What's New
About the Project
NIST
19 Elliptic IntegralsComputation

§19.38 Approximations

Minimax polynomial approximations (§3.11(i)) for K(k) and E(k) in terms of m=k2 with 0m<1 can be found in Abramowitz and Stegun (1964, §17.3) with maximum absolute errors ranging from 4×10⁻⁵ to 2×10⁻⁸. Approximations of the same type for K(k) and E(k) for 0<k1 are given in Cody (1965a) with maximum absolute errors ranging from 4×10⁻⁵ to 4×10⁻¹⁸. Cody (1965b) gives Chebyshev-series expansions (§3.11(ii)) with maximum precision 25D.

Approximations for Legendre’s complete or incomplete integrals of all three kinds, derived by Padé approximation of the square root in the integrand, are given in Luke (1968, 1970). They are valid over parts of the complex k and ϕ planes. The accuracy is controlled by the number of terms retained in the approximation; for real variables the number of significant figures appears to be roughly twice the number of terms retained, perhaps even for ϕ near π/2 with the improvements made in the 1970 reference.