Norm bounds for rational matrix functions (Q794394): Difference between revisions

From MaRDI portal
RedirectionBot (talk | contribs)
Removed claim: author (P16): Item:Q369399
ReferenceBot (talk | contribs)
Changed an Item
 
(2 intermediate revisions by 2 users not shown)
Property / author
 
Property / author: Bernhard A. Schmitt / rank
 
Normal rank
Property / MaRDI profile type
 
Property / MaRDI profile type: MaRDI publication profile / rank
 
Normal rank
Property / cites work
 
Property / cites work: Numerical range of a matrix: Some effective criteria / rank
 
Normal rank
Property / cites work
 
Property / cites work: An implementation of singly-implicit Runge-Kutta methods / rank
 
Normal rank
Property / cites work
 
Property / cites work: Order and stepsize control in extrapolation methods / rank
 
Normal rank
Property / cites work
 
Property / cites work: Iterated defect correction for the efficient solution of stiff systems of ordinary differential equations / rank
 
Normal rank
Property / cites work
 
Property / cites work: On the stability of semi-implicit methods for ordinary differential equations / rank
 
Normal rank
Property / cites work
 
Property / cites work: High-Accuracy Stable Difference Schemes for Well-Posed Initial-Value Problems / rank
 
Normal rank
Property / cites work
 
Property / cites work: Semidiscretization in Time for Parabolic Problems / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q3673995 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q4133641 / rank
 
Normal rank

Latest revision as of 12:06, 14 June 2024

scientific article
Language Label Description Also known as
English
Norm bounds for rational matrix functions
scientific article

    Statements

    Norm bounds for rational matrix functions (English)
    0 references
    1983
    0 references
    In the solution of a linear differential equation \(y'=Ay,\) it is appropriate to consider bounds of the forms \(\| f(A)\| \leq 1\) and \(\| f(hA)\| \leq 1+ch\) for all \(0<h\leq h_ 0\), where f is an approximation to the exponential function characteristic of a particular numerical method. The constants c and \(h_ 0\) should not be dependent on the norm of the matrix A, which corresponds to an arbitrarily stiff problem. Let \(G_ A\) denote the numerical range, that is the set of possible values of the Rayleigh quotient, of A. The main result of the paper deals with rational functions of the form \(f(z)=\prod^{s}_{j=1}(1-z/w_ j)/\prod^{s}_{k=1}(1-z/z_ k)\) in which the poles and zeros can be so ordered that \(| w_ k| \geq | z_ k|\) and \(Re(1/w_ k)<Re(1/z_ k)\) for \(k=1,2,...,s\). In this case it is shown how to construct a stability region S such that \(\| f(A)\| \leq 1\) whenever \(G_ A\subset S\). Furthermore, it is shown how to obtain a stability angle \(\alpha\) and constants c, \(\zeta_ 0>0\) such that for \(G_ A\subset Z_{\alpha}+\beta\), with \(Z_{\alpha}=\{z\in {\mathbb{C}}:| \arg(-z)| \leq \alpha \}\) and \(\beta \in {\mathbb{R}}\), \(\| f(hA)\| \leq 1+h\beta c\quad for\quad h\beta \leq \zeta_ 0.\) The results are applied to Padé approximations and to implicit Runge-Kutta methods based on equidistant knots.
    0 references
    norm bounds
    0 references
    rational matrix functions
    0 references
    spectral norm
    0 references
    stiff problem
    0 references
    Rayleigh quotient
    0 references
    stability region
    0 references
    implicit Runge-Kutta methods
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references