Past exam of the integration course of the University of Oxford / 2014 / 1 / b

Let $S_N(x)={\sum }_{n=1}^N|f_n(x)|$. These are nonnegative measurable functions increasing to the extended-valued function $S(x)={\sum }_{n=1}^{\infty }|f_n(x)|$. The monotone convergence theorem givesTherefore $S$ is finite almost everywhere: if $S=\infty$ on a measurable set of positive Lebesgue measure, its integral would exceed every finite bound. Hence the series ${\sum }_n{f}_n(x)$ converges absolutely outside the null set $N=\{S=\infty \}$. This is summable L1 norms imply almost-everywhere absolute convergence.

Define $s(x)={\sum }_n{f}_n(x)$ off $N$ and set $s=0$ on $N$. The function is Lebesgue measurable, being the pointwise limit of measurable partial sums off a measurable null set. Since $|s|\le S$ almost everywhere,so the sum is Lebesgue integrable.

For $s_N={\sum }_{n=1}^N{f}_n$, the absolute tail satisfiesApply the monotone convergence theorem to the nonnegative tail partial sums. We obtainThus the partial sums converge in L1 norm. The triangle inequality for an integral yieldsConsequentlyThe scalar series on the right also converges absolutely, since ${\sum }_n|{\int }_E{f}_n|\le {\sum }_n\Vert f_n{\Vert }_1<\infty$. This proves integration of a series with summable L1 norms without needing pointwise convergence on the exceptional null set.