Prove the convergence of the following integral: $$ \lim\limits_{n\to\infty}\frac{1}{n}\int_\limits{-n}^{n}\frac{1-e^{-nx^2}}{x^2(1+nx^2)}dx $$

I tried to use the obvious Weierstrass comparison test in the following way:

$$ \lim\limits_{n\to\infty}\frac{1}{n}\int_\limits{-n}^{n}\frac{1-e^{-nx^2}}{x^2(1+nx^2)}dx \leqslant \lim\limits_{n\to\infty}\frac{1}{n}\int_\limits{-n}^{n}\frac{1-e^{-nx^2}}{x^2}dx = \lim\limits_{n\to\infty}\int_\limits{-n}^{n}\frac{1-e^{-nx^2}}{nx^2}dx $$ However after thinking about a variable substitution I concluded that it would complicate the expression.


How should I prove the integral converges? What are your suggestions?

Edit: Tried a new approach but I would like someone to check it out. $\lim\limits_{n\to\infty}\frac{1}{n}\int_\limits{-n}^{n}\frac{1-e^{-nx^2}}{x^2(1+nx^2)}dx=\lim\limits_{n\to\infty}\frac{2}{n}\int_\limits{0}^{n}\frac{1-e^{-nx^2}}{x^2(1+nx^2)}dx$ since the function is even. Substitution $\sqrt{n}x=y$ So we get after some adjustments: $\lim\limits_{n\to\infty}\frac{2}{\sqrt{n}}\int_\limits{0}^{n\sqrt{n}}\frac{1-e^{-y^2}}{y^2(1+y^2)}dy$ Then I am not sure about the following step: $1-e^{-y^2}\sim y^2$ So we would get $\lim\limits_{n\to\infty}\frac{2}{\sqrt{n}}\int_\limits{0}^{n\sqrt{n}}\frac{{y^2}}{y^2(1+y^2)}dy=$$\lim\limits_{n\to\infty}\frac{2}{\sqrt{n}}\int_\limits{0}^{n\sqrt{n}}\frac{1}{(1+y^2)}dy=0$

  • $\begingroup$ I think it works by using the dominated convergence theorem, by considering the function : $\frac{1-e^{-nx^2}} {nx^2(1+nx^2)}1_{[-n;n]}$ which tends almost everywhere to 0 and which is dominated by a function on $L^1(\mathbb{R})$. $\endgroup$ – ChocoSavour Apr 22 '18 at 11:45
  • $\begingroup$ @ChocoSavour I am sorry but I am not supposed to use the Dominated Convergence Theorem. Anyway thanks for your input! $\endgroup$ – Pedro Gomes Apr 22 '18 at 12:24
  • $\begingroup$ @ChocoSavour Have you got any opinion about my edit? If you have one feel free to comment. $\endgroup$ – Pedro Gomes Apr 22 '18 at 12:25
  • $\begingroup$ In my opinion, there is two things : first, using asymptotic analysis is useful to determine if the integral converges or not, but not to determine the limit, cause if $f \sim g$ (and $f$, $g$ positive), I don't think that it permits to say that if the integral converges, then it converges to the same limit. However, here we have $1-e^{-y^2} \leq y^2$, so we have : $| \frac{2}{\sqrt{n}} \int_{0}^{n\sqrt{n}} \frac{1-e^{-y^2}}{y^2(1+y^2} \, \mathrm{d}y | \leq \frac{2}{\sqrt{n}} \int_{0}^{n\sqrt{n}} \frac{y^2}{y^2(1+y^2)} \, \mathrm{d}y$, and from here we can conclude (the second one tends to 0). $\endgroup$ – ChocoSavour Apr 22 '18 at 13:08
  • $\begingroup$ @ChocoSavour I guess you can say that $\lim_{n\to\infty}\frac{1}{n}\int_\limits{-n}^{n}\frac{1-e^{-nx^2}}{x^2(1+nx^2)}dx=0$ once $ \frac{1-e^{-nx^2}}{x^2(1+nx^2)}>0$. Since it cannot be negative and the original integral is lower than $0$ according to my resolution we could state it equals $0$. $\endgroup$ – Pedro Gomes Apr 22 '18 at 13:50

As in the comments indicated, We can just use the change of variables $x \sqrt{n} =y$ to get $$ \frac{1}{n} \int_{-n}^n \frac{1-\exp(-nx^2)}{x^2(1+nx^2)} =\frac{2}{\sqrt{n}} \int_0^{n^{3/2}} \frac{1-\exp(-x^2)}{x^2(1+x^2)} \mathrm{d} x. $$ Noting that $$0 \leq 1- \exp(-x^2) = \int_{0}^{x^2} \exp(-t) \, \mathrm{d} t \leq \min(x^2, \int_0^\infty \exp(-t) \mathrm{d} t) = \min(x^2,1)$$ we see that the last integral is bounded by $$\int_0^\infty \frac{\min(1,x^2)}{x^2(1+x^2)} \leq \int_0^1 \frac{1}{1+x^2} \mathrm{d} x+ \int_1^\infty \frac{1}{x^2(1+x^2)} < \infty.$$ Thus, we can conclude that the limes is $0$ and the convergence rate is $1/\sqrt{n}$.


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.