3
$\begingroup$

I have a definite integral that I am trying to solve. Any hint or reference is urgently sought.

png,

where $r$ is any positive integer while $\psi$ and $\nu$ are positive real numbers.

$\endgroup$
2
  • $\begingroup$ 1. Do you have a reason to believe it has a closed form? 2. Are less-than general results (for example for specific $\psi$ or $r$) of any use? 3. You can type LateX in your question. $\endgroup$
    – user111187
    May 14, 2014 at 13:45
  • $\begingroup$ I am not sure it has a closed form . less than general results is very welcome. and I will edit my question in a minute in latex. $\endgroup$
    – SA-255525
    May 14, 2014 at 13:57

2 Answers 2

6
+50
$\begingroup$

By substituting $x = \nu y$, the integral becomes

$$ \frac{1}{\nu^r} \int_0^{\infty} \frac{x^r e^x \,dx} {\left[1+\psi(e^x-1)\right]^{1+1/\psi}}$$ so the problem is reduced to finding

$$ I(a,n)= \int_0^{\infty} \frac{x^n e^x \,dx} {\left[1+\frac{1}{a}(e^x-1)\right]^{1+a}} $$ where I relabeled some letters.

The case $n = 0$ is solved as follows: substituting $u = 1+ \frac{1}{a}(e^x - 1)$ gives $$ I(a,0) =\int_1^{\infty} \frac{a}{u^{a+1}}\,du=1 $$

After some more testing, I conjecture that for all $n$ and $a$, $$ I(a,n)= n!\ a^{a-n} \, {}_pF_q \left(\underbrace{a,\ldots,a}_{n+1};\underbrace{1+a,\ldots,1+a}_{n};1-a \right) $$ While some may not consider this to be a closed form, its evaluation (for given $n$ and $a$) is about an order of magnitude faster than numerical integration.

For your convenience: if the above conjecture is true (it probably is) the result of the original integral becomes:

$$ I(r,\nu,\psi)= \frac{r!}{\nu^r} \psi^{r-1/\psi} \, {}_pF_q \left(\underbrace{\frac{1}{\psi},\ldots,\frac{1}{\psi}}_{r+1};\underbrace{1+\frac{1}{\psi},\ldots,1+\frac{1}{\psi}}_{r};1-\frac{1}{\psi} \right) $$ (I corrected a small copying error: I first had 2 in the exponent instead of $r$)

Mathematica input form:

i[r_, [Nu]_ , [Psi]_] := r!/[Nu]^r [Psi]^(r - 1/[Psi])HypergeometricPFQ[Array[1/`[Psi] &, r + 1], Array[1 + 1/[Psi] &, r], 1 - 1/[Psi]]

$\endgroup$
6
  • $\begingroup$ @111187 After substituting u=1+1/a(exp(x)-1) what happens with x^n for n not equal to zero. $\endgroup$
    – SA-255525
    May 17, 2014 at 4:39
  • $\begingroup$ @SA-255525 Well, then you get something nasty that can only be evaluated in terms of hypergeometric functions. That is why I only used this substitution for $n = 0$ and left the rest to Mathematica. $\endgroup$
    – user111187
    May 20, 2014 at 12:48
  • $\begingroup$ @111187 You didn't explain what happened with x^n after substitution. $\endgroup$
    – SA-255525
    May 20, 2014 at 13:18
  • $\begingroup$ @111187 I would highly appreciate it if you give me Mathematica input form $\endgroup$
    – SA-255525
    May 20, 2014 at 16:09
  • $\begingroup$ @SA-255525 The $x^n$ becomes $\ln^n(1 + a(u-1))$, which is not a nice enough form to do anything else with. $\endgroup$
    – user111187
    May 20, 2014 at 19:20
3
$\begingroup$

This definite integral is very complicated. I doubt that a simple closed form exists. Theoritically, the integral can be expressed as a finite sum of $r$ derivatives of an hypergeometric function (in attachment). But writting explicitly the derivatives would be arduous, all the more so the derivatives must be considered with respect to a parameter insead of the variable of the hypergeometric function.

enter image description here

$\endgroup$

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.