3
$\begingroup$

Bernstein's theorem states that for any completely monotone function $f$: $f \in C^{\infty}[0,+\infty)$, $(-1)^n f^{(n)}(t) \geqslant 0$ there is a finite Borel measure $\mu$ such that $$ f(t) = \int_{0}^{+\infty} e^{-tx} \mu(dx) $$

Is there some generalisation of this result on the case of $n$ dimensions?

$\endgroup$

2 Answers 2

4
$\begingroup$

Yes, there is and it goes by the name of Bochner theorem. For details see the book of A. Klenke: Probability Theory. A Comprehensive Course, Springer Verlag, 2008.

$\endgroup$
4
  • 1
    $\begingroup$ Another useful reference is Section 4.2 of Bochner's book "Harmonic analysis and the theory of probability" (which has been reprinted by Dover). $\endgroup$
    – Henry Cohn
    Jan 24, 2012 at 13:11
  • $\begingroup$ That;'s a bit of good news.I did not know it was reprinted. $\endgroup$ Jan 24, 2012 at 13:18
  • $\begingroup$ Thanks, but I rather looked for the next generalisation: is there some similar theorem that characterises functions $f \colon \mathbb{R}^n \to \mathbb{R}$ such that $(-1)^{|\alpha|} \frac{ \partial^{\alpha} }{ \partial^{\alpha_1} \partial^{\alpha_2} \cdots \partial^{\alpha_n} } f(x) \geqslant 0$ ? $\endgroup$
    – Appliqué
    Jan 24, 2012 at 15:46
  • 2
    $\begingroup$ See Theorem 4.2.1 in Bochner's book. $\endgroup$
    – Henry Cohn
    Jan 24, 2012 at 15:52
5
$\begingroup$

Further generalizations using the framework of semigroups with involution can be found in the nice book: Harmonic Analysis on Semigroups: theory of positive definite and related functions, by Christian Berg, Jens Peter Reus Christensen, Paul Ressel.

There, the general framework yields the Bernstein-Widder characterization, Bochner's theorem, etc., all as special cases of a more general setup. (The book is quite accessible after Chapter 3).

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.