Numerical representations of the incomplete Gamma function of complex-valued argument (Q701929)

Various approaches to the numerical representation of the incomplete Gamma function \[ \gamma (m + 1/2, z)=2\,{z}^{m+1/2}\int \limits^{1}\limits_{0}t^{2m}e^{-zt^2}dt \] for complex arguments \(z\) and non-negative small integer indices \(m\) are compared with respect to numerical fitness (accuracy and speed). The authors consider power series, Laurent series, classical numerical methods of sampling the basic integral representation, and other methods not yet covered by the literature. Working with standard numerical methods on the fundamental integral representation of \(\gamma (m + 1/2, z)\) is generally inefficient as it demands dense sampling of complex exponentials. Continued fraction and rational function approximations are difficult to control, because the regions of known accuracy in the \(z\)-plane are complicated. The most suitable scheme is the construction of Taylor expansions around nodes of a regular, fixed grid in the \(z\)-plane, which stores a static matrix of higher derivatives. This is the obvious extension of a procedure that is in common use for real-valued \(z\).