I'm trying to determine the limit in sense of distributions of the sequence $f_{n}=\frac{n}{1+n^2x^2}$. First I wrote this:

$$\begin{align}\int_{\mathbb{R}}f_n(x) \phi(x) dx&=\arctan(nx)\phi(x)|_{-\infty}^{+\infty} -\int_{\mathbb{R}}\arctan(nx) \phi'(x) dx\\\\ &=-\int_{-\infty}^{0}\arctan(nx) \phi'(x) dx-\int_{0}^{+\infty}\arctan(nx) \phi'(x) dx \end{align}$$

I tried proving that the sequence converges to $\pi\delta_{0}$, since $\arctan(nx)\rightarrow \text{sgn(x)}\frac{\pi}{2}$, as $n\rightarrow +\infty$, but without success. Could someone hint what the limit could be so I can try to prove it?


Note that

$$\int_{-\infty}^\infty \frac{n}{1+n^2x^2}\phi(x)\,dx\overbrace{=}^{x\mapsto x/n}\int_{-\infty}^\infty \frac1{1+x^2}\phi(x/n)\,dx$$

Now, since $\phi(x)$ is a test function, it is bounded and continuous. Hence, applying the Dominated Convergence Theorem yields

$$\begin{align} \lim_{n\to \infty}\int_{-\infty}^\infty \frac{n}{1+n^2x^2}\phi(x)\,dx&=\lim_{n\to\infty}\int_{-\infty}^\infty \frac1{1+x^2}\phi(x/n)\,dx\\\\ &=\int_{-\infty}^\infty \frac1{1+x^2}\lim_{n\to\infty}\phi(x/n)\,dx\\\\ &=\phi(0)\int_{-\infty}^\infty \frac{1}{1+x^2}\,dx\\\\ &=\pi\phi(0) \end{align}$$

If we wish to proceed as in the OP, then using the fact that $\phi$ is of compact support, we have

$$\int_{-\infty}^\infty \frac{n}{1+n^2x^2}\phi(x)=-\int_{-\infty}^\infty \arctan(nx)\phi'(x)\,dx$$

Since $|\arctan(nx)|\le \pi/2$, then $|\arctan(nx)\phi'(x)|\le \frac\pi2 |\phi'(x)|$. Inasmuch as $\phi'(x)$ is continuous and of compact support, then $\int_{-\infty}^\infty|\phi'(x)|\,dx<\infty$. Again, we appeal to the Dominated Convergence Theorem to find

$$\begin{align} \lim_{n\to \infty}\int_{-\infty}^\infty \arctan(nx)\phi'(x)\,dx&=-\int_{-\infty}^\infty \lim_{n\to\infty}\arctan(nx)\phi'(x)\,dx\\\\ &=-\int_{-\infty}^\infty \phi'(x)\frac\pi 2\text{sgn}(x)\,dx\\\\ &=-\frac\pi2 \left(\int_0^\infty \phi'(x)\,dx-\int_{-\infty}^0 \phi'(x)\,dx\right)\\\\ &=-\frac\pi2 \left(-\phi(0)-\phi(0)\right)\\\\ &=\pi\phi(0) \end{align}$$

  • $\begingroup$ The first argument is definitely the "right" one, because it applies in much greater generality - proceeding as in the OP gives the idea that there's something special about $1/(1+x^2)$. In fact if $f\in L^1$, $\int f=1$, and $f_n(x)=nf(nx)$ then $\int f_n\phi\to\phi(0)$ for every bounded continuous $\phi$, exactly as above. $\endgroup$ – David C. Ullrich Apr 7 '18 at 14:12
  • $\begingroup$ @DavidC.Ullrich Indeed. The specific function is only an example within the general class that you identified. $\endgroup$ – Mark Viola Apr 7 '18 at 14:16

Hint: $$H(x)=\lim_{n\to\infty}\left(\dfrac{1}{2}+ \dfrac{1}{\pi} arctan(nx) \right)$$ where $H(x)$ is heaviside hunction defined by $H'=\delta_0$ or $H(x)=\int_{- \infty}^{x} \delta(s) ds$


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.