Ergodic Jacobi matrices and conformal maps (Q694502)

This paper centers around a general abstract analysis of the fundamental quantities used in the spectral theory of ergodic Jacobi matrices. Starting with an invariant (with respect to a suitably defined shift) probability measure \(\mu\) on \(\mathcal{J}_2\), the set of Jacobi matrices with operator norm \(\|J\|\leq 2\), one can introduce a number of useful objects. In the special case of ergodic systems of Jacobi matrices, these objects are of fundamental importance. In the paper under review, these quantities are introduced and investigated under the more general case of invariant measures. By averaging the spectral measures \(d\rho_0(\,\cdot\,;J)=d\|E_J(\,\cdot\,)\delta_0\|^2\) with respect to \(\mu\), one obtains the density of states measure \(dk\). One can then define the quantity \(A>0\) by the requirement \[ \ln(A) = \int_{\mathcal{J}_2}\ln(a_0(J))\, d\mu(J), \] where \(a_0(J)\) denotes the \((0,1)\) entry of the Jacobi matrix \(J\). In turn, one defines \[ w(z) = \ln(A) - \int_{[-2,2]}\ln(t-z)\, dk(t),\quad z\in \mathbb{C}_+, \] and the logarithm is chosen with \(\Im \{\ln (\zeta)\}\in (-\pi,0)\), which makes \(w\) a Herglotz function. In turn, one defines \(\gamma(z)=-\Re(w(z))\). The function \(\gamma(\,\cdot\,)\) is harmonic on \(\mathbb{C}_+\) and is called the Lyapunov exponent. Going further, one can use the change of variables \(\zeta\mapsto z(\zeta)=-\zeta-\zeta^{-1}\) to define a function \(F\) which is holomorphic on the unit disk \(D\) by setting \[ F(\zeta)=e^{w(z(\zeta))}, \quad \zeta \in D_+, \] and \(D_+=\mathbb{C}_+\cap D\). One then extends \(F\) to all of \(D\) by reflection \(F(\overline{\zeta})=\overline{F(\zeta)}\). The image of \(F\), that is, \(\Omega=F(D)\), is called the image domain and has an associated slit height function \(h\). The authors study the properties of \(A\), \(dk\), \(w\), \(\gamma\), \(F\), \(\Omega\), and \(h\) extensively in the present paper. Given \((A,dk)\), the associated quantities \(w\), \(\gamma\), \(F\), \(\Omega\), and \(h\) are computed as outlined above. On the other hand, for example, the authors show that one of the former is enough to determine \((A,dk)\), and even the invariant measure \(\mu\).