I am taking a somewhat hard measure theory course and I was asked to prove this:

a) Let $\alpha > 0$ be a real number. Prove that $$\Gamma(\alpha):=\int_0^\infty e^{-x}x^{\alpha-1}dx$$ exists. (We are studying the relationship between being Lebesgue-integrable and Riemann-integrable, so I am not quite sure what kind of integral should I prove exists, but I suspect it's the latter).

b)Prove that: $$\lim_{n\to\infty}\int_0^n \left( 1-\frac{x}{n}\right)^n x^{\alpha-1}dx=\Gamma(\alpha).$$ Hint: $(1-\frac{x}{n})^n\le (1-\frac{x}{n+1})^{n+1}$ whenever $\frac{x}{n}<1$.

This reeks of some limit theorem such as Monotone Convergence, Fatou or Lebesgue's Dominated Convergence, but those apply to Lebesgue-integrable functions. I the inequality stated in the hints makes me think I should use Monotone Convergence.

Any insight would be greatly appreciated. We are using Bartle's Measure Theory Book and we are more or less around the $\mathcal{L}_p$ spaces part.

Thank you in advance.

  • $\begingroup$ For (1) subdivide the integral in two integral:$$\int\limits_0^\infty=\int\limits_0^1+\int\limits_1^\infty$$ . For (2) read here a simple extension to Riemann Integral of the monotone convergence theorem: math.ucv.ro/~niculescu/articles/2011/… $\endgroup$ – DonAntonio Oct 18 '13 at 6:03
  • $\begingroup$ I am trying to prove convergence subdividing that integral into those very same integrals, and proving the second one converges is not that hard, is the one from $0$ to $1$ that's giving me a hard time. While I appreciate the reference, the problems are based on the lectures and we didn't get even close to discussing extensions of the MCT for R-integrals. While a proof based on it may be mathematically correct, I think I am expected to solve this using basic measure-theory results and Lebesgue's theorem. $\endgroup$ – alonso s Oct 18 '13 at 6:12

1) Hint: For $x\ge0$ and $n\ge1$, $e^x=\sum\limits_{k=0}^\infty\frac{x^k}{k!}\ge1+\frac{x^n}{n!}$, thus $e^{-x}\le\frac{n!}{n!+x^n}$

2) Hint: For $m\ge n$, $$ \int_0^m\left(1-\frac xm\right)^m x^{\alpha-1}\,\mathrm{d}x \ge\int_0^n\left(1-\frac xm\right)^m x^{\alpha-1}\,\mathrm{d}x \ge\int_0^n\left(1-\frac xn\right)^n x^{\alpha-1}\,\mathrm{d}x $$ and $\left(1-\frac xn\right)^n\le e^{-x}$ for $x\le n$.

  • $\begingroup$ Have you covered the Dominated Convergence Theorem? $\endgroup$ – robjohn Oct 18 '13 at 6:30
  • $\begingroup$ I got the first part, thank you for the tip. How does one use your hint for the second part? From what I can see, your hint shows that se sequence of integrals is monotone non-decreasing, and is bounded above by the Gamma integral, so the limit exists and is less than or equal to the Gamma function. I don't know if that's what you had in mind. $\endgroup$ – alonso s Oct 18 '13 at 7:09
  • $\begingroup$ @user2770617: Let $f_n(x)=[x\lt n]\left(1-\frac{x}{n}\right)^nx^{\alpha-1}$. Then the hint shows that $f_n(x)\le e^{-x}x^{\alpha-1}$ and $\lim\limits_{n\to\infty}f_n(x)=e^{-x}x^{\alpha-1}$. Dominated Convergence then gives $$ \lim_{n\to\infty}\int_0^n\left(1-\frac{x}{n}\right)^nx^{\alpha-1}\,\mathrm{d}x =\int_0^\infty e^{-x}x^{\alpha-1}\,\mathrm{d}x $$ $\endgroup$ – robjohn Oct 18 '13 at 7:56

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.