$f$ and $g$ are measurable functions on a measure space $(\mathbb R, \mathcal{B}, m)$ where $m$ is Lebesgue measure. $f(x)$ satisfies $$\int_\mathbb R f^2(x) dm(x) < \infty$$ Where, $$f_k(x) = \sqrt{k} f(kx).$$

Prove that if $g(x)$ satisfies $$\int_\mathbb R g^2 dm < \infty$$ then $$\lim\limits_{k \rightarrow \infty} \int_\mathbb R f_k(x) g(x) dm(x) = 0.$$

I can prove that $ f_k(x) g(x) $ is integrable, but I didn't figure out how to get the final result.

  • $\begingroup$ @Jack: Yes, Cauchy inequality makes sure that $f_k(x) g(x) $ is integrable. How can I prove the limit is zero? $\endgroup$ – David Nov 15 '17 at 15:29
  • $\begingroup$ @GuyFsone: No, that's all. I thought the way to prove dominated convergence theorem could work. $\endgroup$ – David Nov 15 '17 at 15:47
  • $\begingroup$ I have solved it by approximation of integral by continuous and compactly supported functions it is correct as well $\endgroup$ – Guy Fsone Nov 15 '17 at 16:32
  • 2
    $\begingroup$ Strongly related $\endgroup$ – Giuseppe Negro Nov 16 '17 at 18:11
  • $\begingroup$ @GiuseppeNegro: Thank you! I didn't find this question before. $\endgroup$ – David Nov 17 '17 at 6:16

Lemma: Continuous function of compact support are dense in $L^2(dm)$. Hence,

For, $\varepsilon>0$ we know that there exist a $f_\varepsilon,g_\varepsilon\in C_c(\Bbb R)$ a continuous function of compact support such that, $$ \|g-g_\varepsilon\|_2\le \varepsilon \|f\|_2^{-1} $$ and $$ \|f-f_\varepsilon\|_2\le \varepsilon \|g_\varepsilon\|_2^{-1} $$

Then,using Cauchy Schwartz inequalty we have, \begin{align} \left|\int_\mathbb R f_k(x) g(x) dm(x)\right| &= \left|\int_\mathbb R f_k(x) [g(x)-g_\varepsilon(x)] dm(x)+\int_\mathbb R f_k(x) g_\varepsilon(x) dm(x)\right|\\ &\le \|g-g_\varepsilon\|_2 \left( \int_\mathbb R k f^2(kx)dm(x)\right)^{1/2} +\left|\int_\mathbb R \sqrt{k}f(kx) g_\varepsilon(x) dm(x)\right| \\& \overset{\color{red}{u=kx}}{=} \|g-g_\varepsilon\|_2 \|f\|_2 + \left|\int_\mathbb R\frac{1}{ \sqrt{k}}f(x) g_\varepsilon(\frac{x}{k}) dm(x)\right|\\ & \le \varepsilon + \frac{1}{ \sqrt{k}}\left|\int_\mathbb Rf(x) g_\varepsilon(\frac{x}{k}) dm(x)\right| \end{align}

Similarly we have, $$ \left|\int_\mathbb Rf(x) g_\varepsilon(\frac{x}{k}) dm(x)\right| \le \varepsilon+ \left|\int_\mathbb Rf_\varepsilon(x) g_\varepsilon(\frac{x}{k}) dm(x)\right| \\= \varepsilon +\left|\int_{\text{supp }f_\varepsilon}f_\varepsilon(x) g_\varepsilon(\frac{x}{k}) dm(x)\right|$$

Since, $f_\varepsilon$ $h_\varepsilon$ are continuous functions with compact support, we get $$\color{blue}{\lim_{k\to \infty}\int_{\text{supp }f_\varepsilon}f_\varepsilon(x) g_\varepsilon(\frac{x}{k}) dm(x)=\int_{\text{supp }f_\varepsilon}f_\varepsilon(x) g_\varepsilon(0) dm(x)}$$

Whence, $$ \lim_{k\to \infty} \left|\int_\mathbb R f_k(x) g(x) dm(x)\right| \le\lim_{k\to \infty} \left[\varepsilon+\frac{\varepsilon}{ \sqrt{k}}+ \frac{1}{ \sqrt{k}}\left|\int_{\text{supp }f_\varepsilon}f_\varepsilon(x) g_\varepsilon(\frac{x}{k}) dm(x)\right| \right]=\varepsilon $$

That is for all $\varepsilon>0,$ we have, $$ \limsup_{k\to \infty} \left|\int_\mathbb R f_k(x) g(x) dm(x)\right| \le \varepsilon $$ letting $\varepsilon\to 0$ yields the require result.

$$ \color{red}{\lim_{k\to \infty} \int_\mathbb R f_k(x) g(x) dm(x)=0.} $$

| cite | improve this answer | |
  • $\begingroup$ @peterag you are right Ok I forgotten to remove the squares aftermath $\endgroup$ – Guy Fsone Nov 16 '17 at 13:50

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.