Fourier-Padé approximants for Nikishin systems (Q836081)

Let \(\Delta_0,\dots,\Delta_m \subset \mathbb{R}\) be intervals satisfying \(\Delta_j \cap \Delta_{j+1} = \emptyset\), and let \(\sigma_1,\dots,\sigma_m\) be finite Borel measures with constant sign in \(\Delta_j\), that is the smallest interval containing supp(\(\sigma_j\)). The Nikishin system of measures \((s_1,\dots,s_m) = {\mathcal N}(\sigma_1,\dots,\sigma_m)\) generated by them consists of the following: \[ s_1 = \left\langle \sigma_1 \right\rangle = \sigma_1, \; s_2(x) = \left\langle \sigma_1, \sigma_2 \right\rangle(x) = \int \frac{d \sigma_2(t)}{x-t} d \sigma_1(x) = \hat{\sigma_2}(x) d \sigma_1(x), \dots, \] \[ s_m = \left\langle \sigma_1, \dots, \sigma_m \right\rangle = \left\langle \sigma_1, \left\langle \sigma_2, \dots, \sigma_m \right\rangle \right\rangle. \] Let \((\hat s_1,\dots, \hat s_m)\) be the corresponding Nikishin system of Markov functions. Moreover, a sequence \(\{ l_j \}_{j \in \mathbb{N}_0}\) of orthonormal polynomials w.r.t. to \(\sigma_0\) and a finite Borel measure with constant sign and bounded support consisting of infinitely many points contained in \(\Delta_0\) are considered. For \(n = (n_0,\dots, n_m) \in (\mathbb{N}_0)^{m+1}\) and \(|n| := \sum_{j=0}^m n_i\), a system of polynomials \(A_{n,0},\dots, A_{n,m}\) with (i) deg(\(A_{n,j}\)) \(\leq n_j - 1\), not all identically 0, (ii) for \(k=0,\dots,|n|-2\) \[ \int \left( A_{n,0}(x) + \sum_{j=1}^m A_{n,j}(x) \hat{s_j}(x) \right)l_k(x) d\sigma_0(x)= 0, \] is called a type I Fourier-Padé approximant of \((\hat s_1,\dots, \hat s_m)\) w.r.t. \(n\). We abbreviate the function in brackets by \(L_{n,0}(x)\). Let \(\Lambda = \Lambda(p_0,\dots,p_m)\) be an infinite sequence of distinct multi-indices such that \[ \lim_{n \in \Lambda} \frac{n_j}{|n|} = p_j \in (0,1), \quad P_j := \sum_{k=j}^m p_k. \] Assuming \(\sigma_j \:(j=0,\dots,m)\) to be regular measures, the authors obtain \[ \lim_{n \in \Lambda} | L_{n,0}(z) |^{1/|n|} = \exp\left( P_1 V^{\bar \mu_1}(z) - V^{\bar \mu_0}(z) - 2 \sum_{k=1}^m \frac{\omega_k^{\bar \mu}}{P_k} \right) \] uniformly on each compact subset of \(\mathbb{C} \setminus (\Delta_0 \cup \Delta_1)\). In this expression, \(V\) denotes the logarithmic potential, \(\bar \mu = \bar \mu(\mathcal{C})\) = \(( \bar{\mu}_{0}, \dots, \bar{\mu}_{m} )\) the equilibrium vector measure, and \((\omega_0^{\bar\mu},\dots,\omega_m^{\bar\mu})\) the system of equilibrium constants for the vector potential problem determined by a given interaction matrix \(\mathcal{C}\) on the system of intervals \(\Delta_j\). Their construction is similar to type I Hermite-Padé approximation. Instead of power series expansions of functions in the system, they take their development in a series of orthogonal polynomials.