Throughout, a measurable function X → R X\to\mathbb{R} X → R with values ≥ 0 \ge0 ≥ 0 lies in M + \mathcal{M}^{+} M + , and when it is integrable its integrals as an element of M + \mathcal{M}^{+} M + and of L 1 \mathcal{L}^{1} L 1 coincide; and for a measure ρ \rho ρ and f ∈ L 1 ( ρ ) f\in\mathcal{L}^{1}(\rho) f ∈ L 1 ( ρ ) , ∫ f d ρ = ∫ f + d ρ − ∫ f − d ρ \int f\,d\rho=\int f^{+}\,d\rho-\int f^{-}\,d\rho ∫ f d ρ = ∫ f + d ρ − ∫ f − d ρ (base). By the Hahn decomposition theorem (which uses the axiom of dependent choice ), every finite signed measure on ( X , F ) (X,\mathcal{F}) ( X , F ) has a Hahn set , so its Jordan decomposition exists. Let ν + , ν − \nu^{+},\nu^{-} ν + , ν − and ν ′ + , ν ′ − \nu'^{+},\nu'^{-} ν ′ + , ν ′ − be the Jordan decompositions of ν \nu ν and ν ′ \nu' ν ′ ; these are finite measures, and ∣ ν ∣ = ν + + ν − |\nu|=\nu^{+}+\nu^{-} ∣ ν ∣ = ν + + ν − setwise by Properties of the Total Variation: a Finite Measure, the Jordan Decomposition, Linear Combinations, the Total Variation Norm, and Integrability (ii) .
Step 1: linearity. Let f , g ∈ L 1 ( ν ) = L 1 ( ∣ ν ∣ ) f,g\in\mathcal{L}^{1}(\nu)=\mathcal{L}^{1}(|\nu|) f , g ∈ L 1 ( ν ) = L 1 ( ∣ ν ∣ ) and a , b ∈ R a,b\in\mathbb{R} a , b ∈ R . By linearity of the Lebesgue integral, a f + b g ∈ L 1 ( ∣ ν ∣ ) = L 1 ( ν ) af+bg\in\mathcal{L}^{1}(|\nu|)=\mathcal{L}^{1}(\nu) a f + b g ∈ L 1 ( ∣ ν ∣ ) = L 1 ( ν ) . By Properties of the Total Variation: a Finite Measure, the Jordan Decomposition, Linear Combinations, the Total Variation Norm, and Integrability (viii) , f f f and g g g lie in L 1 ( ν + ) \mathcal{L}^{1}(\nu^{+}) L 1 ( ν + ) and L 1 ( ν − ) \mathcal{L}^{1}(\nu^{-}) L 1 ( ν − ) , and linearity of the integrals against ν + \nu^{+} ν + and ν − \nu^{-} ν − gives
∫ ( a f + b g ) d ν = ∫ ( a f + b g ) d ν + − ∫ ( a f + b g ) d ν − = a ( ∫ f d ν + − ∫ f d ν − ) + b ( ∫ g d ν + − ∫ g d ν − ) = a ∫ f d ν + b ∫ g d ν . \int(af+bg)\,d\nu=\int(af+bg)\,d\nu^{+}-\int(af+bg)\,d\nu^{-}=a\Bigl(\int f\,d\nu^{+}-\int f\,d\nu^{-}\Bigr)+b\Bigl(\int g\,d\nu^{+}-\int g\,d\nu^{-}\Bigr)=a\int f\,d\nu+b\int g\,d\nu . ∫ ( a f + b g ) d ν = ∫ ( a f + b g ) d ν + − ∫ ( a f + b g ) d ν − = a ( ∫ f d ν + − ∫ f d ν − ) + b ( ∫ g d ν + − ∫ g d ν − ) = a ∫ f d ν + b ∫ g d ν .
This proves (i) .
Step 2: the bound. Let f ∈ L 1 ( ν ) f\in\mathcal{L}^{1}(\nu) f ∈ L 1 ( ν ) . Then ∣ f ∣ ∈ L 1 ( ∣ ν ∣ ) |f|\in\mathcal{L}^{1}(|\nu|) ∣ f ∣ ∈ L 1 ( ∣ ν ∣ ) , and f ∈ L 1 ( ν ± ) f\in\mathcal{L}^{1}(\nu^{\pm}) f ∈ L 1 ( ν ± ) by Properties of the Total Variation: a Finite Measure, the Jordan Decomposition, Linear Combinations, the Total Variation Norm, and Integrability (viii) . The inequality ∣ ∫ f d ρ ∣ ≤ ∫ ∣ f ∣ d ρ |\int f\,d\rho|\le\int|f|\,d\rho ∣ ∫ f d ρ ∣ ≤ ∫ ∣ f ∣ d ρ for ρ = ν + , ν − \rho=\nu^{+},\nu^{-} ρ = ν + , ν − and the identity of Properties of the Total Variation: a Finite Measure, the Jordan Decomposition, Linear Combinations, the Total Variation Norm, and Integrability (viii) give
∣ ∫ f d ν ∣ ≤ ∣ ∫ f d ν + ∣ + ∣ ∫ f d ν − ∣ ≤ ∫ ∣ f ∣ d ν + + ∫ ∣ f ∣ d ν − = ∫ ∣ f ∣ d ∣ ν ∣ . \Bigl|\int f\,d\nu\Bigr|\le\Bigl|\int f\,d\nu^{+}\Bigr|+\Bigl|\int f\,d\nu^{-}\Bigr|\le\int|f|\,d\nu^{+}+\int|f|\,d\nu^{-}=\int|f|\,d|\nu| . ∫ f d ν ≤ ∫ f d ν + + ∫ f d ν − ≤ ∫ ∣ f ∣ d ν + + ∫ ∣ f ∣ d ν − = ∫ ∣ f ∣ d ∣ ν ∣.
This proves (ii) .
Step 3: bounded functions. Let f f f and C C C be as in (iii) . Since ∣ ν ∣ |\nu| ∣ ν ∣ is a finite measure (Properties of the Total Variation: a Finite Measure, the Jordan Decomposition, Linear Combinations, the Total Variation Norm, and Integrability (i) ), f ∈ L 1 ( ∣ ν ∣ ) = L 1 ( ν ) f\in\mathcal{L}^{1}(|\nu|)=\mathcal{L}^{1}(\nu) f ∈ L 1 ( ∣ ν ∣ ) = L 1 ( ν ) by Bounded Measurable Functions on a Finite Measure Space Are Integrable (i) . The function with constant value C = C 1 X ( x ) C=C\,\mathbf{1}_{X}(x) C = C 1 X ( x ) lies in M + \mathcal{M}^{+} M + and has integral C ∣ ν ∣ ( X ) = C ∥ ν ∥ C\,|\nu|(X)=C\,\|\nu\| C ∣ ν ∣ ( X ) = C ∥ ν ∥ against ∣ ν ∣ |\nu| ∣ ν ∣ , and ∣ f ∣ ≤ C |f|\le C ∣ f ∣ ≤ C pointwise; by monotonicity of the integral on M + \mathcal{M}^{+} M + and Step 2,
∣ ∫ f d ν ∣ ≤ ∫ ∣ f ∣ d ∣ ν ∣ ≤ C ∥ ν ∥ . \Bigl|\int f\,d\nu\Bigr|\le\int|f|\,d|\nu|\le C\,\|\nu\| . ∫ f d ν ≤ ∫ ∣ f ∣ d ∣ ν ∣ ≤ C ∥ ν ∥.
This proves (iii) .
Step 4: differences of finite measures. Let μ 1 , μ 2 \mu_{1},\mu_{2} μ 1 , μ 2 , f f f and C C C be as in (iv) . For every finite measure ρ \rho ρ on ( X , F ) (X,\mathcal{F}) ( X , F ) , f ∈ L 1 ( ρ ) f\in\mathcal{L}^{1}(\rho) f ∈ L 1 ( ρ ) by Bounded Measurable Functions on a Finite Measure Space Are Integrable (i) ; and f ∈ L 1 ( ν ) f\in\mathcal{L}^{1}(\nu) f ∈ L 1 ( ν ) by Step 3. Since ν + − ν − = ν = μ 1 − μ 2 \nu^{+}-\nu^{-}=\nu=\mu_{1}-\mu_{2} ν + − ν − = ν = μ 1 − μ 2 setwise, ν + ( A ) + μ 2 ( A ) = ν − ( A ) + μ 1 ( A ) \nu^{+}(A)+\mu_{2}(A)=\nu^{-}(A)+\mu_{1}(A) ν + ( A ) + μ 2 ( A ) = ν − ( A ) + μ 1 ( A ) for every A ∈ F A\in\mathcal{F} A ∈ F : the finite measures ν + + μ 2 \nu^{+}+\mu_{2} ν + + μ 2 and ν − + μ 1 \nu^{-}+\mu_{1} ν − + μ 1 of Nonnegative Combinations of Finite Measures and Their Integrals (i) are equal. So Nonnegative Combinations of Finite Measures and Their Integrals (ii) gives
∫ f d ν + + ∫ f d μ 2 = ∫ f d ( ν + + μ 2 ) = ∫ f d ( ν − + μ 1 ) = ∫ f d ν − + ∫ f d μ 1 , \int f\,d\nu^{+}+\int f\,d\mu_{2}=\int f\,d(\nu^{+}+\mu_{2})=\int f\,d(\nu^{-}+\mu_{1})=\int f\,d\nu^{-}+\int f\,d\mu_{1}, ∫ f d ν + + ∫ f d μ 2 = ∫ f d ( ν + + μ 2 ) = ∫ f d ( ν − + μ 1 ) = ∫ f d ν − + ∫ f d μ 1 ,
and hence
∫ f d ν = ∫ f d ν + − ∫ f d ν − = ∫ f d μ 1 − ∫ f d μ 2 . \int f\,d\nu=\int f\,d\nu^{+}-\int f\,d\nu^{-}=\int f\,d\mu_{1}-\int f\,d\mu_{2}. ∫ f d ν = ∫ f d ν + − ∫ f d ν − = ∫ f d μ 1 − ∫ f d μ 2 .
This proves (iv) .
Step 5: linearity in the measure. Let f f f and C C C be as in (v) . By (iii) , applied to ν \nu ν , ν ′ \nu' ν ′ and to the finite signed measures ν + ν ′ \nu+\nu' ν + ν ′ and c ν c\nu c ν , f f f is integrable with respect to each of them. For finite measures ρ 1 , ρ 2 \rho_{1},\rho_{2} ρ 1 , ρ 2 and real a 1 , a 2 ≥ 0 a_{1},a_{2}\ge0 a 1 , a 2 ≥ 0 , f f f lies in L 1 ( ρ 1 ) ∩ L 1 ( ρ 2 ) \mathcal{L}^{1}(\rho_{1})\cap\mathcal{L}^{1}(\rho_{2}) L 1 ( ρ 1 ) ∩ L 1 ( ρ 2 ) by Bounded Measurable Functions on a Finite Measure Space Are Integrable (i) , so Nonnegative Combinations of Finite Measures and Their Integrals (ii) gives
∫ f d ( a 1 ρ 1 + a 2 ρ 2 ) = a 1 ∫ f d ρ 1 + a 2 ∫ f d ρ 2 . ( ∗ ) \int f\,d(a_{1}\rho_{1}+a_{2}\rho_{2})=a_{1}\int f\,d\rho_{1}+a_{2}\int f\,d\rho_{2}.\tag{$*$} ∫ f d ( a 1 ρ 1 + a 2 ρ 2 ) = a 1 ∫ f d ρ 1 + a 2 ∫ f d ρ 2 . ( ∗ )
Now ν + ν ′ = ( ν + + ν ′ + ) − ( ν − + ν ′ − ) \nu+\nu'=(\nu^{+}+\nu'^{+})-(\nu^{-}+\nu'^{-}) ν + ν ′ = ( ν + + ν ′ + ) − ( ν − + ν ′ − ) setwise, a difference of finite measures, so Step 4 applied to ν + ν ′ \nu+\nu' ν + ν ′ , then ( ∗ ) (*) ( ∗ ) , give
∫ f d ( ν + ν ′ ) = ∫ f d ν + + ∫ f d ν ′ + − ∫ f d ν − − ∫ f d ν ′ − = ∫ f d ν + ∫ f d ν ′ . \int f\,d(\nu+\nu')=\int f\,d\nu^{+}+\int f\,d\nu'^{+}-\int f\,d\nu^{-}-\int f\,d\nu'^{-}=\int f\,d\nu+\int f\,d\nu' . ∫ f d ( ν + ν ′ ) = ∫ f d ν + + ∫ f d ν ′ + − ∫ f d ν − − ∫ f d ν ′ − = ∫ f d ν + ∫ f d ν ′ .
If c ≥ 0 c\ge0 c ≥ 0 , then c ν = c ν + − c ν − c\nu=c\nu^{+}-c\nu^{-} c ν = c ν + − c ν − setwise, a difference of finite measures (Nonnegative Combinations of Finite Measures and Their Integrals (i) with μ 2 = μ 1 \mu_{2}=\mu_{1} μ 2 = μ 1 and a 2 = 0 a_{2}=0 a 2 = 0 ), and Step 4 and ( ∗ ) (*) ( ∗ ) give ∫ f d ( c ν ) = c ∫ f d ν + − c ∫ f d ν − = c ∫ f d ν \int f\,d(c\nu)=c\int f\,d\nu^{+}-c\int f\,d\nu^{-}=c\int f\,d\nu ∫ f d ( c ν ) = c ∫ f d ν + − c ∫ f d ν − = c ∫ f d ν . If c < 0 c<0 c < 0 , then c ν = ∣ c ∣ ν − − ∣ c ∣ ν + c\nu=|c|\nu^{-}-|c|\nu^{+} c ν = ∣ c ∣ ν − − ∣ c ∣ ν + setwise, and likewise ∫ f d ( c ν ) = ∣ c ∣ ∫ f d ν − − ∣ c ∣ ∫ f d ν + = c ∫ f d ν \int f\,d(c\nu)=|c|\int f\,d\nu^{-}-|c|\int f\,d\nu^{+}=c\int f\,d\nu ∫ f d ( c ν ) = ∣ c ∣ ∫ f d ν − − ∣ c ∣ ∫ f d ν + = c ∫ f d ν . This proves (v) .
Step 6: indicators. Let A ∈ F A\in\mathcal{F} A ∈ F . The indicator 1 A \mathbf{1}_{A} 1 A is a nonnegative simple function, in particular measurable as a function into R \mathbb{R} R , with ∣ 1 A ∣ ≤ 1 |\mathbf{1}_{A}|\le1 ∣ 1 A ∣ ≤ 1 ; so 1 A ∈ L 1 ( ν ) \mathbf{1}_{A}\in\mathcal{L}^{1}(\nu) 1 A ∈ L 1 ( ν ) by Step 3. For every finite measure ρ \rho ρ , ∫ 1 A d ρ = ρ ( A ) \int\mathbf{1}_{A}\,d\rho=\rho(A) ∫ 1 A d ρ = ρ ( A ) , so
∫ 1 A d ν = ν + ( A ) − ν − ( A ) = ν ( A ) . \int\mathbf{1}_{A}\,d\nu=\nu^{+}(A)-\nu^{-}(A)=\nu(A). ∫ 1 A d ν = ν + ( A ) − ν − ( A ) = ν ( A ) .
This proves (vi) .
Step 7: finite measures. Suppose ν ( A ) ≥ 0 \nu(A)\ge0 ν ( A ) ≥ 0 for every A ∈ F A\in\mathcal{F} A ∈ F . By Properties of the Total Variation: a Finite Measure, the Jordan Decomposition, Linear Combinations, the Total Variation Norm, and Integrability (iv) , ν \nu ν is a finite measure with ∣ ν ∣ = ν |\nu|=\nu ∣ ν ∣ = ν , so the set L 1 ( ∣ ν ∣ ) \mathcal{L}^{1}(|\nu|) L 1 ( ∣ ν ∣ ) of Integrable Functions and the Integral with Respect to a Finite Signed Measure (i) is the set L 1 ( ν ) \mathcal{L}^{1}(\nu) L 1 ( ν ) of the base; and for f f f in it, by the same clause, ∫ f d ν + − ∫ f d ν − \int f\,d\nu^{+}-\int f\,d\nu^{-} ∫ f d ν + − ∫ f d ν − is the Lebesgue integral of f f f against ν \nu ν . This proves (vii) .