A characterization of the finely harmonic morphism in \(R^ n\) (Q1091512)

This article proves the following Theorem: For a finely continuous mapping \(\phi\) from a domain U \((\subset {\mathbb{R}}^ n\), \(n\geq 2)\) into \({\mathbb{R}}^ m\) (m\(\geq 2)\), the following are equivalent: (i) The components \(\phi_ j\) of \(\phi\) (1\(\leq j\leq m)\) are finely harmonic in U, and \(\nabla \phi_ i \cdot \nabla \phi_ j=\delta_{ij} | \nabla \phi_ 1|^ 2\) a. e. on U, where \(\nabla \phi_ i\) denotes a generalized gradient. (ii) The components \(\phi_ j\) of \(\phi\) (1\(\leq i\leq m)\), \(\phi_ i\phi_ j\) (i\(\neq j)\) and \(\phi^ 2_ i-\phi^ 2_ j\) are finely harmonic in U. (iii) \(\phi\) is a finely harmonic morphism on U. (iv) for each \(x\in U\) and for each Brownian motion \(B(t,\omega)\) issued from x, \(\{\phi (B(t))\}_{0\leq t<\tau}\) is a diagonal martingale, where \(\tau\) is the first exit time of B(t) from U. (v) \(\phi\) preserves the paths of Brownian motion. Some closely related earlier articles should be mentioned: For the part (i) - (iii), see \textit{B. Fuglede} [Lect. Notes Math. 747, 123-132 (1979; Zbl 0414.53033) and Ann. Inst. Fourier 28, 107-144 (1978; Zbl 0339.53026)] and for the part (iii) - (v), see \textit{B. Øksendal} [Invent. Math. 75, 179-187 (1984; Zbl 0537.60083)].