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

Let $f_n,f \in L^p$ and $f_n \to f$ uniformly. Further let $g \in L^q$. Now, clearly $\int (f_n - f) g$ is defined. How do I show that

$$\int |f_n - f||g|$$

can be made arbitrary small if I choose $n$ large enough? If we just naively apply Hölder we will not be able to make it small but just finite. Further, we cannot just replace $|f_n - f|$ by the bound "$\epsilon$" because that is larger than this. Does someone have a hint?

share|cite|improve this question
Try choosing a set $A$ of finite measure for which $\int_{A^c} |g|^q$ is small. – Nate Eldredge Nov 21 '10 at 14:44
up vote 6 down vote accepted

Outside of a sufficiently large compact set, $\|g\|_q <\epsilon$. Use Hölder naively, and on that portion the integral is bounded above by $(\|f_n\|_p + \|f\|_p)\epsilon$. On the inside of the compact set $\|g\|_1 \leq C \|g\|_q$ where $C$ is given by the volume of the compact set. Then you can replace $|f_n - f|$ by $\epsilon$.

Note that this requires $\|f_n\|_p$ be uniformly bounded, which would be true if you additionally assume $f_n \to f$ in $L^p$. If you don't assume that:

Consider over $\mathbb{R}$, take $f_n = \frac1n \chi_{[2^n,2^{n+1}]}$, where $\chi$ denotes the characteristic function. $f_n \to 0$ uniformly as a sequence of functions (i.e. $f_n\to f$ in $L^\infty$). For this sequence $\|f_n\|_p = \frac1n \cdot 2^{n/p} \nearrow \infty$ (unless $p = \infty$). (Note that since it is not a bounded sequence, you cannot assert that it has a weakly converging subsequence.)

Now take $g$ be the function such that $g(x) = 2^{-n/(q-\delta)}$ if $2^n < |x| \leq 2^{n+1}$. By construction it is a function in $L^q$.

But $|f_n - f| |g| = \frac1n \cdot 2^{-n / (q - \delta)} \chi_{[2^n,2^{n+1}]}$. If you integrate it, you get

$$ \frac1n \cdot 2^{n( 1 - \frac{1}{q - \delta}) }$$

So unless $q = 1$ and $p = \infty$, you can always choose $\delta$ sufficiently small such that the exponent in the integrated expression is positive, and provide a counterexample to what you want proved. (And of course, in the case where $p = \infty$, you can just replace $|f-f_n|$ by $\epsilon$ and be done with it.

share|cite|improve this answer
Thanks. I got the case where the $f_n$ are uniformly bounded in $L^p$ but not where this is not true. Is it possible to give an intuitive meaning why it should be false for the $f_n$ not uniformly bounded in $L^p$? – Jonas Teuwen Nov 21 '10 at 15:54
Hum, that's a good question. It has to do with the interplay between compactness of the domain and the compactness inside the function space. See, if instead of $\mathbb{R}^d$, you worked on a space with finite total measure, then uniform convergence in $L^\infty$ will imply boundedness in $L^p$ also. This means that any possible modes of failure must take advantage of the fact that the support of $f_n$ can run off to infinity. – Willie Wong Nov 21 '10 at 16:01
Now, if $f_n$ are uniformly bounded, and their supports run off to infinity, then they converge weakly to zero, which means that what you want is also true. So necessarily a counterexample will require $f_n$ to be unbounded. As to an intuitive meaning: I guess one way to think about it is that while on sets of finite measure, the topology given by $L^p$ is finer than that of $L^q$ if $p > q$, on sets of infinite measure, it is possible that the topology given by $L^p$ and $L^q$ be not comparable for $p\neq q$. – Willie Wong Nov 21 '10 at 16:06
(Sorry for the poor explanation. The extent to which this is intuitive to me is just that there are certain lack of inclusions between Lebesgue spaces on non-compact domains, and a bit of concentration compactness for sequences in Lebesgue spaces. Which is to say, not very intuitive at all.) – Willie Wong Nov 21 '10 at 16:10

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.