Problem Set Three: Convergence and Completeness Definitions: Let (S,d) be a metric space. A sequence (f n ) in S converges to a point f in S iff ε 0 m N so that n m d(f n , f ) ε . Convergence to f is denoted by writing f n f or lim f n f . Convergence to f is equivalent to each of these statements. (i) For any open ball B centered at f, m N so that n m f n B . (ii) The real sequence d(f n , f ) converges to zero, i.e., d(f n , f ) 0 . Example: In Frechet space R N if f n ( 1, 2, 3, ... , n,1,1, ... etc ... ) for each n and f ( 1, 2, 3, ... , n, n 1, n 2, ... etc ... ) then f n f . Example: In C[0,1] with the 2-norm, the sequence f n (x) x n converges to the constantly zero function. Note f n (1) 1 for all n. Definitions: Let (S,d) be a metric space and (f n ) be a sequence in S. (a) (b) (c) (d) (f n ) is Cauchy iff ε 0 m N so that n & k m d(f n , f k ) ε . S is complete iff each Cauchy sequence in S converges to some point in S. A set E in S is bounded iff E is contained in some open ball. (f n ) is bounded iff its set of values { f1 , f 2 , ... , f n , ... } is bounded in S. Example: d(x, y) 1/x 1/y define a new metric on S = (0,1). (S,d) has the same convergent sequences as (0,1) with the usual metric. S isn't d-bounded. w n 1/n isn't d-Cauchy. Theorem: In a metric space a convergent sequence is Cauchy. A Cauchy sequence is bounded. The sign or signum function is defined by sgn(x) 1 if x 0 , and sgn(x) 1 otherwise. Example: Define f n on [ 1, 1] by f n (x) nx for x 1/n and f n (x) sgn(x) otherwise. f n sgn 0 . Under the 2-norm C[ 1, 1] is not complete. 2 PROBLEMS Problem 3-1: Use the Cauchy-Schwartz Inequality to show (again) that for any n-tuple x in R n , x 2 max k 1 x ( k ) y( k ) : n y 2 1. Problem 3-2 (Arc Length Inequality): Here is a way to use the Cauchy-Schwartz Inequality to verify an inequality you know from calculus. If u(t) and v(t) are continuously differentiable functions on the interval [a,b] then ( u ( b ), v ( b )) ( u ( a ), v ( a )) 2 b a 2 2 u ( t ) v ( t ) dt . In other words the straight line distance from (u(a),v(a)) to (u(b),v(b)) can be estimated from above by the integral. Read the following proof and supply a reason for each step or assertion. Proof: Let x and y be any real numbers with x 2 y 2 1. x [ u ( b ) u ( a )] y [ v ( b ) v ( a )] x b a b a a a u ( t ) dt y b a v ( t ) dt x u ( t ) y v ( t ) dt b b x 2 y u ( t ) 2 2 u ( t ) v ( t ) 2 2 v ( t ) 2 dt dt . Since that’s true for all x and y with x 2 y 2 1, the desired inequality holds. Problem 3-3 (Orlitz Inequality, generalizes 3-2) If f1 , f 2 , ... , f n are continuous functions on [a,b] and 2 1/2 f(x) { nk 1 f k (x) } is their square function, then 2 1/2 { nk 1 [ ab f k (x) d x ] } ab f(x) d x (You only need continuity to make sure the integrals are defined.) Problem 3-4: This problem is about R N , the space of all real sequences with the metric d(f, g) k 1 2 k min{ 1, f(k) g(k) } . if 1 k n n f n (k) 0 if n k For each n let f n be the sequence . (a) Show that d(f n , f m ) 1 2 m for 1 n m . (b) Explain why the sequence is bounded in R N . (c) Explain why the sequence ( f n ) does not converge in R N . Problem 3-5: (a) For any n-tuple x in R n , x x 2 n x . (b) For any n-tuple x in R n , use the Cauchy-Schwartz Inequality to show x x n x . 2 1 2 (c) Use parts (a) and (b) to show that if a sequence is convergent (respectively, Cauchy, bounded) in one p-norm then it is convergent (resp., Cauchy, bounded) in all p-norms, p 1, 2, . Problem 3-6: Let (S,d) be a metric space. (a) For any points x, y and z in S, d ( x , z ) d ( y , z ) d ( x , y ) (b) If x n x then d ( x n , z ) d ( x , z ) for any z in S. (c) If x n x and y n y then d ( x n , y n ) d ( x , y ). FACT FROM ADVANCED CALCULUS: With the usual metric d ( x , y ) x y the reals are a complete metric space Problem 3-7: (a) For any sequence of real numbers min{ 1, x n x } 0 iff x n x 0 . (b) With the metric d ( x , y ) min{ 1, x y } , the reals are a complete metric space. Problem 3-8: In the reals with the railroad metric, a Cauchy sequence is either eventually constant or converges to zero in the usual sense. R with the railroad metric is a complete metric space.