Digital Library of Mathematical Functions
About the Project
NIST
19 Elliptic IntegralsSymmetric Integrals

§19.24 Inequalities

Contents

§19.24(i) Complete Integrals

The condition y\leq z for (19.24.1) and (19.24.2) serves only to identify y as the smaller of the two nonzero variables of a symmetric function; it does not restrict validity.

19.24.2\tfrac{1}{2}\leq z^{{-1/2}}\mathop{R_{G}\/}\nolimits\!\left(0,y,z\right)\leq%
\tfrac{1}{4}\pi,0\leq y\leq z,
19.24.3\left(\frac{y^{{3/2}}+z^{{3/2}}}{2}\right)^{{2/3}}\leq\frac{4}{\pi}\mathop{R_{%
G}\/}\nolimits\!\left(0,y^{2},z^{2}\right)\leq\left(\frac{y^{2}+z^{2}}{2}%
\right)^{{1/2}},y>0, z>0.

If y, z, and p are positive, then

19.24.4\frac{2}{\sqrt{p}}(2yz+yp+zp)^{{-1/2}}\leq\frac{4}{3\pi}\mathop{R_{J}\/}%
\nolimits\!\left(0,y,z,p\right)\leq(yzp^{2})^{{-3/8}}.

Inequalities for \mathop{R_{D}\/}\nolimits\!\left(0,y,z\right) are included as the case p=z.

A series of successively sharper inequalities is obtained from the AGM process (§19.8(i)) with a_{0}\geq g_{0}>0:

where

19.24.6
a_{{n+1}}=(a_{n}+g_{n})/2,
g_{{n+1}}=\sqrt{a_{n}g_{n}}.

Other inequalities can be obtained by applying Carlson (1966, Theorems 2 and 3) to (19.16.20)–(19.16.23). Approximations and one-sided inequalities for \mathop{R_{G}\/}\nolimits\!\left(0,y,z\right) follow from those given in §19.9(i) for the length L(a,b) of an ellipse with semiaxes a and b, since

For x>0, y>0, and x\neq y, the complete cases of \mathop{R_{F}\/}\nolimits and \mathop{R_{G}\/}\nolimits satisfy

Also, with the notation of (19.24.6),

with equality iff a_{0}=g_{0}.

§19.24(ii) Incomplete Integrals

Inequalities for \mathop{R_{{-a}}\/}\nolimits\!\left(\mathbf{b};\mathbf{z}\right) in Carlson (1966, Theorems 2 and 3) can be applied to (19.16.14)–(19.16.17). All variables are positive, and equality occurs iff all variables are equal.

Examples

19.24.10\frac{3}{\sqrt{x}+\sqrt{y}+\sqrt{z}}\leq\mathop{R_{F}\/}\nolimits\!\left(x,y,z%
\right)\leq\frac{1}{(xyz)^{{1/6}}},

Inequalities for \mathop{R_{C}\/}\nolimits\!\left(x,y\right) and \mathop{R_{D}\/}\nolimits\!\left(x,y,z\right) are included as special cases (see (19.16.6) and (19.16.5)).

Other inequalities for \mathop{R_{F}\/}\nolimits\!\left(x,y,z\right) are given in Carlson (1970).

If a (\neq 0) is real, all components of \mathbf{b} and \mathbf{z} are positive, and the components of z are not all equal, then

19.24.13
\mathop{R_{{a}}\/}\nolimits\!\left(\mathbf{b};\mathbf{z}\right)\mathop{R_{{-a}%
}\/}\nolimits\!\left(\mathbf{b};\mathbf{z}\right)>1,
\mathop{R_{{a}}\/}\nolimits\!\left(\mathbf{b};\mathbf{z}\right)+\mathop{R_{{-a%
}}\/}\nolimits\!\left(\mathbf{b};\mathbf{z}\right)>2;

see Neuman (2003, (2.13)). Special cases with a=\pm\frac{1}{2} are (19.24.8) (because of (19.16.20), (19.16.23)), and

The same reference also gives upper and lower bounds for symmetric integrals in terms of their elementary degenerate cases. These bounds include a sharper but more complicated lower bound than that supplied in the next result:

with equality iff y=z.