Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Define $T: L^2(\mathbb{R})\to L^2(\mathbb{R})$ by $(Tf)(x)=\int_{\mathbb{R}}\frac{f(y)}{1+|x|+|y|}dy$.

Is this operator bounded? If it is, then is it also compact?

I got stuck in simply applying Cauchy-Schwarz inequality to show it is bounded.

Any hint on this problem? Thanks in advance!


1, I have tried to set $f_n(y)=\frac{1}{\sqrt{n}}1_{[0,n]}$, then $\{||Tf_n||_2\} $ are uniformly bounded, so I guess $T$ is a bounded operator.

More details as following:

As above, take $f_n(y)=\frac{1}{\sqrt{n}}1_{[0,n]}$, then $(Tf_n)(x)=\frac{1}{\sqrt{n}}\int_0^n\frac{1}{1+|x|+y}dy$,

so $||Tf_n||_2^2=\int_{\mathbb{R}^1}\frac{1}{n}|\int_0^n\frac{1}{1+|x|+y}dy|^2dx=\int_{\mathbb{R}^1}\frac{1}{n}(ln(1+|x|+y)|_{y=0}^{y=n})^2dx=\int_{\mathbb{R}^1}\frac{1}{n}(ln(\frac{1+|x|+n}{1+|x|}))^2dx=\frac{2}{n}\int_0^{\infty}(ln(\frac{1+x+n}{1+x}))^2dx=\frac{2}{n}\int_1^{\infty}(ln(1+\frac{n}{y}))^2dy\overset{z=\frac{n}{y}}{=}2\int_{0}^n\frac{ln(1+z)^2}{z^2}dz\leq 2\int_{0}^{\infty}\frac{ln(1+z)^2}{z^2}dz$, while since $z\to 0, \frac{ln(1+z)^2}{z^2}\to 1; z\to\infty, \frac{ln(1+z)^2}{z^2}\leq \frac{1}{z^{1+\epsilon}}, \forall 0<\epsilon<1$, so value of this integral is finite.

2, Note that the kernel $K(x,y)=\frac{1}{1+|x|+|y|}$ appearing here does not belong to $L^2(\mathbb{R})$; otherwise, all the problems can be solved, for example, see Hilbert-Schmidt operator

share|cite|improve this question
$\frac{f(x)}{1+|x|+|y|}\leq f(x)$ – akkkk Dec 17 '12 at 16:41
@akkkk, but $f$ might not be in $L^1(\mathbb{R})$. – ougao Dec 17 '12 at 16:56
Are you sure it is $f(x)$ and not $f(y)$ inside the integral defining $T$? – Julián Aguirre Dec 17 '12 at 17:29
@Julián, sorry, it should be $f(y)$, I have fixed it. – ougao Dec 17 '12 at 17:35
What did you find for $\lVert Tf_n\rVert$ in the first remark? Can you show your computations? – Davide Giraudo Dec 18 '12 at 12:28
up vote 8 down vote accepted

$T$ is bounded but not compact.

To see $T$ is bounded, given $f\in L^2(\mathbb{R})$, for the change of variables $y=xt$(when $x\ne 0$ is fixed), we have $$|Tf(x)|\le\int_{\mathbb{R}}\frac{|f(y)|}{|x|+|y|}dy=\int_{\mathbb{R}}\frac{|f(xt)|}{1+|t|}dt.\tag{1}$$ Applying Minkowski's integral inequality to $(1)$ and using the changes of variables $x=\frac{y}{t}$(when $t\ne 0$ is fixed) and $t=s^2$ successively, we have $$\|Tf\|_2\le\int_{\mathbb{R}}\frac{\big(\int_{\mathbb{R}}|f(xt)|^2dx\big)^{\frac{1}{2}}dt}{1+|t|}=\int_{\mathbb{R}}\frac{\big(\int_{\mathbb{R}}|f(y)|^2dy\big)^{\frac{1}{2}}dt}{|t|^\frac{1}{2}(1+|t|)}=\int_{\mathbb{R}}\frac{2\|f\|_2ds}{1+s^2}=2\pi\|f\|_2.$$ It follows that $\|T\|_2\le 2\pi$, i.e. $T$ is bounded.

To see $T$ is not compact, as in your question, let $f_n(y)=\frac{1}{\sqrt{n}}1_{[0,n]}$. Since $\|f_n\|_2=1$, it suffices to show that $\{Tf_n\}$ has no Cauchy subsequence in $L^2(\mathbb{R})$. By definition, $$Tf_n(x)=\frac{1}{\sqrt{n}}\log(1+\frac{n}{1+|x|}).$$ Given $n\in\mathbb{N}$, let $m\gg n$, say $m\ge 144n$. When $m\le |x|\le 2m$,


and $$|Tf_n(x)|\le \frac{1}{\sqrt{n}}\log(1+\frac{n}{m})\le\frac{\sqrt{n}}{m}\le\frac{1}{12\sqrt{m}}.$$

It follows that when $m\le |x|\le 2m$, $|Tf_m(x)-Tf_n(x)|\ge\frac{1}{4\sqrt{m}}$, and hence

$$||Tf_m-Tf_n||_2\ge 2\big(\int_m^{2m}|Tf_m-Tf_n|^2dx\big)^{\frac{1}{2}}\ge \frac{1}{2}.\tag{2}$$ $(2)$ implies that for any subsequence $\{f_{n_k}\}$ of $\{f_n\}$, $\{Tf_{n_k}\}$ cannot be a Cauchy sequence in $L^2(\mathbb{R})$, i.e. $T$ is not compact.

share|cite|improve this answer
The integration is incorrect: $$\int_{\mathbb{R}} \frac{dy}{(1+|x|+|y|)^2}=\frac{2}{1+|x|}.$$ – Julián Aguirre Dec 17 '12 at 17:45
@JuliánAguirre: Thank you. – 23rd Dec 17 '12 at 17:48
+1 Very nice... – copper.hat Dec 19 '12 at 16:04
@copper.hat: Thank you! – 23rd Dec 19 '12 at 16:14
@richard, thank you !!! – ougao Dec 21 '12 at 18:09

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.