I'm doing some practice questions for a real analysis qualifying exam coming up in a few weeks. I have a couple questions, namely on the "if the statement is true, prove it. Otherwise, give a counterexample" questions.

a) In an infinite-dimensional Hilbert space $H$, for any weakly convergent sequence $\left\lbrace x_n \right\rbrace$, there exists a subsequence that is convergent with respect to the norm

b) Since two iterated integrals exist and $$\int_{(0,1)}\int_{(0,1)} \frac{x^2-y^2}{(x^2+y^2)^2} dm(x)dm(y) = \int_{(0,1)}\int_{(0,1)} \frac{x^2-y^2}{(x^2+y^2)^2} dm(y)dm(x)$$ we can conclude, via the Tonelli-Fubini theorem, that the double integral exists.

c) There exists a function $f \geq 0$ on $(0, \infty))$ such that $f \in L^p((0,\infty)$ if and only if $p=1$

I have no idea how to solve part a.

For b I think the answer is true since I can switch the order of integration if the inside integral is finite, and since the integral exists by assumption, that solves b.

For c, I think the answer should be true and the function should be some modified version of $\frac{1}{x}$. The space on the paper for this answer is rather short, so there should be simple counterexample of this form I would think, but I can't construct it properly.

Any help would be tremendously appreciated!-


a. $\{\sin(n\cdot)\}$ converges weakly to $0$ in $L^2([0,2\pi])$ but $$\int_0^{2\pi}\sin^2(nx)\,dx = \pi - \frac{\sin(4\pi n)}{4n} \ge 1.$$ To show that this provides a counterexample to the statement recall that if a subsequence converges strongly then it converges strongly to the weak limit, which in this case is $0$. Hence we would have $$1 \le \lim_{k\to \infty}\|\sin(n_k\cdot)\|_{L^2} = 0.$$

b. The iterated integrals are not equal.

c. Let $$ f = \begin{cases} \frac{1}{(\log x)^2 x} & \text{if}\ x \in (0,0.5) \\ \frac{1}{x^2} &\ \text{otherwise.} \end{cases} $$

Then $f$ is positive and integrable, but $f \notin L^p$ for any $p > 1$ since the singularity at $0$ is not integrable. Indeed let $p = 1 + 2\epsilon$, with $\epsilon > 0$. Then, for $x$ small, $$|f(x)|^p = \frac{1}{x^{1 + \epsilon}}\frac{1}{|\log x|^{2p}x^{\epsilon}} \ge \frac{1}{x^{1 + \epsilon}} \notin L^1.$$ The last inequality follows from the fact that $$\lim_{x \to 0} (\log x)^{2p}x^{\epsilon} = 0,$$ so that we can find $\delta < \frac 12$ such that if $x < \delta$ then $$\frac{1}{|\log x|^{2p}x^{\epsilon}} \ge 1.$$

  • $\begingroup$ Can you explain how the last function is in $L^1$ but not in $L^p$? I see that it's in $L^1$, but not why it isn't in $L^p$: $\int_0^{0.5} |\frac{1}{x(logx)^2}|^p dx = \int_{-\infty}^{log(0.5)} |\frac{1}{u^{2p}}| dx$ which is still finite $\endgroup$ – Brenton May 1 '16 at 3:15
  • $\begingroup$ @Brenton: I have edited with a proof that the integral blows up. Make sure to double check my computations, but I think it works :) $\endgroup$ – Giovanni May 1 '16 at 3:32
  • $\begingroup$ @Brenton: check the change of variables again: the substitution $u = \log x$ doesn't give the integral that you get, unless I am mistaken. $\endgroup$ – Giovanni May 1 '16 at 3:37
  • $\begingroup$ You're right, I think I made a mental mistake doing it instead of writing it out carefully $\endgroup$ – Brenton May 1 '16 at 15:46

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.