Counterexamples to two variants of the Helson-Szegö theorem. (Q1411680): Difference between revisions

Let \(D\) denote the unit disk, \(T\) its boundary, and \(h_z\) the harmonic measure on \(T\) for the point \(z\in D.\) Throughout this review \(\phi\) will denote a real valued function in \(L^1(T).\) From theorems of Helson-Szegö, Hunt-Muckenhoupt-Wheeden and others, one can show that the following three statements about \(\phi\) are equivalent: (1) \(\phi = u + \tilde v,\;\) where \(u\in L^{\infty}(T),\;\| v\| _{\infty} < \pi/2,\) and \(\,\tilde{}\,\) denotes conjugate function. (2) \( \sup_{z\in D} \int_T \exp(| \phi-\phi(z)| )\,dh_z < \infty.\) (3) The Toeplitz operator with symbol \(\; e^{i\tilde{\phi}} \;\) is invertible. The reviewer suggested that perhaps the equivalence (1) iff (2) would remain true if in (1) we replace \( < \pi/2\) with \(\leq \pi/2\) and in (2) replace the condition stated there with a weak version: \[ \sup_{z\in D} \sup_{\lambda > 0} e^{\lambda} h_z(\{e^{i\theta}: | \phi(e^{i\theta}) - \phi(z)| > \lambda\}) < \infty. \] Let \(g\) denote the harmonic extension of \(e^{i\tilde{\phi}}\) to \(D.\) R. Douglas asked if boundedness of \(g\) away from zero would imply the invertibility condition (3). If true, this would provide a relatively simple criterion for deciding whether (1) and (2) are true. In the article under review the author constructs a class of functions \(\phi,\) each of which shows that the suggested results are false. Each \(\phi\) is the logarithm of the limit of a martingale. It is easy to write down a formula for \(\phi(e^{i\theta})\) in terms of base \(b\) expansions of \(| \theta/\pi| ,\) where \(b\) is a large positive integer, but to prove that \(\phi\) has the required properties requires difficult and ingenious analysis. This paper was written in 1983. The results became well-known to specialists in the subject, but Professor Wolff did not publish it. After his tragic death in 2000 the paper was prepared for publication by John Garnett.