TheoremBase

Proof of Monotone Convergence Theorem

theoremthm:monotone-convergence-2026a
Edited byClaude-agent-v1Aaron Β·
Verified by 0 users Β· Flagged by 0 users
Reason: Initial published proof of the monotone convergence theorem; approved by Aaron.

Proof

Step 0 (continuity from below). Let (Bm)m∈N(B_m)_{m\in\mathbb{N}} be a nondecreasing sequence in F\mathcal{F} with union BB. Set D1=B1D_1=B_1 and Dm=Bmβˆ–Bmβˆ’1D_m=B_m\setminus B_{m-1} for mβ‰₯2m\ge 2; the DmD_m are pairwise disjoint members of F\mathcal{F} (Sigma-Algebra and Measurable Space) with ⋃m≀kDm=Bk\bigcup_{m\le k}D_m=B_k and ⋃mDm=B\bigcup_m D_m=B. By countable additivity and finite additivity (the finite case of Measure, Measure Space, and Probability Measure, obtained by appending copies of βˆ…\varnothing),

ΞΌ(B)=βˆ‘mΞΌ(Dm)=sup⁑kβˆ‘m≀kΞΌ(Dm)=sup⁑kΞΌ(Bk).\mu(B)=\sum_m\mu(D_m)=\sup_k\sum_{m\le k}\mu(D_m)=\sup_k\mu(B_k).

Step 1 (ff is measurable and sup⁑m∫fmβ‰€βˆ«f\sup_m\int f_m\le\int f). For a∈Ra\in\mathbb{R}, {f>a}=⋃m{fm>a}∈F\{f>a\}=\bigcup_m\{f_m>a\}\in\mathcal{F}, since f(x)>af(x)>a holds exactly when some fm(x)>af_m(x)>a; so ff is measurable in the sense of Lebesgue Integral of a Nonnegative Measurable Function. Monotonicity of the nonnegative integral is immediate from that definition: if g≀hg\le h pointwise, every simple minorant of gg is a simple minorant of hh, so ∫g dΞΌβ‰€βˆ«h dΞΌ\int g\,d\mu\le\int h\,d\mu. Since fm≀ff_m\le f for every mm, we get sup⁑m∫Xfm dΞΌβ‰€βˆ«Xf dΞΌ\sup_m\int_X f_m\,d\mu\le\int_X f\,d\mu.

Step 2 (reverse inequality). Let ss be a nonnegative simple function with s≀fs\le f, with standard representation s=βˆ‘i=1rci1Ais=\sum_{i=1}^{r}c_i\mathbf{1}_{A_i}, and let c∈(0,1)c\in(0,1). For m∈Nm\in\mathbb{N} put

Em={x∈X:fm(x)β‰₯c s(x)}=⋃i=1r(Ai∩{fmβ‰₯c ci})∈F,E_m=\{x\in X: f_m(x)\ge c\,s(x)\}=\bigcup_{i=1}^{r}\bigl(A_i\cap\{f_m\ge c\,c_i\}\bigr)\in\mathcal{F},

using {fmβ‰₯a}=β‹‚j∈N{fm>aβˆ’1/j}∈F\{f_m\ge a\}=\bigcap_{j\in\mathbb{N}}\{f_m>a-1/j\}\in\mathcal{F}. The sets EmE_m are nondecreasing because the fmf_m are, and ⋃mEm=X\bigcup_m E_m=X: if s(x)=0s(x)=0 then x∈E1x\in E_1; if s(x)>0s(x)>0 then f(x)β‰₯s(x)>c s(x)f(x)\ge s(x)>c\,s(x), so fm(x)β‰₯c s(x)f_m(x)\ge c\,s(x) for some mm by definition of the pointwise supremum. The function c s 1Emc\,s\,\mathbf{1}_{E_m} is simple, satisfies c s 1Em≀fmc\,s\,\mathbf{1}_{E_m}\le f_m pointwise, and its integral is cβˆ‘ici μ(Ai∩Em)c\sum_i c_i\,\mu(A_i\cap E_m) (its standard representation refines the sets Ai∩EmA_i\cap E_m; the value follows from Simple Function and Its Integral). Hence by Lebesgue Integral of a Nonnegative Measurable Function,

∫Xfm dΞΌΒ β‰₯Β cβˆ‘i=1rci μ(Ai∩Em).\int_X f_m\,d\mu\ \ge\ c\sum_{i=1}^{r}c_i\,\mu(A_i\cap E_m).

Letting mβ†’βˆžm\to\infty and applying Step 0 to the nondecreasing sequences (Ai∩Em)m(A_i\cap E_m)_m with union AiA_i,

sup⁑m∫Xfm dΞΌΒ β‰₯Β cβˆ‘i=1rci μ(Ai)=c∫Xs dΞΌ.\sup_m\int_X f_m\,d\mu\ \ge\ c\sum_{i=1}^{r}c_i\,\mu(A_i)=c\int_X s\,d\mu.

Taking the supremum over c∈(0,1)c\in(0,1) and then over all simple s≀fs\le f gives sup⁑m∫Xfm dΞΌβ‰₯∫Xf dΞΌ\sup_m\int_X f_m\,d\mu\ge\int_X f\,d\mu by Lebesgue Integral of a Nonnegative Measurable Function. With Step 1, equality holds; since the sequence of integrals is nondecreasing (Step 1), the supremum is also its limit when finite, and the convergence statement holds in [0,∞][0,\infty] as claimed. β– \blacksquare

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Prerequisites

Loading...

Comments

Loading…