2
$\begingroup$

Consider an indefinite "elliptic integral”, $$ J_n(r):=\int^r_{r_0}\frac{x^n}{\sqrt{f(x)}}, \quad f(x):=(x-x_1)(x-x_2)(x-x_3). $$ where $r_0>0, n=1,0,βˆ’1$, and $x_{1,2,3}$ are roots of the cubic equation $𝑓(π‘₯)=0$.

I am interested in a case when $𝑓(π‘₯)=0$ has one real root $π‘₯_1>0$ and two complex roots $π‘₯_2$ and $π‘₯_3$ (since it's a cubic equation, they are also complex conjugate, $π‘₯_3=\bar{x_2}$).

I want to know if $𝐽_{βˆ’1}$ can be written without imaginary number $𝑖$, because this integral appears in a problem of physics, $𝐽_𝑛$ should be real. But, so far, I failed to make it. I write down what I faced:

I have tried an integration by substitution. In the present case, let $π‘₯_2=π‘Ž+𝑖𝑏$ and thus $π‘₯_3=\bar{x_2}=π‘Žβˆ’π‘–π‘$. Then we have $𝑓(π‘₯)=(π‘₯βˆ’π‘₯_1)((π‘₯βˆ’π‘Ž)^2+𝑏^2)$.

Now, transform the variable from $π‘₯$ to $𝑦$ with $$ π‘₯=π‘₯_1+𝑝 \tan^2𝑦, 𝑝:=\sqrt{(π‘₯_1βˆ’π‘Ž)^2+𝑏^2}. $$ After some manipulation, $𝐽_𝑛$ reduces to $$ \frac{\sqrt{p}}{2}𝐽_𝑛(π‘Ÿ)=\int\frac{(x_1+𝑝\tan^2y)^n {\rm d}y}{\sqrt{1βˆ’π΄\sin^2𝑦\cos^2𝑦}}, $$ where $𝐴:=2(π‘βˆ’(π‘₯_1βˆ’π‘Ž))/𝑝$ and thus $0<𝐴<2$ by definition.

For $𝑛=0$ and $n=+1$, there is no problem: For $𝑛=0$, a software "Mathematica" returns $$ \frac{\sqrt{p}}{2}𝐽_0(π‘Ÿ)=\frac{1}{2}𝐹(2𝑦,\frac{A}{4}), $$ where $𝐹(\phi,π‘š)$ is Elliptic integral of the first kind. For $𝑛=1$, Mathematica returns a combination of elliptic functions and elementary functions (I do not express them here, though).

Note that there is no imaginary number $i$ so far.

For $𝑛=βˆ’1$, however, integration returns a messy form with $i$: enter image description here

where $\Pi(n,\phi,m)$ is incomplete Elliptic integral of the third kind.

I guess there should be a simpler form of $𝐽_{βˆ’1}$ without $i$. Maybe the above transformation is not good for $𝐽_{βˆ’1}$? Is there some identities that eliminate $i$? I am not comfortable that in the messy form, because $\sqrt{A-4}$ appears, although $0<𝐴<2$.

I am beginner of elliptic integral, any comments are welcome. Thank you.

$\endgroup$
4
  • 2
    $\begingroup$ Let $\pm \sqrt {1 - u^2} = (x - x_1 - p)/(x - x_1 + p)$, then $$\int_{x_1 + p}^r \frac {dx} {x \sqrt {(x - x_1) ((x - a)^2 + b^2)}} = \frac 1 {4 x_1 p^{3/2}} \int_{u_0}^1 \frac {\pm ((x_1 - p) u^2 + 2 p) - 2 p \sqrt {1 - u^2}} {(1 - n u^2) \sqrt {1 - u^2} \sqrt {1 - m u^2}} du, \\ m = \frac 1 2 \left( 1 + \frac {a - x_1} p \right), \quad n = -\frac {(x_1 - p)^2} {4 x_1 p}, \quad u_0 = \frac {2 \sqrt {(r - x_1) p}} {r - x_1 + p},$$ with the plus sign corresponding to $r > x_1 + p$ and the minus sign to $r <x_1 + p$. This gives $F$, $\Pi$ and an elementary function. $\endgroup$
    – Maxim
    Jun 25, 2022 at 13:36
  • $\begingroup$ Thanks! How did you come up with the transformation? Just amazing for me... $\endgroup$ Jun 26, 2022 at 7:33
  • 2
    $\begingroup$ Those are fairly standard steps in the theory of elliptic integrals. If we apply the substitution $v = (x - A)/(x - B)$, we get a quartic polynomial under the square root. Choosing $(A, B) = x_1 \pm p$ eliminates the $v^1$ and $v^3$ terms. The quartic becomes $(1 - v^2) (C^2 + v^2)$, so we take $u^2 = 1 - v^2$. $\endgroup$
    – Maxim
    Jun 27, 2022 at 20:08
  • $\begingroup$ @Maxim Thank you. If you would like to post your answer, I can click your answer that was helpful. $\endgroup$ Jun 28, 2022 at 2:55

0

You must log in to answer this question.

Browse other questions tagged .