Zero location for nonstandard orthogonal polynomials (Q5959025): Difference between revisions
From MaRDI portal
Added link to MaRDI item. |
Set profile property. |
||
Property / MaRDI profile type | |||
Property / MaRDI profile type: MaRDI publication profile / rank | |||
Normal rank |
Revision as of 23:48, 4 March 2024
scientific article; zbMATH DE number 1722103
Language | Label | Description | Also known as |
---|---|---|---|
English | Zero location for nonstandard orthogonal polynomials |
scientific article; zbMATH DE number 1722103 |
Statements
Zero location for nonstandard orthogonal polynomials (English)
0 references
29 May 2003
0 references
The authors introduce an interesting method to derive bounds for the zeros of orthogonal polynomials with respect to an inner product of the form \[ \langle p,q\rangle = \sum_{k=0}^N \int p^{(k)}(z) {\overline q^{(k)}(z)}w_k(z) d\mu(z),\quad N\geq 1, \tag{a} \] where \(\mu\) is a finite positive Borel measure with support \(\Delta\subset {\mathbb C}\) (containing infinitely many points) and \(w_k\in L^1(\mu), w_k\geq 0, 0\leq k\leq N\). The main result is: 1. Assume \(w_k/w_{k-1}\in L^{\infty}(\mu)\) and write \(C_k=||w_k/w_{k-1}||_{\infty}\) \((1\leq k\leq N)\); if \(z_0\) is a zero of an orthogonal polynomial with respect to (a), then \[ d(z_0,\text{Co}(\Delta))\leq {1\over 2}\sqrt{\sum_{k=1}^N k^2C_k} \] (\(\text{Co}(\Delta)\) is the convex hull of \(\Delta\)). Moreover, the bound is sharp. 2. If, furthermore, \(0<\alpha_i, \beta_i< 2\) satisfy \(\alpha_i+\beta_i=2\) (\(1\leq i\leq N-1\)), \(\alpha_0=\beta_N=2\), then \[ d(z_0,\text{Co}(\Delta))\leq\max\left\{{k\sqrt{C_k}\over \sqrt{\alpha_{k-1}\beta_k}}, k=1,\ldots,N\right\}. \] The authors also study sequences of polynomials \(\{p_n\}_n\) satisfying a \((2N+3)\)-term recurrence relation \[ t^{N+1}p_n(t)=c_{n,0}p_n(t)+\sum_{k=1}^{N+1} \left[c_{k,k}p_{n-k}(t) + c_{n+k,k}p_{n+k}(t)\right], \] with initial conditions \(p_k=0\) \((k<0)\) and \(p_k(t)\) a polynomial of degree \(k\) for \(k=0,1,\ldots,N+1\). These are orthogonal with respect to \[ \langle p,q\rangle = \int (T_0(p),\ldots T_N(p))W(z)(T_0(q),\ldots,T_N(q))^{*}d\mu(z),\tag{b} \] where \[ T_m(p)=\sum_n {p^{(n(N+1)+m)}(0)\over (n(N+1)+m)!}t^n \] and \(\mu\) a positive measure supported on \({\mathbb R}\). Restricting themselves to the case \(W=\text{diag}(w_0,\ldots,w_N)\) and \(\mu\) a finite positive Borel measure with compact support in the complex plane, the authors then prove 3. Let the \(w_k\) be as before, \(|z|^2w_0(z)/w_N(z)\in L^{\infty}(\mu)\) and define \(C_0= |||z|^2w_0(z)/w_N(z)||_{\infty}\); if \(z_0\) is a zero of an orthogonal polynomial with respect to (b), then \[ |z_0|\leq\max\{\sqrt{C_0},\sqrt{C_1},\ldots,\sqrt{C_N}\}. . \]
0 references
zeros
0 references
higher-order recurrence relation
0 references
matrix polynomials
0 references
Sobolev inner product
0 references