Throughout fix m≥1 and write sp=Tp2−m for p=0,1,…,2m, so that 0=s0<s1<⋯<s2m=T, Im,p=[sp−1,sp) for p<2m, and Im,2m=[s2m−1,T]. Write λ=λ[0,T] and B=B[0,T].
Claim 1. Partition. Let t∈[0,T). Then 0≤T−1t<1, so there is exactly one p∈{1,…,2m} with sp−1≤t<sp, namely the least element p of {1,…,2m} with t<sp, a set that is nonempty because t<T=s2m; indeed sp−1≤t holds for that p, when p=1 because s0=0≤t and when p>1 by minimality, and the strict increase of s0<s1<⋯<s2m makes p unique. Hence t lies in Im,p for exactly one p (noting that for p=2m the condition s2m−1≤t<s2m does place t in Im,2m=[s2m−1,T]). The remaining point t=T lies in Im,2m and, since sp<T for p<2m, in no other atom. So the atoms are pairwise disjoint with union [0,T].
Measurability and measure. Each atom is the intersection with [0,T] of an interval of R, and intervals are Borel sets; so each atom lies in B by the description B={S∩[0,T]:S Borel} in claim 1 of the restricted Lebesgue toolkit. Since Lebesgue measure assigns to an interval its length, λ(Im,p)=sp−sp−1=T2−m, whence PT(Im,p)=T−1⋅T2−m=2−m.
Gm is a σ-algebra. By definition Gm={⋃p∈SIm,p:S⊆{1,…,2m}}. It contains ∅ (take S=∅) and [0,T] (take S to be everything). Because the atoms partition [0,T], the complement of ⋃p∈SIm,p in [0,T] is ⋃p∈/SIm,p, again a member. A union of countably many members corresponds to the union of the associated index sets, hence is a member. So Gm is a σ-algebra on [0,T], and it is contained in B because each atom is and B is closed under finite unions.
Refinement. Let 1≤p<2m. Then Im+1,2p−1=[T(2p−2)2−(m+1),T(2p−1)2−(m+1)) and Im+1,2p=[T(2p−1)2−(m+1),T(2p)2−(m+1)), both half-open because 2p<2m+1; their union is [T(2p−2)2−(m+1),T(2p)2−(m+1))=[sp−1,sp)=Im,p. For p=2m, Im+1,2m+1−1=[T(2m+1−2)2−(m+1),T(2m+1−1)2−(m+1)) and Im+1,2m+1=[T(2m+1−1)2−(m+1),T], whose union is [T(2m−1)2−m,T]=Im,2m. Hence every level-m atom is a union of level-(m+1) atoms, so every union of level-m atoms is one of level-(m+1) atoms, giving Gm⊆Gm+1. Iterating, for m′≥m every level-m atom is a union of level-m′ atoms; since the level-m′ atoms are pairwise disjoint with union [0,T] and every level-m′ atom is nonempty, each level-m′ atom meets, hence is contained in, exactly one level-m atom.
Claim 2. Write G∞=σ(⋃m≥1Gm).
The inclusion G∞⊆B. By claim 1 each Gm is contained in B, so B is a σ-algebra on [0,T] containing ⋃mGm; by the minimality in the definition of the generated σ-algebra, G∞⊆B.
The inclusion B⊆G∞. Put H={S⊆R:S∩[0,T]∈G∞}. This is a σ-algebra on R: R∩[0,T]=[0,T]∈G1⊆G∞; if S∈H then (R∖S)∩[0,T]=[0,T]∖(S∩[0,T])∈G∞; and if S1,S2,⋯∈H then (⋃nSn)∩[0,T]=⋃n(Sn∩[0,T])∈G∞.
Let U⊆R be open for the metric (s,t)↦∣s−t∣. For m≥1 let Vm be the union of those level-m atoms that are contained in U; then Vm∈Gm⊆G∞. We show U∩[0,T]=⋃m≥1Vm. Each Vm is contained in U and in [0,T], which gives one inclusion. For the other, let t∈U∩[0,T] and choose a real ε>0 with {s∈R:∣s−t∣<ε}⊆U. Since 2m≥m for every natural m≥1 and the real numbers are Archimedean, there is m≥1 with T2−m<ε. By claim 1 there is a (unique) p with t∈Im,p, and every s∈Im,p satisfies ∣s−t∣≤T2−m<ε, because Im,p is contained in an interval of length T2−m containing t. Hence Im,p⊆U, so Im,p is one of the atoms forming Vm and t∈Vm. Therefore U∩[0,T]∈G∞, that is U∈H.
So H is a σ-algebra on R containing every open set. Since the Borel σ-algebra is the σ-algebra generated by the open sets, minimality gives B(R)⊆H; that is, S∩[0,T]∈G∞ for every Borel S. As B={S∩[0,T]:S Borel}, we conclude B⊆G∞, and with the previous inclusion, G∞=B.
Claim 3. Each 1Im,p is a bounded B-measurable function, hence a square-integrable random variable on ([0,T],B,PT); by the definition of square-integrability, the product X1Im,p of two square-integrable random variables is integrable, so ap:=2mE[X1Im,p] is a well-defined real number and AmX=∑p=12map1Im,p.
(i) Gm-measurability. AmX takes the value ap on Im,p and, the atoms partitioning [0,T], no other values. For a Borel set B, (AmX)−1(B)=⋃{Im,p:ap∈B}∈Gm.
(ii) Square-integrability. With C=maxp∣ap∣, a maximum over a finite set, ∣AmX∣≤C everywhere, so (AmX)2≤C2 and E[(AmX)2]≤C2<∞ by monotonicity of the integral and PT([0,T])=1. Thus AmX is bounded and square-integrable.
(iii) The defining identity. Let A∈Gm, say A=⋃p∈SIm,p with S⊆{1,…,2m}. By disjointness of the atoms, 1A=∑p∈S1Im,p pointwise, and likewise AmX1A=∑p∈Sap1Im,p. By linearity of the integral and E[1Im,p]=PT(Im,p)=2−m from claim 1,
E[AmX1A]=p∈S∑ap2−m=p∈S∑E[X1Im,p]=E[Xp∈S∑1Im,p]=E[X1A].
Conditions (i), (ii) and (iii) of the definition of conditional expectation are therefore satisfied, so AmX is a conditional expectation of X given Gm.