Unicité du plongement d'une mesure de probabilité dans un semi-groupe de convolution gaussien. Cas non-abélien (Q1056964)

The author proves the following remarkable theorem: Let G be a simply connected step two nilpotent Lie group and let \(\mu\) be a probability measure on G. Then there exists at most one Gaussian convolution semigroup \((\nu_ t)_{t>0}\) on G such that \(\nu_ 1=\mu.\) If \(G={\mathbb{R}}^ n\) this result is classical. In the present (non- Abelian) situation the proof relies on the following fact: the diffusion process \(\gamma\) associated with a Gaussian semigroup on G can be constructed by a certain recurrence procedure, cf. \textit{B. Roynette}, z. Wahrscheinlichkeitstheor. Verw. Geb. 32, 133-138 (1975; Zbl 0312.60036). Reviewer's remark: It would be interesting to have an answer to the following related question: Given a Gaussian measure \(\mu\) on G can there exists a non-Gaussian convolution semigroup \((\mu_ t)_{t>0}\) on G such that \(\mu_ 1=\mu\)?