精華區beta Math 關於我們 聯絡資訊
※ 引述《kemowu (小展)》之銘言: : Suppose ∫ f(y) dy = 0 for every subinterval I of R. Show that f = 0 a.e. : I : on R. For any finitely measurable set A, there corresponds a descending sequence of open sets O_n, each containg A, such that the intersection of all O_n differ only by a measure zero set from A. Let χ_n be the characteristic function of O_n, χ be that of A. Since the integral of f vanishes on each open interval, so does the integral of fχ_n. Hnece by Lebesgue dominated convergence theorem, ∫f χ dx = ∫_A f dx =0. Namely, the integral of f on each finitely measurable set equals zero, which implies the same property of f on each measurable set. For all positive a define E(a) := {x|f(x)>a}. 0 =∫ f dx ≧ a m(E(a)). E(a) Hence m(E(a)) = 0 for all positive a. Consequently f(x)≦0 a.e.. The reversed inequality holds similarly. Another proof: If we take Mikusinski formulation to Daniell integral as our definition of integration, f is integrable ( on (a,b) ) iff there exists a sequence of step functions s_n such that (i) ∫|s_m-s_n| dx approaches zero as m and n tend to infinity, (ii) s_n tends to f a.e., whence ∫f dx is defined by lim ∫s_n dx. n→∞ Hence │∫s_n dx│≦ ∫│s_m-s_n│dx + │∫ s_m dx│ I I I t_n Put s_n = Σ c χ , where χ is the charateristic function of k=1 k,n k,n k,n an interval I_{k,n}. Subsequently, t_n t_n Σ│∫ s_n dx│ = ∫│s_n│ dx ≦∫│s_m-s_n│dx + │Σ ∫ s_m dx│. k=1 I_{k,n} k=1 I_{k,n} Given ε>0, there correspond a large N such that ∫|s_m-s_n| dx < ε whenever m > n > N. Fix n > N, and let m approaches infinity: ∫│s_n│ dx < ε, for all n > N. Hence for all positive a, m({x│|s_n(x)| > a, for some n > N }) ≦ ε/a, which implies m({x│|f(x)| ≧ a }) ≦ ε/a. Since ε was arbitrary, m({ x | |f(x)| ≧ a }) = 0. 基本上提到 interval 就可以想想上面第二種證明一開始列的那個性質, 如果從 measure formulation 來看這個性質可以當作 Egoroff theorem 的一種變形, 第一種方法基本上就是把這件事的證明跑過一遍, 大概是這樣 ※ 編輯: ppia 來自: 114.32.4.99 (08/18 20:11) ※ 編輯: ppia 來自: 114.32.4.99 (08/18 20:12)