Lecture #22 MATH 321: Real Variables II University of British Columbia Lecture #22: Instructor: Scribe: March 3, 2008 Dr. Joel Feldman Peter Wong Theorem. (Weierstraß) If f : [a, b] → C is continuous, then there is a sequence of polynomials { Pn }n∈N such that sup |f (x) − Pn (x)| → 0 as n → ∞. x∈[a,b] Proof. Reduction #1: It suffices to consider [a, b] = [0, 1]. Reduction #2: It suffices to consider the case in which f (0) = f (1) = 0. R1 Define Qn (x) = cn (1 − x2 )n , where cn = { −1 (1 − x2 )n }−1 . Qn obeys (1) Qn (x) ≥ 0 on [−1, 1], R1 (2) −1 Qn (x) dx = 1, (3) 0 < δ ≤ |x| ≤ 1 =⇒ 0 ≤ Qn (x) ≤ cn (1 − δ 2 ) ≤ √ n(1 − δ 2 )n . Define for 0 ≤ x ≤ 1, Pn (x) = = Z 1 f (x + t)Qn (t) dt, −1 Z 1+x∈[1,2] −1+x∈[−1,0] setting s = x + t f (s)Qn (s − x) ds = Z 1 f (s) 0 P 2n Qn (s − x) | {z } m=0 (polynomial ds in S)xm is obviously a polynomial in x, which is real if f : [0, 1] → R. Let ε > 0. Since f is continuous on [0, 1], (i) f is bounded: ∃M such that |f (x)| ≤ M on [0, 1], (ii) f is uniformly continuous: ∃δ > 0 such that |f (x) − f (y)| < ε 2 if |x − y| < δ. Then for all 0 ≤ x ≤ 1, |f (x) − Pn (x)| = f (x) ≤ ≤ < = Z 1 Z−1 ε 2 ε 2 | 1 −1 Qn (t) dt − Pn (x) = Z 1 −1 [f (x) − f (t − x)]Qn (t) dt |f (x) − f (t + x)|Qn (t) dt |t|<δ Z Z |f (x) − f (t + x)| Qn (t) dt + {z } | |t|<δ Z δ≤|t|≤1 ε/2 √ Qn (t) dt +2 · 2M · n(1 − δ 2 )n {z ≤1 } √ + 4M n(1 − δ 2 )n < ε √ as long as n is big enough so that 4M n(1 − δ 2 )n < 2ε . |f (x) − f (t + x)| Qn (t) | {z } | {z } 2M √ n(1−δ 2 )n dt 2 MATH 321: Lecture #22 Idea: Generalize f : [a, b] → C continuous to f : K → C continuous, where K is any compact metric space. Problem: If x ∈ K, then x2 , x3 , . . . are not defined (i.e., there is no concept of polynomials on K.) Notation: Let • K be a compact metric space, • F = { K : R →| f bounded } with the metric d(f, g) = supx∈K |f (x) − g(x)|, • C = { K : R →| f continuous } with the metric d(f, g) = supx∈K |f (x) − g(x)|, • A ⊂ C as approximators. Definition. A is an algebra if (1) f, g ∈ A =⇒ f + g ∈ A, (2) f, g ∈ A =⇒ f g ∈ A, (3) f ∈ A, c ∈ R =⇒ cf ∈ A. Conjecture. Let K be a compact metric space and C = { f : K → R | f continuous } and let A ⊂ C be an algebra. Then A = C. ( A is the set of all uniform limits of sequences in A.) To search for conter-examples, we use the following approach: If (1) every p ∈ A has property X, (2) for every pn has property X, limpn has the property X, and (3) there exists an f ∈ C which does not have the property X, then the conjecture fails. Exercise. Find as many such properties X as possible. Lecture #23 MATH 321: Real Variables II University of British Columbia Lecture #23: Instructor: Scribe: March 5, 2008 Dr. Joel Feldman Peter Wong Notation: Let • K = a compact metric space, • F = { K : R →| f bounded } with the metric d(f, g) = supx∈K |f (x) − g(x)|, • C = { K : R →| f continuous } with the same metric d, • A ⊂ C the set of approximators. Definition. A is an algebra if (1) f, g ∈ A =⇒ f + g ∈ A, (2) f, g ∈ A =⇒ f g ∈ A, (3) c ∈ R, f ∈ A =⇒ cf ∈ A. Definition. A = the closure of A = the set of all limits of sequences in A Theorem. If A ⊂ F is an algebra, then A is a closed algebra. Proof. A is closed since any closure is closed. A is an algebra since f, g ∈ A =⇒ ∃ sequences { fn } , { gn } ⊂ A such that f = lim fn , g = lim gn (uniformly) f + g = lim fn + lim gn = lim (fn + gn ) ∈ A n→∞ n→∞ n→∞ f g = lim fn · lim gn = lim (fn gn ) ∈ A =⇒ n→∞ n→∞ n→∞ cf = c lim fn = lim (cfn ) ∈ A n→∞ n→∞ Conjecture. If A ⊂ C is an algebra. Then A = C. Conclusion: False: (1) If ∃ x ∈ K and A ⊂ { f : K → R | f (x) = 0 }, then A ⊂ { f : K → R | f (x) = 0 } (2) If ∃ x1 , x2 ∈ K, x1 6= x2 and A ⊂ { f : K → R | f (x1 ) = f (x2 ) }, then A ⊂ { f : K → R | f (x1 ) = f (x2 ) } Theorem. (Stone, 1937) If K is a compact metric space and A ⊂ C = { f : K → R | f continuous } obeys (1) A is an algebra, (2) A vanishes nowhere (i.e., ∀x ∈ K, ∃f ∈ A with f (x) 6= 0.) (3) A separates points (i.e., x1 , x2 ∈ K, x1 6= x2 implies ∃f ∈ A with f (x1 ) 6= f (x2 ).) Then A = C. 2 MATH 321: Lecture #23 Proof. Step 1: Show that f ∈ A =⇒ |f | ∈ A. Proof. Let ε > 0. It suffieces to find g ∈ A such that ∀x ∈ K, |g(x) − |f (x)|| < ε. Let a = supx∈K |f (x)|. By the Weierstraß Theorem: n X ∃c0 , . . . , cn ∈ R such that j=0 cj y j − |y| < 2ε , ∀|y| ≤ a =⇒ |c0 | < ε/2 (Set y = 0) =⇒ n X cj y j − |y| < n X cj f (x)j − |y| < ε, ∀x ∈ K j=1 =⇒ j=1 ε 2 + ε 2 = ε, ∀|y| ≤ a Step 2: If f1 , f2 ∈ A, then the pointwise max(f1 , f2 ), min(f1 , f2 ) ∈ A, where ( ( f1 (x), if f1 (x) ≥ f2 (x) f1 (x), max(f1 , f2 )(x) = and min(f1 , f2 )(x) = f2 (x), if f1 (x) ≤ f2 (x) f2 (x), if f1 (x) ≤ f2 (x) if f1 (x) ≥ f2 (x) Proof. max(f1 , f2 ) = |f1 − f2 | f1 + f2 + 2 2 and min(f1 , f2 ) = f1 + f2 |f1 − f2 | − 2 2 Step 3: Show that if f1 , . . . , fn ∈ A, then max{f1 , . . . , fn }, min f {f1 , . . . , fn } ∈ A. Proof. Apply the proof of Step 2 inductively. Step 4: Let x1 , x2 ∈ K, x1 6= x2 . Let c1 , c2 ∈ R. Then there is an f ∈ A with f (x1 ) = c1 and f (x2 ) = c2 . Proof. Since A separates points, ∃g ∈ A with g(x1 ) 6= g(x2 ) and A vanishes nowehere ∃h ∈ A with h(x1 ) 6= 0, and ∃k ∈ A with k(x2 ) 6= 0, g(x) − g(x2 ) h(x) g(x) − g(x1 ) k(x) f (x) = c1 +c2 g(x1 ) − g(x2 ) h(x1 ) g(x2 ) − g(x1 ) k(x2 ) | | {z } {z } = 1 0 at x = x1 at x = x2 = 0 1 at x = x1 at x = x2 Lecture #24 MATH 321: Real Variables II University of British Columbia Lecture #24: Instructor: Scribe: March 7, 2008 Dr. Joel Feldman Peter Wong Theorem. (Stone, 1937) If K is a compact metric space and A ⊂ C = { f : K → R | f continuous } with d(f, g) = supx∈K |f (x) − g(x)|, A ⊂ C is an algebra which vanishes nowhere and separates points, then A = C. Proof. Step 1: f ∈ A =⇒ |f | ∈ A Step 2: f1 , f2 ∈ A =⇒ max(f1 , f2 ), min(f1 , f2 ) ∈ A, pointwise-speaking Step 3: f1 , . . . , fn ∈ A =⇒ max{f1 , . . . , fn }, min f {f1 , . . . , fn } ∈ A Step 4: x1 , x2 ∈ K, x1 6= x2 , c1 , c2 ∈ R =⇒ ∃ f ∈ A with f (x1 ) = c1 and f (x2 ) = c2 Step 5: Let f ∈ C. Let x ∈ K. Let ε > 0. Then ∃ gx ∈ A (gx does not mean ∂g ∂x ) that obeys (1) gx (x) = f (x) (2) gx (t) > f (t) − ε, for all t ∈ K. Proof. By Step 4, for each y ∈ K, there is an hy ∈ A obeying hy (x) = f (x), hy (y) = f (y). Since hy − f is continuous and hy (y) − f (y) = 0, there is an open set Jy with y ∈ Jk with hy − f (t) > −ε for all t ∈ Jy (i.e., hy (t) > f (t) − ε.) { Jy }y∈K is an open cover of the compact set K =⇒ there is a finite subcover Jy1 , . . . , Jym . Then gx (t) = max { hy1 (t), . . . , hym (t) } works. We check that (a) hy1 , . . . , hym ∈ A ⊂ A =⇒ gx ∈ A. (By Step 3.) (b) gx (x) = max { f (x), . . . f (x) } = f (x). (c) gx ( |{z} t ) = max{hy1 (t), . . . , hy1001 (t), . . . , hym (t)} | {z } ∈Jy1001 >f (t)−ε Step 6: (Final Step) Let f ∈ C. Let ε > 0. Then ∃ g ∈ A such that d(f, g) < ε. Proof. By Step 5, for each x ∈ K, ∃ gx ∈ A that obeys (1) and (2). So gx − f is continuous and gx (x) − f (x) = 0 =⇒ ∃ an open set Kx obeying x ∈ Kx and gx (t) − f (t) < ε for all t ∈ Kx . Thus, { Kx }x∈K is an open cover of K, which is compact =⇒ there is a finite subcover Kx1 , . . . , Kxn . Taking g(t) = min{ gx1 , . . . , gxn } ∈ A |{z} |{z} ∈A ∈A works since (a) g(t) = min{ gx1 (t) , . . . , gxn (t) } > f (t) − ε | {z } | {z } >f (t)−ε >f (t)−ε (b) g( |{z} t ) = min{ gx1 (t) , . . . , gx1001 (t), . . . , gxn (t) } > f (t) + ε | {z } | {z } | {z } ∈Kx1001 >f (t)−ε f (t)+ε >f (t)−ε Theorem. (Generalized Stone-Weierstraß for Complex numbers) If K is a compact metric space, C = { f : K → C | f continuous } with the metric d(f, g) = supx∈K |f (t) − g(t)|, A ⊂ C is an algebra in C which vanishes nowhere, separates points, and is self-adjoint (i.e., f ∈ A =⇒ f ∗ ∈ A, where f ∗ is the complex conjugate of f . See Problem Set 7 Question 2.) Then A = C. 2 MATH 321: Lecture #24 Proof. Set AR = { f ∈ A | f is real-valued }. Since AR is a real algebra which vanishes nowhere and spearates point. By Step 4, there is for each x1 6= x2 , and f ∈ A which f (x1 ) = 1 and f (x2 ) = 0. By Stone-Weierstraß for the reals, A = { f : K → R } f continuous. For any f ∈ C, f (x) = f (x) − f ∗ (x) f (x) + f ∗ (x) +i ∈A 2 2i | {z } | {z } =<(f ) ∈ AR ⊂ A ==(f ) ∈ AR ⊂ A