MA2224 (Lebesgue integral) Tutorial sheet 3 [February 5, 2016] Name: Solutions 1. Let I be the (usual) interval algebra (of subsets of R) and define µ : I → [0, ∞] by ( 0 if m(E) < ∞ µ(E) = 1 if m(E) = ∞ [Aside: This will not be a useful way to do things.] Show that µ is monotone and finitely subadditive, but not finitely additive. Solution: µ monotone: if E1 , E2 ∈ I with E1 ⊆ E2 , we consider the two cases m(E1 ) < ∞ and m(E1 ) = ∞ separately. If m(E1 ) < ∞, then µ(E1 ) = 0 ≤ µ(E2 ). If m(E1 ) = ∞, then since m is monotone, ∞ = m(E1 ) ≤ m(E2 ) and so m(E2 ) = ∞ also. Thus 1 = µ(E1 ) ≤ 1 = µ(E2 ) holds in this case also. µ finitely subadditive: it is enough to show that if E1 , E2 ∈ I , then µ(E1 ∪ E2 ) ≤ µ(E1 ) + µ(E2 ). If m(E1 ∪ E2 ) < ∞, then µ(E1 ∪ E2 ) = 0 ≤ µ(E1 ) + µ(E2 ). If m(E1 ∪ E2 ) = ∞, then since m is finitely subadditive m(E1 ∪ E2 ) ≤ m(E1 ) + m(E2 ) and so at least one of m(E1 ), m(E2 ) is ∞. So one of µ(E1 ), µ(E2 is 1 and so 1 = µ(E1 ∪ E2 ) ≤ µ(E1 ) + µ(E2 ). µ not finitely additive: if E1 = (−∞, 0] and E2 = (0, ∞), then E1 , E2 ∈ I are disjoint and 1 = µ(E1 ∪ E2 ) = µ(R) 6= 2 = µ(E1 ) + µ(E2 ). 2. If S = {x0 } has just one point x0 ∈ R, show that m∗ (S) = 0. Solution: For ε > S 0, consider E1 = (x0 − ε, x0 ] ∈ I , En = ∅ ∈ I (for n = 2, 3, . . .). Then S = {x0 } ⊆ ∞ n=1 En and ∞ X m(En ) = m(E1 ) + 0 = ε. n=1 Thus m∗ (S) ≤ ε. As ε > 0 is arbitrary, we must have m∗ (S) = 0. 3. If F ⊆ R is a countable set, show that m∗ (F ) = 0. S ∗ Solution: If F = {x1 , x2 , . . .} is countably infinite, F = ∞ n=1 {xn }. As m is countably subadditive ∞ X ∗ m (F ) ≤ m∗ ({xn }) = 0 n=1 ∗ (using Q2). Hence m (F ) = 0 in this case. If F = {x1 , x2 , . . . , xn } is finite (countable sets can be finite or countably S infinite), we can use finite subadditivity of m∗ in the same way to get m∗ (F ) = 0 (F = nj=1 {xj }). 4. Show carefully that m∗ ([a, b]) = b − a if a ≤ b. [Hint: if a < b compare (a, b] with [a, b]. If a = b, what is [a, b]?] Solution: As (a, b] ⊂ [a, b] and m∗ is monotone, m∗ ((a, b]) ≤ m∗ ([a, b]). But m∗ ((a, b]) = m((a, b]) = b − a since (a, b] ∈ I (even if a = b when (a, b] = (a, a] = ∅). So b − a ≤ m∗ ([a, b]). By finite subadditivity of m∗ applied to [a, b] = {a} ∪ (a, b] we also have m∗ ([a, b]) ≤ m∗ ({a}) + m∗ ((a, b]) = 0 + m((a, b]) = b − a (using Q2). Richard M. Timoney 2