For positive integer $n$, an integral of the form $\int_0^\infty x^n e^{-ax}\,\mathrm{d}x$ will reduce to $\frac{n!}{a^{n+1}}$. This can be shown by successively applying integration by parts and observing that the $\left.f(x)g(x)\right|_0^\infty$ terms vanish. The first such term $f(x)g(x)$ is $\left.-\frac{1}{a}e^{-ax}x^n\right|_0^\infty$. It is clear that an exponential will "beat" any power function as $x$ tends to infinity, but how does one show this analytically given the indeterminate form of the expression in the infinite limit? I presume by application of L'Hôpital's rule $n$ times, but I wanted to verify.

After applying integration by parts $n$ times, the only term that does not vanish is of the form $\frac{n!}{a^n} \int_0^\infty e^{-ax}\,\mathrm{d}x$, which quite simply evaluates to $\frac{n!}{a^n} \left(\left.\frac{-1}{a} e^{-ax}\right)\right|_0^\infty = \frac{n!}{a^{n+1}}$.

However, for non-integer $n$, I am getting stuck after integrating by parts $\lfloor n\rfloor$ times. After consulting tables of integration, I realize that the solution is $\frac{\Gamma(n+1)}{a^{n+1}}$, which makes sense given that factorials for non-integer $n$ are evaluated by the gamma function. However, I am not sure how to show this analytically. If one proceeds as before by successively applying integration by parts ($\lfloor n \rfloor$ times) to reduce the integral $\int_0^\infty x^n e^{-ax}\,\mathrm{d}x$, then the $\lfloor n \rfloor$ terms of the form $\left.f(x)g(x)\right|_0^\infty$ vanish and one is left with a final term of the form $\frac{n!}{a^{n}} \int_0^\infty e^{-ax}x^{n-\lfloor n \rfloor}\,\mathrm{d}x$, where $0<n-\lfloor n \rfloor<1$. This is analogous to the previous case of positive $n\in\mathbb{Z}$, except for the residual $x^{n-\lfloor n \rfloor}$ term. Another application of integration by parts would not help, as it would simply reduce the exponent of $x$ to a negative value, and integration by parts could continue ad infinitum. Rather than continuing integration by parts ad infinitum, the $\Gamma$ function appears. How does one resolve this paradox and thereby arrive at the solution $\frac{\Gamma(n+1)}{a^{n+1}}$? Thank you very much.


You can use substitution, and see the $\Gamma$ function appears right away. Integration by parts kind of mixes things up since we define factorials for integers by the iterated multiplication, so with any non-integer number you'll have problems, so is better to use $\Gamma$ directly:

$$ax = u$$

$$\frac{1}{{{a^{n + 1}}}}\int\limits_0^\infty {{u^n}{e^{ - u}}dx} = \frac{{\Gamma \left( {n + 1} \right)}}{{{a^{n + 1}}}}$$

  • $\begingroup$ Thank you. I think that I didn't realize that the expression was in fact the definition of the gamma function (adjusted by a constant). I thought I had to reduce the power of $x$ (by integration by parts, for instance) to be a real number between $0$ and $1$ in order to invoke the gamma function. $\endgroup$ – user001 Feb 17 '12 at 0:30
  • $\begingroup$ @user001 That's the magic of maths. You can give yourself a name or nickname so people know you. $\endgroup$ – Pedro Tamaroff Feb 17 '12 at 0:35
  • $\begingroup$ So, does the gamma function count as being an analytic solution to the integral posed in the question? If that is the case, I could assign a name to any integral and declare that name to be the analytic solution. Also, why is this particular integral so special to have received a name? (I presume because it is "useful" or "interesting," but maybe there is a more satisfying reason.) Thanks again. $\endgroup$ – user001 Feb 17 '12 at 0:54
  • $\begingroup$ @user001 You should do some research on the Gamma function or ask here, you'll find out it pops up everywhere (I might be exaggerating, but well...) $\endgroup$ – Pedro Tamaroff Feb 17 '12 at 1:19

For your first question: Yes, for $a>0$, write $x^n e^{-ax}$ as $x^n\over e^{ax}$ and note that applying L'Hopital's rule $n$-times is justified and will leave you with finding $\lim\limits_{x\rightarrow\infty} {n!\over a^n e^{ax}}$ . Then you can evaluate this limit by other reasoning see that it is 0. (note also when "plugging" in the lower limit $x=0$, you get 0).


$$ {-1\over a} x^ne^{-ax} \Bigl|_0^\infty = {-1\over a} \lim\limits_{x\rightarrow\infty} { x^n\over e^{ ax}}-0 = {-1\over a} \lim\limits_{x\rightarrow\infty} {n!\over a^n e^{ax}} =0. $$

  • $\begingroup$ Thank you for confirming my suspicions. $\endgroup$ – user001 Feb 17 '12 at 0:27

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.