial F/\partial x\) and \(\partial F/\partial t\) exist as Radon-Nikodým and are locally bounded, then \[ \begin{aligned} F(X_{t},t)=F(X_{0},0) & +\int_0^t (\partial F/\partial t)(X_{s},s)ds +\int_0^t (\partial F/\partial x)(X_{s-},s)dX_{s}\\ & + \sum_{0<s\leq t}(F(X_{s},s)-F(X_{s-},s) -(\partial F/\partial x)(X_{s-},s) (X_{s}-X_{s-})) \\ & -\frac12 \int_0^t \int_R (\partial F/\partial x)(x,s)dL_{s}^{x}.\end{aligned} \] If \(b:[0,1]\to {\mathbb R}\) is measurable, \(\int_0^1 \int_R 1_{(-\infty,b(s))}(x)dL_{s}^{x}\) is denoted by \(L_{t}^{b(\cdot )}\). If \(b\) is continuous, \(F\) is continuous, \(F\) is \(C^{2,1}\) on the closure of \((x<b(s))\) and on that of \((x>b(s))\) and \(\partial F/\partial x\) exists Radon-Nikodým, denoted \(\partial F_{1}/\partial x\) on \((x<b(s))\) and to \(\partial F_{2}/\partial x\) on \((x\geq b(s))\), then the author establishes an expression for \(F(X_{t},t)\), differing from the previous one by the fact that the last two terms are replaced by \[ \begin{multlined} \frac12 \int_0^t ((\partial^{2}F_{1}/\partial x^{2}) (X_{s},s)1_{(x<b(s))} +(\partial^{2}F_{2}/\partial x^{2})(X_{s},s)1_{(x\geq b(s))})d[X]_{s}^{c}\\+ \frac12 \int_0^t (\partial F_{2}/\partial x-\partial F_{1}/\partial x)(b(s),s) d_{s} L_{s}^{b(\cdot )}.\end{multlined} \] Also a multidimensional generalization of the first formula for \(F(X_{t},t)\) is proved. Two applications, to functions \(F\) of the type in the second \(1\)-dimensional formula, are given.
    0 references
    0 references
    Lévy processes
    0 references
    Brownian component
    0 references
    local times on curves
    0 references
    variant of stochastic calculus
    0 references
    Itô formula
    0 references
    0 references