A strong Szegö-Widom Limit Theorem for operators with almost periodic diagonal (Q610747)

The classical Szegő-Widom Limit Theorem describes the asymptotic behavior of the determinants of finite sections of Toeplitz operators, namely those operators which have constant entries along each diagonal. In particular, for certain classes of functions, this theorem states that for a matrix function \(a(t)\) such that \(\det a(t)\neq 0\) for all \(t\in\mathbb{T}\) and such that \(\det a(t)\) has winding number \(0\), then for \[ G[a]=\exp\left(\frac{1}{2\pi}\int_{0}^{2\pi}\log\det a(e^{ix})\,dx\right) \] we have \[ \lim_{n\to\infty}\frac{\det T_n(a)}{G[a]^n}=\det T(a)\det T(a^{-1}). \] The authors of this paper seek to generalize these results to operators which have almost periodic sequences as their diagonals. To state the main result of the paper, we first recall some definitions. First, recall that the Toeplitz operator \(T(a)\) is the operator acting on \(\ell^2(\mathbb{Z}^+)\) that has a matrix representation given by \(T(a)_{j,k}=a_{j-k}\) for \(j,k\in\mathbb{Z}^+\). The finite Toeplitz matrices are the finite sections of the Toeplitz operators \(T_n(a)=P_nT(a)P_n\), where \(P_n\) is the projection from \(\ell^2(\mathbb{Z}^+)\) onto the first \(n\) entries. Also, let \(P\) denote the standard Riesz projection in \(\ell^2(\mathbb{Z})\), namely the projection that takes an arbitrary sequence \(a_n\) to the one supported only on the non-negative integers, i.e., \((Pa)_n=a_n\) when \(n\geq 0\) and \((Pa)_n=0\) when \(n<0\). Recall that the set \(AP\) of almost periodic sequences consists of all \(a\in\ell^\infty(\mathbb{Z})\) for which the set \[ \{U_na:n\in\mathbb{Z}\} \] is relatively compact in the norm topology of \(\ell^\infty(\mathbb{Z})\). Here, \((U_n(a))_k=a_{k-n}\), namely \(U_n\) is the shift by \(n\) on the space \(\ell^\infty(\mathbb{Z})\). Let \(\Xi\) be an additive subgroup of \(\mathbb{R}/\mathbb{Z}\). An admissible weight \(\beta:\Xi\to\mathbb{R}^+\) is a function for which \[ 1\leq\beta(\xi_1+\xi_2)\leq\beta(\xi_1)\beta(\xi_2). \] For any such admissible weight \(\beta\), the class \(APW(\mathbb{Z},\Xi,\beta)\) is the set of sequences from \(\ell^\infty(\mathbb{Z})\) of the form \[ a=\sum_{\xi\in\Xi} a_{\xi}e^{2\pi i\xi n}, \] where \[ \left\| a\right\|_{\Xi, \beta}=\sum\beta(\xi)\left| a_\xi\right|<\infty. \] Under the conditions on the weight \(\beta\), this is a Banach algebra. Given two numbers \(\alpha_1, \alpha_2\geq 0\), define \(\mathcal{R}\) to be the set of all operators acting on \(\ell^2(\mathbb{Z})\) which are of the form \[ A=\sum_{k\in\mathbb{Z}} a_{k} U_k \] such that \[ \left\| A\right\|_{\mathcal{R}}=\left\| a_0\right\|_{APW}+\sum_{k=1}^\infty\left(\left(1+k\right)^{\alpha_1}\left\| a_k\right\|_{APW}+\left(1+k\right)^{\alpha_2}\left\| a_{-k}\right\|_{APW}\right)<\infty. \] A strictly increasing sequence of non-negative integers \(h\) is called distinguished for \(\Xi\) if \[ \lim_{n\to\infty} e^{2\pi i\xi h_n}=1. \] Finally, call a weight \(\beta\) is compatible on \(\Xi\) if \[ \inf_{\xi\in\Xi, \xi\neq 0}\beta(\xi)\left\| \xi\right\|_{\mathbb{R}/\mathbb{Z}}>0. \] One of the main results by the authors is then the following Theorem. Assuming the notation from above and that \(\alpha_1,\alpha_2>0\) and \(\alpha_1+\alpha_2=1\), let \(h\) be a distinguished sequence for \(\Xi\). Then for \(A_1,\dots, A_r\in\mathcal{R}\) and \(A=e^{A_1}\cdots e^{A_r}\), we have \[ \lim_{n\to \infty}\frac{\det\left(P_{h_n}PAPP_{h_n}\right)}{\exp\left(\text{trace}\left(P_{h_n}P(A_1+\cdots A_r)PP_{h_n}\right)\right)}=\det\left(PAPA^{-1}P\right). \] Another result of the authors is the following. Let \(\beta\) be an admissible and compatible weight on \(\Xi\) and suppose that \(h\) is a distinguished sequence. Then, for \(a\in APW\), letting \(M(a)\) denote the mean of the sequence, we have as \(n\to\infty\) that \[ \sum_{k=0}^{h_n-1}a_k=h_nM(a)+o(1). \] Combining these two theorems then gives the desired extension of the Strong-Szegő-Widom Limit Theorem, \[ \lim_{n\to\infty}\frac{\det\left(P_{h_n}PAPP_{h_n}\right)}{G^{h_n}}=\det\left(PAPA^{-1}P\right), \] where \(G=\exp(M(a))\). While the paper is very technical, the authors do a good job of motivating their results and provide examples to show that the conditions imposed on the weights \(\beta\) are reasonable.