A probabilistic proof of Schoenberg's theorem (Q2633838)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | A probabilistic proof of Schoenberg's theorem |
scientific article |
Statements
A probabilistic proof of Schoenberg's theorem (English)
0 references
10 May 2019
0 references
Let \(g:(0,\infty)\to(0,\infty)\) be a given function such that for every dimension \(k\in \mathbb{N}\), the function \(\mathbb{R}^k \ni \xi \mapsto g(|\xi|^2)\) is the characteristic function of an infinitely divisible probability measure on \(\mathbb{R}^k\). Then a classical result of Schoenberg states that \(f:= - \log g\) is a Bernstein function. The authors give a new probabilistic proof of this result and obtain some extensions of it. In particular, they show that it is sufficient to have the above condition for all odd dimensions \(k\) and they also show that it is necessary and sufficient that for some (equivalently, all) \(k\in \mathbb{N}\) there exists a \(k\)-dimensional Lévy process \((X_t^k)_{t\geq 0}\) with characteristic exponent \(\mathbb{R}^k \ni \xi \mapsto f(|\xi|^2)\) such that \(X_t^k\) has distribution \(\mathrm{e}^{-ct} \delta_0(\mathrm{d}x) + p_t(x) \, \mathrm{d}x\) for all \(t>0\), where \(\delta_0\) denotes the Dirac measure at \(0\) (in \(\mathbb{R}^k\)), \(c\in [0,\infty]\) is a constant and \(p_t :\mathbb{R}^k \to [0,\infty)\) is rotationally invariant such that \(p_t(\sqrt{\cdot})\) is completely monotone. The Lévy process \(X^k\) is then a Brownian motion time-changed by a subordinator with Laplace exponent \(f\). The authors obtain various corollaries of their results and proofs, among them a gradient estimate for the transition semigroup of a subordinated Brownian motion.
0 references
negative definite function
0 references
subordination
0 references
Lévy process
0 references
transition density
0 references
Hartman-Wintner condition
0 references
0 references