This is the exercise 5 in page 109 of Analysis III of Amann and Escher.

Let $f_j,f:X\to\overline{\Bbb R}^+$ measurable functions such that $(f_j)\to f$ in measure. Show that

$$\int_Xf\,d\mu\le\varliminf\int_X f_j\,d\mu$$

Note: $X$ is a $\sigma$-finite space.

From a previous result I know that if $(f_j)\to f$ in measure then there is a subsequence $(f_{j_k})\to f$ almost everywhere.

My work so far: I set $g_j:=\inf_{k\ge j} f_j$. Then $(g_j)$ is increasing and by the existence of $(f_{j_k})$ we knows that $g_j\le f$ almost everywhere, thus WLOG we can consider that $g_j\le f$ in $X$.

Then by Fatou's lemma we have that $$ \int_X\varliminf f_j\,d\mu=\int_X \lim g_j\,d\mu\le\varliminf\int_X f_j\,d\mu\tag1 $$ Then if we shows that $(g_j)\to f$ almost everywhere we are done. However I was unable to prove it, I set $$ L:=\{x\in X: \lim |f(x)-g_j(x)|=0\}\\ A_{j,n}:=\{x\in X:|f(x)-f_j(x)|\ge 1/n\}\\ B_n:=\{x\in X:|f(x)-g_j(x)|\ge 1/n,\;\forall j\in\Bbb N\} $$ Then I tried to show that $L^\complement$ is a null set from the convergence in measure of $(f_j)$, trying to see if the $B_n$ are null, but I didnt find a way, and Im thinking that it is not necessarily true.

Then I started a different approach: if $\int_X f\le\varliminf\int_X f_j$ then eventually $\int_X f\le\int_X f_j$. This means that if $\int_X f=\infty$ then eventually $\int_X f_j=\infty$, and if $\int_X f=K<\infty$ then eventually $\int_X f_j\ge K$.

But from here I doesn't find something useful. Some help will be appreciated, thank you.


There exists a subsequence $(f_{j_k})_{k \in \mathbb{N}}$ of $(f_j)_{j \in \mathbb{N}}$ such that

$$\liminf_{j \to \infty} \int_X f_j \, d\mu = \lim_{k \to \infty} \int_X f_{j_k} , d\mu. \tag{1}$$

Since $f_{j_k} \to f$ in measure, we can take a further subsequence $f_{j_{k_{m}}}$ which converges almost everywhere to $f$. Applying Fatou's lemma we get

$$\int_X f \, d\mu = \int_X \lim_{m \to \infty} f_{j_{k_m}} \, d\mu \leq \liminf_{m \to \infty} \int_X f_{j_{k_m}} \, d\mu \stackrel{(1)}{=} \liminf_{j \to \infty} \int_X f_j \, d\mu.$$


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.