The space-fractional Poisson process (Q434734): Difference between revisions

For the probabilities \(p_{k}(t)=\operatorname{P}[N(t)=k]\) of a renewal process, let us consider the counting number \(N(t)\) as varying in space and \(t\) as usual denoting time. The classical homogeneous Poisson process is characterized by a system of difference-differential equations \(D_{t} p_{k} (t) = -\lambda (1 - B)p_{k} (t))\), \(\lambda > 0\), \(t \geq 0\), \(k = 1, 2, 3,\dots\), with \(D_t\) denoting first-order differentiation, \(B p_{k} = p_{k-1}\) (backward shift). Now, assume \(0 < \alpha \leq 1\), \(0 < nu \leq 1\). Replacing \(D_t\) by the Caputo time-derivative of order \(\nu\), one obtains the time-fractional Poisson process, whereas replacing \(\lambda (1-B)\) by \(\lambda^\alpha (1-B)^\alpha\) gives us the space-fractional Poisson process. Both replacements applied simultaneously lead to the space-time fractional Poisson process. The appropriate initial conditions are \(p_{0} (0)=1\), \(p_{k} (0)=0\) for \(k > 0\). The authors compare these processes to each other, present formulas and generating functions for the state probabilities \(p_{k} (t)\) and exhibit a relation of subordination for the space-fractional process involving the \(\gamma\)-stable subordinator, \( \gamma \in (0, 1]\).