9.18 Tables9.20 Software

§9.19 Approximations

Contents

§9.19(i) Approximations in Terms of Elementary Functions

  • Martín et al. (1992) provides two simple formulas for approximating \mathop{\mathrm{Ai}\/}\nolimits\!\left(x\right) to graphical accuracy, one for -\infty<x\leq 0, the other for 0\leq x<\infty.

  • Moshier (1989, §6.14) provides minimax rational approximations for calculating \mathop{\mathrm{Ai}\/}\nolimits\!\left(x\right), {\mathop{\mathrm{Ai}\/}\nolimits^{{\prime}}}\!\left(x\right), \mathop{\mathrm{Bi}\/}\nolimits\!\left(x\right), {\mathop{\mathrm{Bi}\/}\nolimits^{{\prime}}}\!\left(x\right). They are in terms of the variable \zeta, where \zeta=\tfrac{2}{3}x^{{3/2}} when x is positive, \zeta=\tfrac{2}{3}(-x)^{{3/2}} when x is negative, and \zeta=0 when x=0. The approximations apply when 2\leq\zeta<\infty, that is, when 3^{{2/3}}\leq x<\infty or -\infty<x\leq-3^{{2/3}}. The precision in the coefficients is 21S.

§9.19(ii) Expansions in Chebyshev Series

These expansions are for real arguments x and are supplied in sets of four for each function, corresponding to intervals -\infty<x\leq a, a\leq x\leq 0, 0\leq x\leq b, b\leq x<\infty. The constants a and b are chosen numerically, with a view to equalizing the effort required for summing the series.

  • Prince (1975) covers \mathop{\mathrm{Ai}\/}\nolimits\!\left(x\right), {\mathop{\mathrm{Ai}\/}\nolimits^{{\prime}}}\!\left(x\right), \mathop{\mathrm{Bi}\/}\nolimits\!\left(x\right), {\mathop{\mathrm{Bi}\/}\nolimits^{{\prime}}}\!\left(x\right). The Chebyshev coefficients are given to 10-11D. Fortran programs are included. See also Razaz and Schonfelder (1981).

  • Németh (1992, Chapter 8) covers \mathop{\mathrm{Ai}\/}\nolimits\!\left(x\right), {\mathop{\mathrm{Ai}\/}\nolimits^{{\prime}}}\!\left(x\right), \mathop{\mathrm{Bi}\/}\nolimits\!\left(x\right), {\mathop{\mathrm{Bi}\/}\nolimits^{{\prime}}}\!\left(x\right), and integrals \int _{0}^{x}\mathop{\mathrm{Ai}\/}\nolimits\!\left(t\right)dt, \int _{0}^{x}\mathop{\mathrm{Bi}\/}\nolimits\!\left(t\right)dt, \int _{0}^{x}\int _{0}^{v}\mathop{\mathrm{Ai}\/}\nolimits\!\left(t\right)dtdv, \int _{0}^{x}\int _{0}^{v}\mathop{\mathrm{Bi}\/}\nolimits\!\left(t\right)dtdv (see also (9.10.20) and (9.10.21)). The Chebyshev coefficients are given to 15D. Chebyshev coefficients are also given for expansions of the second and higher (real) zeros of \mathop{\mathrm{Ai}\/}\nolimits\!\left(x\right), {\mathop{\mathrm{Ai}\/}\nolimits^{{\prime}}}\!\left(x\right), \mathop{\mathrm{Bi}\/}\nolimits\!\left(x\right), {\mathop{\mathrm{Bi}\/}\nolimits^{{\prime}}}\!\left(x\right), again to 15D.

  • Razaz and Schonfelder (1980) covers \mathop{\mathrm{Ai}\/}\nolimits\!\left(x\right), {\mathop{\mathrm{Ai}\/}\nolimits^{{\prime}}}\!\left(x\right), \mathop{\mathrm{Bi}\/}\nolimits\!\left(x\right), {\mathop{\mathrm{Bi}\/}\nolimits^{{\prime}}}\!\left(x\right). The Chebyshev coefficients are given to 30D.

§9.19(iii) Approximations in the Complex Plane

  • Corless et al. (1992) describe a method of approximation based on subdividing \Complex into a triangular mesh, with values of \mathop{\mathrm{Ai}\/}\nolimits\!\left(z\right), {\mathop{\mathrm{Ai}\/}\nolimits^{{\prime}}}\!\left(z\right) stored at the nodes. \mathop{\mathrm{Ai}\/}\nolimits\!\left(z\right) and {\mathop{\mathrm{Ai}\/}\nolimits^{{\prime}}}\!\left(z\right) are then computed from Taylor-series expansions centered at one of the nearest nodes. The Taylor coefficients are generated by recursion, starting from the stored values of \mathop{\mathrm{Ai}\/}\nolimits\!\left(z\right), {\mathop{\mathrm{Ai}\/}\nolimits^{{\prime}}}\!\left(z\right) at the node. Similarly for \mathop{\mathrm{Bi}\/}\nolimits\!\left(z\right), {\mathop{\mathrm{Bi}\/}\nolimits^{{\prime}}}\!\left(z\right).

§9.19(iv) Scorer Functions

  • MacLeod (1994) supplies Chebyshev-series expansions to cover \mathop{\mathrm{Gi}\/}\nolimits\!\left(x\right) for 0\leq x<\infty and \mathop{\mathrm{Hi}\/}\nolimits\!\left(x\right) for -\infty<x\leq 0. The Chebyshev coefficients are given to 20D.