I'm trying to integrate

$$\int_0^{x<a} {\frac{e^x}{\sqrt{x(a-x)}}\,\mathrm dx}$$

I've tried various substitutions to no avail. Any ideas?

  • $\begingroup$ Is $a$ a constant? $\endgroup$ – Brian Mar 21 at 22:50
  • $\begingroup$ Yes, sorry! I'll edit the question. $\endgroup$ – SabrinaChoice Mar 21 at 22:53
  • $\begingroup$ Your system has a singularity when $x=a$, so the interval of integration needs to be specified or else the integral will inevitably blow up. $\endgroup$ – aghostinthefigures Mar 21 at 22:54
  • $\begingroup$ Thanks, you're right. I'll edit as such. $\endgroup$ – SabrinaChoice Mar 21 at 22:55
  • $\begingroup$ @DanielMcLaury I don't have any reason to believe it can be written in closed form. Just hoping it's possible. $\endgroup$ – SabrinaChoice Mar 21 at 23:31

To expand on my comment, trying the $a = 1$ and $a = 2$ cases in WolframAlpha gives

$$\int_0^1 \frac{e^x}{\sqrt{x(1-x)}}\, dx = \sqrt{e} \, \pi \, I_0(1/2)$$


$$\int_0^2 \frac{e^x}{\sqrt{x(2-x)}}\, dx = e \, \pi \, I_0(1)$$

suggesting that the general solution might be

$$\int_0^a \frac{e^x}{\sqrt{x(a-x)}}\, dx = e^{a/2} \pi I_0(a/2).$$

where $I_0$ is a modified Bessel function of the first kind.


Now that I have a second, I was able to do the substitutions suggested by the form of the expressions above and it indeed works out:

$$\int_0^a \frac{e^x \, dx}{\sqrt{x(a-x)}} = \int_0^{2b} \frac{e^x \, dx}{\sqrt{x(2b-x)}}$$

(where $a=2b$),

$$= \int_0^{2b} \frac{e^x \, dx}{\sqrt{b^2 - (x-b)^2}} = \int_{-1}^{1} \frac{b e^{b u + b} \, du}{\sqrt{b^2 - (b u)^2}}$$

(where $b u = x + b$),

$$= e^b \int_{-1}^{1} \frac{e^{b u} \, du}{\sqrt{1 - u^2}} = e^b \int_{-\pi/2}^{\pi/2} \frac{e^{b \sin\theta} \, \cos \theta \, d\theta}{\sqrt{1 - \sin^2 \theta}}$$

(where $u = \sin \theta$)

$$= e^b \int_{-\pi/2}^{\pi/2} e^{b \sin\theta} d\theta = e^b \pi I_0(b) = e^{a/2} \pi I_0\left(\frac{a}{2}\right)$$

Here $$I_0(x) = \frac{1}{\pi} \int_{-\pi/2}^{\pi/2} e^{x \sin\theta} \, d\theta$$

  • $\begingroup$ This is for sure a very nice solution for the particular case of $\int_0^a \frac{e^x \, dx}{\sqrt{x(a-x)}}$ . We still the problem with $\int_0^t \frac{e^x \, dx}{\sqrt{x(a-x)}}$ when $t < a$. Then $\to +1$. Cheers. $\endgroup$ – Claude Leibovici Mar 24 at 2:57
  • $\begingroup$ Perhaps I didn't understand the question, then. I thought x was the variable of integration. If we have an arbitrary upper bound then as you say the problem is harder. $\endgroup$ – Daniel McLaury Mar 24 at 3:17
  • $\begingroup$ For this particular case $t=a$, it is an interesting solution. May I confess that I tried even the case $t=a$ using several CAS with no success at all. $\endgroup$ – Claude Leibovici Mar 24 at 3:22

I am not sure that any closed form could be found and, may be, only series expansions could be the way to go.

First, let $x=ay$ to make $$I=\int_0^{t} {\frac{e^x}{\sqrt{x(a-x)}}\, dx}=\int_0^{\frac ta} {\frac{e^{ay}}{\sqrt{y(1-y)}}\, dy}$$ Expanding around $y=0$ would give $$\frac{1}{\sqrt{y(1-y) }}=\sum_{n=0}^\infty (-1)^n \binom{-\frac{1}{2}}{n}y^{n-\frac 12}$$ and $$\int e^{ay}\,y^{n-\frac 12}\,dy=-y^{n+\frac{1}{2}} E_{\frac{1}{2}-n}(-a y)$$ where appears the exponential integral function.


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.