1 Homework 1 Weiyu Li I. S TEIN 1.2: C ANTOR SET DESCRIBED IN TERNARY EXPANSIONS Some notations are shown as follows: On construction of Cantor set C : Let C0 = [0, 1]. This interval is divided into three parts: the middle third open interval of C0 is E1,1 = (1/3, 2/3) which is excluded, while I1,1 = [0, 1/3], I1,2 = [2/3, 1] is included to k−1 obtain C1 = I1,1 ∪ I1,2 . Generally, when we get Ck = ∪2i=1 Ik,i , each Ik,i is divided into three parts with the middle open one denoted as Ek+1,i , the other two are Ik+1,2i−1 , Ik+1,2i . Then C := ∩∞ k=0 Ck . A. Cantor sets are points represented in 0 and 2. Proof: Notice that x ∈ / E1,1 (i.e., x ∈ C1 ) if and only if x has a decomposition with a1 6= 1. By deduction, we claim that Ck consists of x that has a decomposition with aj 6= 1, j ≤ k . The definition of C completes the proof. B. Well-definedness and continuity of Cantor-Lebesgue function. Proof: First, to show the well-definedness, it is sufficient to show that the ternary expansion of 0 and 2 is unique. Suppose that x= ∞ X ak 3−k = k=1 ∞ X bk 3−k , ak , bk ∈ {0, 2}. k=1 Let k0 := inf{k : ak 6= bk } < ∞, and WLOG (without loss of generality), ak0 = 0, bk0 = 2. Then 0= ∞ ∞ X X 2 · 3−k = 3−k0 , (bk − ak )3−k ≥ 2 · 3−k0 − k=1 k=k0 +1 which is a contradiction! This argument also shows that ∀|x − y| < 3−k0 , x, y ∈ C , the first k0 expansions are the same. Second, the continuity. ∀ > 0, ∃k0 , s.t. 2−k0 < . Therefore, for any x ∈ C , choose δ = 3−k0 , then P −k = 2−k0 < . ∀y ∈ C, |x−y| < δ, the first k0 expansions of x and y are the same, thus |F (x)−F (y)| ≤ ∞ k0 +1 2 From the definition, we know that F is continuous. Moreover, F (0) = 0, F (1) = 1 follows directly from 0 = P∞ k=1 0 · 3−k and 1 = P∞ k=1 2 · 3−k . C. Surjectiveness of Cantor-Lebesgue function. Proof: Every y ∈ [0, 1] has a binary expression y = P∞ x = k=1 2bk · 3−k ∈ C . March 4, 2019 P∞ k=1 bk · 2−k , bk ∈ {0, 1}, from which we can recover DRAFT 2 D. Continuity of extended Cantor-Lebesgue function. Proof: Note that F |C is non-decreasing and the extended definition can be interpreted as F (x) := F (sup{y : y ∈ C, y ≤ x} = sup{F (y) : y ∈ C, y ≤ x} for any x ∈ [0, 1], or F (x) := inf{F (y) : y ∈ C, y ≥ x} due to the monotonicity of F |C and closedness of C . ∀ > 0, x ∈ [0, 1], choose δ = 3−k0 same as in (I-B), (s.t. 2−k0 < ,) then ∀|x − y| < δ, if x, y ∈ C, P −k = 2−k0 < ; else, WLOG y > x, let x0 = inf{z : z ∈ C, z ≥ x} and y 0 = inf{z : |F (x) − F (y)| ≤ ∞ k0 +1 2 z ∈ C, z ≤ y}, then |F (x) − F (y)| = |F (x0 ) − F (y 0 )| < . From the definition, F is continuous. II. S TEIN 1.3: C ANTOR SETS Cξ A. Interpretation of Cξc . Proof: Same notations as in the construction of C . In the space of C0 , |C1 | = 1 − ξ , and by deduction c ∞ c |Ck | = (1 − ξ)k . Also note that Ckc is unions of finite open intervals. Therefore, Cξc = (∩∞ k=0 Ck ) = ∪k=0 Ck is the union of open intervals with total length no less than 1 − |Ck | → 1. Constrained in [0, 1], the total length equals to 1. B. m∗ (Cξ ) = 0. Proof: From the definition, 0 ≤ m∗ (Cξ ) = inf{ ∞ X |Qj | Cξ ⊂ ∪∞ j=1 Qj , Qj are closed intervals} ≤ inf |Ck | = 0. j=1 Thus, m∗ (Cξ ) = 0. III. S TEIN 1.14: O UTER J ORDAN CONTENT - FINITE COVERING INTERVALS A. J∗ (E) = J∗ (Ē). P N N N ¯ Proof: If Ē ⊂ ∪N j=1 Ij , then E ⊂ Ē ⊂ ∪j=1 Ij , thus J∗ (E) ≤ J∗ (Ē). If E ⊂ ∪j=1 Ij , then Ē ⊂ ∪j=1 Ij with P ¯ |Ij | = |Ij |, thus J∗ (E) ≥ J∗ (Ē). B. J∗ (E) = 1, m∗ (E) = 0. E = Q ∩ [0, 1] is an example. March 4, 2019 DRAFT 3 IV. S TEIN 1.15: O UTER MEASURE mR ∗ BY COUNTABLE COVERING RECTANGLES Proof: Given E , since cubes are rectangles, m∗ (E) ≥ mR ∗ (E). R To prove m∗ (E) ≤ mR ∗ (E), WLOG, m := m∗ (E) < ∞. For any > 0, there exists a countable covering P E ⊂ ∪Rj , such that m ≤ ∞ j=1 |Rj | < m + . For each Rj = [a1j , b1j ] × . . . × [adj , bdj ], consider the grid method in proof of Theorem 1.4, but with side length small enough and accept all the tentatively accepted cubes. The almost disjoint accepted cubes {Qji , i = 1, . . . , nj } make up of a rectangle Qj ⊃ Rj and from Pnj Lemma 1.1 satisfy |Qj | = i=1 |Qji | ≤ |Rj | + /2j . In summary, we have ∪j≥1,1≤i≤nj Qji ⊃ ∪Rj ⊃ E with P P j ij |Qji | ≤ j (|Rj | + /2 ) ≤ m + 2. Since is arbitrary chosen, m∗ (E) ≤ m. Remark 1. Lemma 1.1 is the finite sum. If you want to use the infinity summation, please use Observation 5 in P15 instead. March 4, 2019 DRAFT