Each result cited is universally quantified over the data in its own statement. Throughout, λ is Lebesgue measure on B(R) and, as in Finite Products of Lebesgue Measure and Coordinate Integration on Rl and Lebesgue Measure on Rn, points of R1 are identified with real numbers, so that B(R1)=B1=B(R) and λ1=λ. For every m∈N the measure λm is σ-finite by Lebesgue Measure on Euclidean Space is Sigma-Finite §sigma-finite; hence for m,n∈N the product measure λm⊗λn on B(Rm)⊗B(Rn) exists by Existence and Uniqueness of the Product Measure, and Tonelli and Fubini Theorems applies to the pair λm,λn.
Step 1 (Composition with a concatenation). Let m,n∈N and let F:Rm+n→[0,∞] be Borel. By Pairs of Euclidean Points: Coordinate Projections, Pairings, the Product Measure on a Euclidean Space, Borel Norm Functions and Finite Sets §product the map ιm,n is measurable with respect to B(Rm)⊗B(Rn) and B(Rm+n). For real c the set {F∘ιm,n>c} is the preimage under ιm,n of {F>c}∈B(Rm+n), so F∘ιm,n is measurable with respect to B(Rm)⊗B(Rn) in the sense of Measure Spaces and the Lebesgue Integral: Standing Notation §measurable. By the Sections and Tonelli parts of Tonelli and Fubini Theorems, for every u∈Rm the map v↦F(ιm,n(u,v)) is Borel on Rn, the map u↦∫RnF(ιm,n(u,v))λn(dv) is Borel on Rm, and
∫Rm×RnF∘ιm,nd(λm⊗λn)=∫Rm(∫RnF(ιm,n(u,v))λn(dv))λm(du).(1)
It therefore suffices to show that for all m,n∈N and every Borel F:Rm+n→[0,∞]
∫Rm+nFdλm+n=∫Rm×RnF∘ιm,nd(λm⊗λn),(2)
because (1) and (2) with m=q, n=p and F=f give all assertions of the lemma. Let S be the set of n∈N such that (2) holds for every m∈N and every Borel F:Rm+n→[0,∞]. We show 1∈S and that n∈S implies n+1∈S; since n+1 is the successor of n by claim 1 of Arithmetic of Addition on the Natural Numbers, Principle of Induction for the Natural Numbers then gives S=N.
Step 2 (1∈S). Let m∈N. Since m+1 is the successor of m and m+1=1 (claims 1 and 7 of Arithmetic of Addition on the Natural Numbers), Finite Products of Lebesgue Measure and Coordinate Integration on Rl with l=m+1 defines Bm+1=Bm⊗B(R) and λm+1=λm⊗λ, transported to Rm+1 along the bijection J:Rm×R→Rm+1, J((θ1,…,θm),t)=(θ1,…,θm,t); thus E∈Bm+1 exactly when J−1(E)∈Bm⊗B(R), and then λm+1(E)=(λm⊗λ)(J−1(E)). By Lebesgue Measure on Rn, B(Rk)=Bk and Lebesgue measure on it is the measure λk of that lemma, for k=m and k=m+1. By the definition of the concatenation in Concatenation Identifies a Product of Euclidean Spaces with a Euclidean Space, ιm,1(u,t) has kth coordinate uk for k∈[m] and (m+1)th coordinate t, so ιm,1=J. Hence λm+1(E)=(λm⊗λ1)((ιm,1)−1(E)) for every E∈B(Rm+1), that is, λm+1 is the image measure of λm⊗λ1 under the measurable map ιm,1, and claim 2 of Image Measures, Measures with Densities, and Change of Variables gives (2) for n=1.
Step 3 (Rebracketing). Let m,n∈N, u∈Rm, w∈Rn and t∈R1. By claim 3 of Arithmetic of Addition on the Natural Numbers, m+(n+1)=(m+n)+1, and we claim
ιm,n+1(u,ιn,1(w,t))=ιm+n,1(ιm,n(u,w),t).(3)
By the description of indices in Concatenation Identifies a Product of Euclidean Spaces with a Euclidean Space, every k∈[m+(n+1)] satisfies exactly one of k∈[m] and k=m+j with a unique j∈[n+1], and by claim 5 of Properties of the Order on the Natural Numbers such a j satisfies j∈[n] or j=n+1. If k∈[m], then k∈[m+n] by claim 6 of Properties of the Order on the Natural Numbers, and both sides of (3) have kth coordinate uk. If k=m+j with j∈[n], the left side has kth coordinate equal to the jth coordinate of ιn,1(w,t), namely wj; and k∈[m+n] by claim 6 of Properties of the Order on the Natural Numbers, so the right side has kth coordinate equal to the kth coordinate of ιm,n(u,w), again wj. If k=m+(n+1)=(m+n)+1, both sides have kth coordinate t. This proves (3).
Step 4 (n∈S implies n+1∈S). Let n∈S, let m∈N and let F:Rm+(n+1)→[0,∞] be Borel. For u∈Rm let gu(w)=F(ιm,n+1(u,w)) for w∈Rn+1 and K(u)=∫Rn+1gudλn+1; by Step 1 (with n+1 in place of n) each gu is Borel and
∫Rm×Rn+1F∘ιm,n+1d(λm⊗λn+1)=∫RmKdλm.(4)
Define G:Rm+n→[0,∞] by G(z)=∫R1F(ιm+n,1(z,t))λ1(dt), which is Borel by Step 1 applied to F on R(m+n)+1=Rm+(n+1). By 1∈S (Step 2) applied with n in place of m to gu, then Step 1 with (n,1) in place of (m,n), and then (3),
K(u)=∫Rn×R1gu∘ιn,1d(λn⊗λ1)=∫Rn(∫R1F(ιm+n,1(ιm,n(u,w),t))λ1(dt))λn(dw)=∫RnG(ιm,n(u,w))λn(dw).
On the other hand, by 1∈S applied with m+n in place of m, Step 1 with (m+n,1), then n∈S applied to G, and Step 1 with (m,n),
∫Rm+(n+1)Fdλm+(n+1)=∫Rm+nGdλm+n=∫Rm×RnG∘ιm,nd(λm⊗λn)=∫Rm(∫RnG(ιm,n(u,w))λn(dw))λm(du).
The inner integral on the right is K(u), so by (4) the equality (2) holds for m and n+1. As m and F were arbitrary, n+1∈S.
Step 5 (Conclusion). By Steps 1 to 4, S=N, so (2) holds for all m,n∈N; as noted in Step 1, (1) and (2) with m=q, n=p and F=f give the Borel measurability of v↦f(ιq,p(u,v)) for each u, the Borel measurability of u↦∫Rpf(ιq,p(u,v))λp(dv), and the displayed identity of the lemma.