Digital Library of Mathematical Functions
About the Project
NIST
12 Parabolic Cylinder FunctionsProperties

§12.14 The Function \mathop{W\/}\nolimits\!\left(a,x\right)

Contents

§12.14(i) Introduction

In this section solutions of equation (12.2.3) are considered. This equation is important when a and z (=x) are real, and we shall assume this to be the case. In other cases the general theory of (12.2.2) is available. \mathop{W\/}\nolimits\!\left(a,x\right) and \mathop{W\/}\nolimits\!\left(a,-x\right) form a numerically satisfactory pair of solutions when -\infty<x<\infty.

§12.14(ii) Values at z=0 and Wronskian

12.14.1\mathop{W\/}\nolimits\!\left(a,0\right)=2^{{-\frac{3}{4}}}\left|\frac{\mathop{%
\Gamma\/}\nolimits\!\left(\tfrac{1}{4}+\tfrac{1}{2}ia\right)}{\mathop{\Gamma\/%
}\nolimits\!\left(\tfrac{3}{4}+\tfrac{1}{2}ia\right)}\right|^{{\frac{1}{2}}},
12.14.2{\mathop{W\/}\nolimits^{{\prime}}}\!\left(a,0\right)=-2^{{-\frac{1}{4}}}\left|%
\frac{\mathop{\Gamma\/}\nolimits\!\left(\tfrac{3}{4}+\tfrac{1}{2}ia\right)}{%
\mathop{\Gamma\/}\nolimits\!\left(\tfrac{1}{4}+\tfrac{1}{2}ia\right)}\right|^{%
{\frac{1}{2}}}.

§12.14(iii) Graphs

For the modulus functions \widetilde{F}(a,x) and \widetilde{G}(a,x) see §12.14(x).

See accompanying text
Figure 12.14.1: k^{{-\ifrac{1}{2}}}\mathop{W\/}\nolimits\!\left(3,x\right), k^{{\ifrac{1}{2}}}\mathop{W\/}\nolimits\!\left(3,-x\right), \widetilde{F}(3,x), 0\leq x\leq 8. Magnify
See accompanying text
Figure 12.14.2: k^{{-\ifrac{1}{2}}}{\mathop{W\/}\nolimits^{{\prime}}}\!\left(3,x\right), k^{{\ifrac{1}{2}}}{\mathop{W\/}\nolimits^{{\prime}}}\!\left(3,-x\right), \widetilde{G}(3,x), 0\leq x\leq 8. Magnify
See accompanying text
Figure 12.14.3: k^{{-\ifrac{1}{2}}}\mathop{W\/}\nolimits\!\left(-3,x\right), k^{{\ifrac{1}{2}}}\mathop{W\/}\nolimits\!\left(-3,-x\right), \widetilde{F}(-3,x), 0\leq x\leq 8. Magnify
See accompanying text
Figure 12.14.4: k^{{-\ifrac{1}{2}}}{\mathop{W\/}\nolimits^{{\prime}}}\!\left(-3,x\right),k^{{%
\ifrac{1}{2}}}{\mathop{W\/}\nolimits^{{\prime}}}\!\left(-3,-x\right), \widetilde{G}(-3,x), 0\leq x\leq 8. Magnify

§12.14(iv) Connection Formula

where

12.14.5
k=\sqrt{1+e^{{2\pi a}}}-e^{{\pi a}},
1/k=\sqrt{1+e^{{2\pi a}}}+e^{{\pi a}},
12.14.6\rho=\tfrac{1}{8}\pi+\tfrac{1}{2}\phi_{2},
12.14.7\phi_{2}=\mathop{\mathrm{ph}\/}\nolimits\mathop{\Gamma\/}\nolimits\!\left(%
\tfrac{1}{2}+ia\right),

the branch of \mathop{\mathrm{ph}\/}\nolimits being zero when a=0 and defined by continuity elsewhere.

§12.14(v) Power-Series Expansions

12.14.8\mathop{W\/}\nolimits\!\left(a,x\right)=\mathop{W\/}\nolimits\!\left(a,0\right%
)w_{1}(a,x)+{\mathop{W\/}\nolimits^{{\prime}}}\!\left(a,0\right)w_{2}(a,x).

Here w_{1}(a,x) and w_{2}(a,x) are the even and odd solutions of (12.2.3):

12.14.9w_{1}(a,x)=\sum_{{n=0}}^{\infty}\alpha_{n}(a)\frac{x^{{2n}}}{(2n)!},
12.14.10w_{2}(a,x)=\sum_{{n=0}}^{\infty}\beta_{n}(a)\frac{x^{{2n+1}}}{(2n+1)!},

where \alpha_{n}(a) and \beta_{n}(a) satisfy the recursion relations

12.14.11
\alpha_{{n+2}}=a\alpha_{{n+1}}-\tfrac{1}{2}(n+1)(2n+1)\alpha_{{n}},
\beta_{{n+2}}=a\beta_{{n+1}}-\tfrac{1}{2}(n+1)(2n+3)\beta_{{n}},

with

12.14.12
\alpha_{0}(a)=1,
\alpha_{1}(a)=a,
\beta_{0}(a)=1,
\beta_{1}(a)=a.

Other expansions, involving \mathop{\cos\/}\nolimits\!\left(\tfrac{1}{4}x^{2}\right) and \mathop{\sin\/}\nolimits\!\left(\tfrac{1}{4}x^{2}\right), can be obtained from (12.4.3) to (12.4.6) by replacing a by -ia and z by xe^{{\ifrac{\pi i}{4}}}; see Miller (1955, p. 80), and also (12.14.15) and (12.14.16).

§12.14(vi) Integral Representations

These follow from the contour integrals of §12.5(ii), which are valid for general complex values of the argument z and parameter a. See Miller (1955, p. 26).

§12.14(vii) Relations to Other Functions

Bessel Functions

For the notation see §10.2(ii). When x>0

12.14.13\mathop{W\/}\nolimits\!\left(0,\pm x\right)=2^{{-\frac{5}{4}}}\sqrt{\pi x}%
\left(\mathop{J_{{-\frac{1}{4}}}\/}\nolimits\!\left(\tfrac{1}{4}x^{2}\right)%
\mp\mathop{J_{{\frac{1}{4}}}\/}\nolimits\!\left(\tfrac{1}{4}x^{2}\right)\right),

Confluent Hypergeometric Functions

For the notation see §13.2(i).

The even and odd solutions of (12.2.3) (see §12.14(v)) are given by

§12.14(viii) Asymptotic Expansions for Large Variable

Write

where

with \phi_{2} given by (12.14.7). Then as x\to\infty

The coefficients c_{{2r}} and d_{{2r}} are obtainable by equating real and imaginary parts in

12.14.22c_{{2r}}+id_{{2r}}=\frac{\mathop{\Gamma\/}\nolimits\!\left(2r+\tfrac{1}{2}+ia%
\right)}{\mathop{\Gamma\/}\nolimits\!\left(\tfrac{1}{2}+ia\right)}.

Equivalently,

§12.14(ix) Uniform Asymptotic Expansions for Large Parameter

The differential equation

follows from (12.2.3), and has solutions \mathop{W\/}\nolimits\!\left(\tfrac{1}{2}\mu^{2},\pm\mu t\sqrt{2}\right). For real \mu and t oscillations occur outside the t-interval [-1,1]. Airy-type uniform asymptotic expansions can be used to include either one of the turning points \pm 1. In the following expansions, obtained from Olver (1959), \mu is large and positive, and \delta is again an arbitrary small positive constant.

Positive a, 2\sqrt{a}<x<\infty

uniformly for t\in[1+\delta,\infty). Here {\cal{A}}_{s}(t) is as in §12.10(ii), \sigma is defined by

12.14.27\sigma=\mu^{2}\xi+\tfrac{1}{4}\pi,

with \xi given by (12.10.7), and

12.14.28l(\mu)=\sqrt{2}e^{{\frac{1}{8}\pi\mu^{2}}}e^{{i(\frac{1}{2}\phi_{2}-\frac{1}{8%
}\pi)}}g(\mu e^{{-\frac{1}{4}\pi i}}),

with g(\mu) as in §12.10(ii). The function l(\mu) has the asymptotic expansion

with

12.14.30
l_{0}=1,
l_{1}=-\tfrac{1}{1152},
l_{2}=-\tfrac{16123}{398\;13120}.

Positive a, -2\sqrt{a}<x<2\sqrt{a}

uniformly for t\in[-1+\delta,1-\delta], with \eta given by (12.10.23) and {\widetilde{\cal A}}_{s}(t) given by (12.10.24).

The expansions for the derivatives corresponding to (12.14.25), (12.14.26), and (12.14.31) may be obtained by formal term-by-term differentiation with respect to t; compare the analogous results in §§12.10(ii)12.10(v).

Negative a, -\infty<x<\infty

In this case there are no real turning points, and the solutions of (12.2.3), with z replaced by x, oscillate on the entire real x-axis.

uniformly for t\in\Real, where

12.14.36{\overline{\sigma}}=\mu^{2}{\overline{\xi}}+\tfrac{1}{4}\pi,

and {\overline{\xi}} and the coefficients \overline{u}_{{s}}(t) and \overline{v}_{{s}}(t) as in §12.10(v).

§12.14(x) Modulus and Phase Functions

As noted in §12.14(ix), when a is negative the solutions of (12.2.3), with z replaced by x, are oscillatory on the whole real line; also, when a is positive there is a central interval -2\sqrt{a}<x<2\sqrt{a} in which the solutions are exponential in character. In the oscillatory intervals we write

where k is defined in (12.14.5), and \widetilde{F}(a,x) (>0), \widetilde{\theta}(a,x), \widetilde{G}(a,x) (>0), and \widetilde{\psi}(a,x) are real. \widetilde{F} or \widetilde{G} is the modulus and \widetilde{\theta} or \widetilde{\psi} is the corresponding phase. Compare §12.2(vi).

For properties of the modulus and phase functions, including differential equations and asymptotic expansions for large x, see Miller (1955, pp. 87–88). For graphs of the modulus functions see §12.14(iii).

§12.14(xi) Zeros of \mathop{W\/}\nolimits\!\left(a,x\right), {\mathop{W\/}\nolimits^{{\prime}}}\!\left(a,x\right)

For asymptotic expansions of the zeros of \mathop{W\/}\nolimits\!\left(a,x\right) and {\mathop{W\/}\nolimits^{{\prime}}}\!\left(a,x\right), see Olver (1959).