26
$\begingroup$

I am interested in a general closed-form formula for integrals of the following form: $$\mathcal{J}_q=\int_0^1x^{2\,q-1}\,K(x)^2dx,\tag0$$ where $K(x)$ is the complete elliptic integral of the 1ˢᵗ kind: $$K(x)={_2F_1}\left(\frac12,\frac12;\ 1;\ x^2\right)\frac\pi2\\=\int_0^{\pi/2}\frac{d\phi}{\sqrt{1-x^2\sin^2\phi}}=\int_0^1\frac{dz}{\sqrt{1-x^2z^2\vphantom{|^2}}\sqrt{1-z^2\vphantom{|^2}}}.\tag1$$ I could not find a suitable integral in DLMF or Gradshteyn—Ryzhik or Prudnikov tables of integrals. Mathematica was not able to evaluate it either, even when the parameter $q$ was fixed to an integer (Note: In Mathematica you would write EllipticK[x^2] for $K(x)$ because of a different convention for the modulus parameter).

But, numerical integration and lookups in WolframAlpha and ISC+ suggest that there are closed forms for (at least some) postitive integer values of the parameter $q$. For example,

$$\mathcal{J}_1\stackrel?=\frac{7\,\zeta(3)}4,\ \mathcal{J}_2\stackrel?=\frac{7\,\zeta(3)}8+\frac14,\\\mathcal{J}_3\stackrel?=\frac{77\,\zeta(3)}{128}+\frac{17}{64},\ \mathcal{J}_4\stackrel?=\frac{119\,\zeta(3)}{256}+\frac{881}{3456},\tag2$$ where $\zeta(3)$ is the Apéry constant. It looks that the general formula for postitive integer values of $q$ is $$\mathcal{J}_q\stackrel?=a_q\,\zeta(3)+b_q,\tag3$$ where $a_q,\,b_q$ are some rational coefficients. I tried to find general formulae for these coefficients, and actually discovered plausible candidates for both.

Conjecturally, $a_q$ can be expressed in terms of a generalized hypergeometric function: $$a_q\stackrel?=\frac{7\,\pi}4\cdot\frac{{_4F_3}\left(\begin{array}{c}\frac12,\,\frac12,\,1-q,\,1-q\\1,\,\frac32-q,\,\frac32-q\end{array}\middle|\ 1\right)}{\Gamma\left(\frac32-q\right)^2\ \Gamma(q)^2}.\tag4$$ Somewhat suprisingly, this scarish expression seems to evaluate only to rationals for all $q\in\mathbb{Z}^+$.

For $b_q$ I found only a conjectural recurrence relation: $$b_1\stackrel?=0,\ b_2\stackrel?=\frac14,\ b_q\stackrel?=\frac{2 \,(2\,q-3)\,\left(2\,q^2-6\,q+5\right)\cdot b_{q-1}-4\,(q-2)^3\cdot b_{q-2}+1}{4\,(q-1)^3},\tag5$$ but I could not find a general term formula for it.

  • Can we prove that the closed forms shown in $(2)$ are correct?
  • Can we prove that the formula $(3)$ holds for all $q\in\mathbb{Z}^+$ with some rational coefficients $a_q,\,b_q$?
  • Are my conjectured formulae for these coefficients $(4),\,(5)$ correct?
  • Can the formula $(4)$ be expanded in terms of simpler functions?
  • Is there a general term formula for $(5)$?
  • Can we find (or at least conjecture) a general closed-form formula for $\mathcal{J}_q$ where $q$ is not necessarily an integer?
$\endgroup$
17
+250
$\begingroup$

As pointed out in a comment, these and many related results are derived in the excellent paper "Moments of Elliptic Integrals", by James Wan (http://arxiv.org/abs/1101.1132). Specifically, the author shows that $$ K_{1}=\int_{0}^{1}xK(x)^2dx =\frac{7}{4}\zeta(3), $$ that $$ K_{3}=\int_{0}^{1}x^3 K^2(x)dx= \frac{1}{4}+\frac{7}{8}\zeta(3), $$ and finally that the recurrence $$ (n+1)^3K_{n+2} - 2n(n^2+1)K_{n}+(n-1)^3 K_{n-2}=2 $$ holds in general. In terms of your integrals, where ${\cal J}_q=K_{2q-1}$, we find that $$ {\cal J}_{q}=K_{2q-1}=\frac{2+2(2q-3)((2q-3)^2+1)K_{2q-3}-(2q-4)^3 K_{2q-5}}{(2q-2)^3} \\ =\frac{1}{4(q-1)^3}+\frac{(2q-3)(2q^2-6q+5)}{2(q-1)^3}{\cal J}_{q-1}-\frac{(q-2)^3}{(q-1)^3}{\cal J}_{q-2}. $$ Then indeed ${\cal J}_q=a_q\zeta(3)+b_q$, where $a_1=7/4$, $a_2=7/8$, $b_1=0$, $b_2=1/4$; and $a_q$ and $b_q$ satisfy these recurrence relations: $$ a_{q}=\frac{(2q-3)(2q^2-6q+5)}{2(q-1)^3}a_{q-1}-\frac{(q-2)^3}{(q-1)^3}a_{q-2} $$ and $$ b_{q}=\frac{1}{4(q-1)^3}+\frac{(2q-3)(2q^2-6q+5)}{2(q-1)^3}b_{q-1}-\frac{(q-2)^3}{(q-1)^3}b_{q-2} $$ (where the latter matches your conjecture).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.