Paper 4, Section II, J

Probability and Measure
Part II, 2012

State and prove Fatou's lemma. [You may use the monotone convergence theorem.]

For (E,E,μ)(E, \mathcal{E}, \mu) a measure space, define L1:=L1(E,E,μ)L^{1}:=L^{1}(E, \mathcal{E}, \mu) to be the vector space of μ\mu integrable functions on EE, where functions equal almost everywhere are identified. Prove that L1L^{1} is complete for the norm 1\|\cdot\|_{1},

f1:=Efdμ,fL1.\|f\|_{1}:=\int_{E}|f| d \mu, \quad f \in L^{1} .

[You may assume that 1\|\cdot\|_{1} indeed defines a norm on L1L^{1}.] Give an example of a measure space (E,E,μ)(E, \mathcal{E}, \mu) and of a sequence fnL1f_{n} \in L^{1} that converges to ff almost everywhere such that fL1f \notin L^{1}.

Now let

D:={fL1:f0 almost everywhere ,Efdμ=1}\mathcal{D}:=\left\{f \in L^{1}: f \geqslant 0 \text { almost everywhere }, \int_{E} f d \mu=1\right\} \text {. }

If a sequence fnDf_{n} \in \mathcal{D} converges to ff in L1L^{1}, does it follow that fD?f \in \mathcal{D} ? If fnDf_{n} \in \mathcal{D} converges to ff almost everywhere, does it follow that fDf \in \mathcal{D} ? Justify your answers.