On isomorphisms between Hardy spaces on complex balls (Q1823343)

The paper centers around the famous result that the Hardy space \(H^ p(B)\) on the unit ball B in \({\mathbb{C}}^ n\) \((n>1)\) is isomorphic to the Hardy space \(H^ p(D)\), \(1\leq p<\infty\), on the unit disk D in \({\mathbb{C}}\). Indeed, when \(1<p<\infty\) it was shown by \textit{A. B. Aleksandrov} [Math. USSR, Sb. 46, 143-159 (1983); translation from Mat. Sb., Nov. Ser. 118(160), 147-163 (1982; Zbl 0503.32001)] and by \textit{B. Wojtaszczyk} [Ann. Math., II. Ser. 118, 21-34 (1983; Zbl 0546.32003)] that \(H^ p(B)\) is isomorphic to \(L^ p([0,1])\) and hence to \(H^ p(D)\). The case of \(p=1\), however, is much harder, and the proof that \(H^ 1(B)\) is isomorphic to \(H^ 1(D)\) appeared first in the above mentioned paper of Wojtaszczyk. Unfortunately, a crucial argument in the latter contains an error, and the aim of the paper under review is to present a new proof of this fact. Part of Wojtaszczyk's proof was based on the intuitive but erronous fact that non-isotropic distances on \(\partial B\) and on \(R^{2n-1}\) give locally equivalent metrics. A proof of this non-equivalence is also contained in the present paper. To prove that \(H^ 1(B)\) is isomorphic to \(H^ 1(D)\), the author uses a probabilistic approach developed by \textit{B. Maurey} [Acta Math. 145, 79- 120 (1980; Zbl 0509.46045)]. In this approach the standard Brownian motion is replaced by a diffusion associated with the invariant Laplacian in B. This gives an isomorphic embedding of \(H^ 1(B)\) onto a complemented subspace of a martingale \(H^ 1\)-space which is isomorphic to a complemented subspace of \(H^ 1(D)\). One then is in a position to exploit the first part of Wojtaszczyk's proof, which shows that \(H^ 1(D)\) is isometric to a complemented subspace of \(H^ 1(B)\), and hence the desired isomorphism follows by the decomposition principle.