A method has been developed for an efficient approximation of incomplete gamma function that occurs in molecular orbital calculation with Gaussian-type basis sets. The method employs rational Chebyshev approximation and forward and backward recurrence relations to reduce the memory-space requirement without compromising accuracy. An example is illustrated for the cases wherein the sum of quantum numbers of orbitals is less than or equal to 8 and the rational functions are evaluated using double-precision numbers. In sucha case, results accurate to 50 significant bits can be obtained.
In this paper, we derive efficient quadratic and quartic iteration algorithms from the improvement of Gauss'arithmetic-geometric mean (AGM) algorithm. The number of multiplications in the improved quadratic algorithm is only half the number of the original algorithm, but the number of the square root operations in the improved AGM iterations is equal to the number of the original algorithm. So we derive an efficient simultaneous Newton iteration for the square root calculation. Next, weimplement a fast multiple-precision computation for the proposed algorithms and estimate the number of floating point operations and the execution time to compute the AGM iterations.