Let $ 1 \leq p_1 < p_2 < \infty$, and suppose that $f_n$ is a sequence of functions in $L^{p_1}[a,b]$ such that $f_n \to f$ pointwise a.e. on $[a,b]$. Suppose in addition that $ ||f_n||_{p_2} \leq 1$ for every $n$, where $|| \cdot ||_{p_2}$ denotes the $L^{p_2}$ norm, then how can we show that $f_n \to f$ strongly in $L^{p_1}$ ?

What I have tried:

Since $ 1 < p_2 < \infty$ and $\{f_n\}$ has bounded $L^{p_2}$ norm, it follows that $f_n \to f$ weakly in $L^{p_2}$. Now since $[a,b]$ has finite measure, if $f_n \to f$ strongly in $L^{p_2}$, then $f_n \to f$ strongly in $L^{p_1}$ also, so this is what I am trying to show, although I do not know whether we actually do have strong convergence in $L^{p_2}$.

Given $f_n \to f$ weakly in $L^{p_2}$, we have strong convergence in $L^{p_2}$ if and only if the norms converge, i.e. $||f_n||_{p_2} \to ||f||_{p_2}$. But how can we argue that we do have convergence of the norms in this case? Maybe using the condition that $ ||f_n||_{p_2} \leq 1$ ?

  • 1
    $\begingroup$ In general strong convergence in $L^{p_2}$ will not hold, so that route won't work. $\endgroup$ – Nate Eldredge Jan 22 '13 at 1:51
  • 1
    $\begingroup$ You could mimic the proof in the first answer here. In place of Cauchy-Schwarz, use (or prove) $(\int_A |g|^{p_1})^{1/p_1}\le (\int_A |g|^{p_2})^{1/p_2}\cdot\bigl(\mu(A)\bigr)^{1/p_1-1/q_1}$ for $g\in L_{p_2}$. $\endgroup$ – David Mitra Jan 22 '13 at 2:01

For simplicity, consider the case $p_1 = 1$. (The general case follows, by replacing $f_n$ by $|f_n|^{p_1}$ and $p_2$ by $p_2/p_1$.)

Show that the condition $\|f_n\|^{p_2} \le 1$ implies that $\{f_n\}$ is uniformly integrable. Then use Vitali's convergence theorem to conclude that $f_n$ converges in $L^1$.

  • $\begingroup$ 1) Does this argument work for $L^1(\Omega)$, $\Omega \subset \mathbb{R}^n$ either bounded or unbounded? 2) If $p_2=p_1$ (=1) then can we claim $f_n$ converges WEAKLY in $L^1$ to $f$? $\endgroup$ – Behnam Esmayli Jul 24 '18 at 21:39
  • $\begingroup$ @BehnamEsmayli: (1) It works if $\Omega$ has finite measure (and in particular if $\Omega$ is bounded). Otherwise you can pretty easily find counterexamples. (2) No, consider for example $[a,b] = [0,1]$, $f_n = n 1_{(0, 1/n)}$, $f = 0$. $\endgroup$ – Nate Eldredge Jul 28 '18 at 20:13
  • $\begingroup$ But doesn't that seq in (2) converge weakly to $f$? $\endgroup$ – Behnam Esmayli Jul 31 '18 at 2:33
  • $\begingroup$ @BehnamEsmayli: If it did, then $\int f_n g$ should converge to $\int fg$ for every $g \in L^\infty([0,1])$, since integration against $g$ is a continuous linear functional on $L^1$. Try it with $g=1$. $\endgroup$ – Nate Eldredge Jul 31 '18 at 2:45

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.