[Paper Review] Differentiation matrices for meromorphic functions

Rafael G. Campos, Claudio Meneses|ArXiv.org|Jul 1, 2004

Meromorphic and Entire Functions8 references7 citations

TL;DR

This paper extends differentiation matrices to meromorphic functions using complex Hermite interpolation, enabling accurate numerical differentiation of rational and elliptic functions in the complex domain. The method achieves spectral convergence for derivatives of Jacobi and Weierstrass elliptic functions, and successfully approximates solutions to singular differential equations like Kummer's equation with high accuracy using $N=21$ nodes.

ABSTRACT

A procedure to obtain differentiation matrices is extended straightforwardly to yield new differentiation matrices useful to obtain derivatives of complex rational functions. Such matrices can be used to obtain numerical solutions of some singular differential problems defined in the complex domain. The potential use of these matrices is illustrated with the case of elliptic functions.

Motivation & Objective

• To extend differentiation matrices to meromorphic functions, particularly rational and elliptic functions, in the complex domain.
• To develop a numerical method for solving singular differential equations with poles using interpolation-based differentiation matrices.
• To demonstrate the method's effectiveness on periodic and doubly periodic meromorphic functions such as Jacobi and Weierstrass elliptic functions.
• To validate the approach on Kummer’s differential equation using eigenvalue formulation and spectral convergence.

Proposed method

• Uses complex Hermite interpolation to construct differentiation matrices that yield exact derivatives for meromorphic functions with poles.
• Applies the residue theorem to derive the matrix entries from interpolation nodes, ensuring exactness for rational functions of degree ≤ N−1.
• Employs trigonometric and algebraic polynomial bases with nodes chosen on complex contours or rays to capture periodicity and singular behavior.
• Constructs the differentiation matrix $ D $ as a $ N \times N $ matrix whose entries are derived from the logarithmic derivative of the nodal polynomial $ \omega(z) $.
• Solves singular differential equations by transforming them into discrete eigenvalue problems using $ L = ZD^2 + (bI - Z)D $, where $ Z $ is a diagonal matrix of nodes.
• Normalizes vectors and compares approximate solutions with exact confluent hypergeometric function values to assess accuracy.

Experimental results

Research questions

• RQ1 Can differentiation matrices be generalized to meromorphic functions with poles, beyond analytic or polynomial functions?
• RQ2 How accurately can differentiation matrices approximate the derivatives of doubly periodic meromorphic functions like Weierstrass ℘-functions?
• RQ3 What is the convergence behavior of the method when applied to singular differential equations such as Kummer’s equation?
• RQ4 Can the eigenvalue formulation of the discrete operator recover the correct parameters of special functions like $ M(a,b,z) $?

Key findings

• For the Jacobi elliptic function, the error norm decreased from $ 1.6 \times 10^{-4} $ at $ N=10 $ to $ 1.0 \times 10^{-8} $ at $ N=20 $, showing spectral convergence.
• For the Weierstrass ℘-function, the error norm decreased from $ 1.5 \times 10^{-5} $ at $ N=10 $ to $ 1.0 \times 10^{-8} $ at $ N=20 $, confirming high-order accuracy.
• The Kummer equation solution using $ N=21 $ nodes on the ray $ z_k = 5(1+i)k/N $ achieved absolute errors of $ 0.0675659 $ for $ b=5/2 $ and $ 0.0426948 $ for $ b=3+2i $, with high agreement in real and imaginary parts.
• The method successfully recovers the confluent hypergeometric function $ M(a,b,z) $ as an eigenvector of the discrete operator, with the eigenvalue $ \lambda $ approximating the parameter $ a $.
• The differentiation matrix acts as a nonlocal operator, computing derivatives via function values at $ N $ distinct points, even though the derivative is a local operation.
• The approach maintains high accuracy even for functions with poles, demonstrating robustness for singular problems in the complex domain.