Nevanlinna and Smirnov classes on the upper half plane (Q1185417)

Let \(\mathbb{U}\) be the open disc \(| z|<1\), and suppose \(f(z)\) is holomorphic there, in symbols, \(f\in{\mathcal O}(\mathbb{U})\). The function \(f(z)\) belongs to the Nevanlinna class \(N(\mathbb{U})\) if the nonnegative subharmonic function \(\log(1+| f(z)|)\) has a harmonic majorant (on \(\mathbb{U}\)). Alternatively, \(f\in N(\mathbb{U})\) if \(f(z)\) is the quotient of bounded holomorphic functions (with denominator vanishing nowhere in \(\mathbb{U}\)). Like the more studied \(H^ p(\mathbb{U})\), \(0<p\leq\infty\), the Nevanlinna class carries an invariant metric. In the metric, the distance between \(f\) and \(g\) is \[ d(f,g)\overset\text{def}= \lim_{r\to 1} {1\over 2\pi} \int_ 0^{2\pi} \log(1+| f(re^{i\theta})- g(re^{i\theta})|) d\theta. \] Unlike the distance from \(f\) to 0 in the \(H_ p(\mathbb{U})\) metric, \(d(f,0)\) usually cannot be calculated using only the boundary values of \(f\). If it can, the function \(f(z)\) belongs to the Smirnov class \(N_ *(\mathbb{U})\). This is to say, \(f\in N_ *(\mathbb{U})\) if \(f\in N(\mathbb{U})\) and \[ d(f,0)={1\over2\pi}\int_ 0^{2\pi}\log(1+| f(e^{i\theta})| d\theta. \] Alternatively, \(f\in N_ *(\mathbb{U})\) if once more \(f(z)\) is the quotient of bounded holomorphic functions, but now with denominator outer. In recent years, \textit{J. H. Shapiro} and \textit{A. L. Shields} [Am. J. Math. 97 (1975), 915- 936 (1976; Zbl 0323.30033)], \textit{J. W. Roberts} [Ill. J. Math. 19, 553- 559 (1975; Zbl 0313.30033)] and \textit{K. Stephenson} [Indiana Univ. Math. J. 26, 307-324 (1977; Zbl 0326.30025)] have studied the geometry of the Nevanlinna and Smirnov classes. For example, Stephenson shows that the only isometries of \(N_ *(\mathbb{U})\) are the obvious ones: a rotation of the circle followed by multiplication by a unimodular constant. In the paper here, Nevanlinna and Smirnov classes on the upper half plane, Mochizuki replaces the disc \(| z|<1\) with the half-plane \(y>0\), which he denotes by \(D\). However, he studies not \(N(D)\), but a somewhat different Nevanlinna class which he calls \(N_ 0(D)\). To describe this class, we need the Poisson kernel of the half-plane \(y>0\). This is the positive harmonic function \[ P(z)=-{1\over\pi}{\mathfrak J}{1\over z}={1\over\pi} {y\over {x^ 2+y^ 2}}. \] Suppose \(f\in{\mathcal O}(D)\). The function \(f\) belongs to \(N_ 0(D)\) if the line \(y=0\) carries bounded measure \(\mu\) whose convolution with \(P\) bounds \(\log(1+| f|)\): \[ \log(1+| f(z)|)\leq(P*\mu)(z)=\int_{-\infty} ^ \infty P(z-t)d\mu(t). \] If \(\mu\) is absolutely continuous with respect to linear measure, Mochizuki writes \(f\in N_ *(D)\). Let \(\varphi\) be the usual map of the disc \(\mathbb{U}\) onto the half-plane \(D\), and suppose \(f\in N_ 0(D)\). Then \(f\in N_ *(D)\) iff \(f(\varphi)\in N_ *(\mathbb{U})\). The proof uses work of V. I. Krylov. The Nevanlinna class \(N_ 0(D)\) like \(N(\mathbb{U})\), carries an invariant metric. The distance between \(f\) and \(g\) is now \[ d(f,g)\overset\text{def} = \lim_{y\to 0} \int_{- \infty} ^ \infty \log(1+| f(x+iy)-g(x+iy)|)dx. \] Mochizuki's Smirnov class may be described in terms of the metric on his larger Nevanlinna class: \(f\in N_ *(D)\) iff \(f\in N_ 0(D)\) and \[ d(f,0)=\int_{-\infty} ^ \infty \log(1+| f(x)|)dx. \] Mochizuki shows that \(N(D)\) and its subalgebra \(N_ *(D)\) have properties like those found earlier for \(N(\mathbb{U})\) and \(N_ *(\mathbb{U})\) by Shapiro and Shields, Roberts, and Stephenson. For example, he shows that the only isometries of \(N_ *(D)\) are the obvious ones: a translation of the line followed by multiplication by a unimodular constant. He also shows that neither \(N(\mathbb{U})\) and \(N_ 0(D)\) nor \(N_ *(\mathbb{U})\) and \(N_ *(D)\) are isometric.