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

I have the following question about $L^p$ spaces:

Suppose that $f,f_1,f_2,\ldots$ are functions in $L^p$ for some $p \geq 1$ and the sequence converges in $L^p$ to $f$, i.e. $||f_n-f||_p \to 0$.

Does this imply that the sequence $|f_1|^p,|f_2|^p,\ldots$ converges in $L^1$ to $|f|^p$?

Is it also true if we remove the absolute value? That is, the sequence $f_1^p,f_2^p,...$ converges to $f^p$, in case $p$ doesn't create problem with power (for example $p=3$).

I am using necessary and sufficient condition from the Vitali's convergence theorem and it seems to work.

Do you have some advices or some references for this result?

Many thanks

share|cite|improve this question
What are you asking? – checkmath Mar 30 '12 at 15:12
Is true that $||\,|f_n|^p-|f|^p \, ||_1 \to 0$? – Kolmo Mar 30 '12 at 15:19
Hint you have $\int |f_n|^p \to \int |f|^p$. – checkmath Mar 30 '12 at 15:19
So if you have $|f_n|$ dominated by $|f|$ then it follows directly! – checkmath Mar 30 '12 at 15:26
@chessmath you mean $|f_n| \leq |f|$ a.e for all n, then $|| \, |f_n|^p -|f|^p \, || = \int |f|^p-|f_n|^p $ and it is ok. But I dont have that dominance. – Kolmo Mar 30 '12 at 15:33
up vote 8 down vote accepted

If $a,b\in \mathbb{R}$ and $p\ge1$, then it follows from the mean value theorem that $$ |a^p-b^p|\le p\max(|a|,|b|)^{p-1}|a-b|\le p(|a|+|b|)^{p-1}|a-b|. $$ Let $q$ be the conjugate exponent of $p$. Then $$ \int\bigl|\,|f_n|^p-|f|^p\,\bigr|\le p\int\bigl(|f_n|+|f|\bigr)^{p-1}|f_n-f|\le p\Bigl(\int\bigl(|f_n|+|f|\bigr)^{(p-1)q}\Bigr)^{1/q}\Bigl(\int|f_n-f|^{p}\Bigr)^{1/p}, $$ that is $$ \||f_n|^p-|f|^p\|_1\le p(\||f_n|+|f|\|_p)^{p/q}\|f_n-f\|_p\le p(\|f_n\|_p+\|f\|_p)^{p/q}\|f_n-f\|_p, $$ proving that $|f_n|^p\to|f|^p$ in $L^1$.

Observe that we have proved that the operators $T_1, T_2\colon L^p\to L^1$ defined by $T_1(f)=|f|^p$ and $T_2(f)=f^p$( when $T_2$ makes sense) are locally Lipschitz.

share|cite|improve this answer
Thanks very much Julian. So direct and clean. Thanks again. – Kolmo Mar 31 '12 at 9:13

If we assume that the result is not true, we would be able to find a subsequence $\{f_{n_k}\}$ and a $\delta>0$ such that $\lVert |f_{n_k}|^p-|f|^p\rVert_{L^1}\geq \delta_0$. Taking if necessary a further subsequence, we can assume that $f_{n_k}$ converges to $f$ almost everywhere. We denote $g_k:=|f_{n_k}|^p$, and let $h_k:=|g_k|+|f|^p-|g_k-|f|^p|$. Then almost everywhere, $h_k$ converges to $2|f|^p$, and since $h_k\geq 0$, by Fatou lemma: $$2\lVert f\rVert_{L^p}^p\leq \liminf_{k\to +\infty}\int (|g_k|-|g_k-|f|^p|)+\int|f|^p$$ and since $\int g_k\to \int |f|^p$ we have that $\limsup_{k\to \infty}\int |g_k-|f||^p=\limsup_{k\to \infty}\int ||f_{n_{k}}|^p-|f|^p|=0$. I followed a method similar to here.

For $f_n^p$ just choose $g_k =f_{n_k}^p$.

share|cite|improve this answer
If I take a=2, b=1 and p=2 we have $|2^2-1^2|=3>1=|2-1|^2$ – Kolmo Mar 30 '12 at 15:48
@Kolmo My attempt was wrong, I hope this one is correct. – Davide Giraudo Mar 30 '12 at 16:17
thanks for the responce. I have a doubt. This reasoning shows that $\limsup_{k\to\infty} \int |\ |f_{n_k}|^p-|f|^p \ | \leq 0$. Why this should be true for the whole sequecence? – Kolmo Mar 31 '12 at 9:02
Let $a_n:=\int ||f_n|^p-|f_n||$. We have shown that for each subsequence of $\{a_n\}$ we can find a subsequence which converges to $0$. – Davide Giraudo Mar 31 '12 at 10:53
so $a_{n_k}=\int ||f_{n_k}|^p-|f|^p|$. Since $f_n$ converge in $L^p$, the same does $f_{n_k}$ and so we know that exists a subsequence of $f_{n_k}$ that converge a.e. to f. Then if we say $h_l=|f_{n_{k_l}}|^p$ this converges a.e. to $|f|^p$ and $||h_l||_1 \to |||f|^p||_1$. So by the trick you linked $a_l \to 0$. Did I understand correct. I guess yes because it works. Thanks. – Kolmo Mar 31 '12 at 11:14

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.