Page 363 - DMTH401_REAL ANALYSIS
P. 363

Unit 30: Riemann's and Lebesgue




                                                   d
                                 d
                     d
          (ii)  If z   , there is r    such that z – r  [0, 1] . Then there is y  Y such that y ~ z  – r and  Notes
               so there is r    such that y + r = z – r or z = y + r + r. This shows that   =   r d (Y + r).
                            d
                                                                          d
               By [102](iii) and [102](viii), we conclude that  m * L,d  [Y] > 0. Let  =  m  * L,d  [Y] and n   be such
                                                                               d
                       d
               that n > 2 . Choose distinct elements x ,…,x   X. Then  å n  m *  [Y + x ] = n > 2  again by
                                                             =
                                              1   n          i 1  L,d  i
                                                                                     d
                                                                         [
               translation invariance. On the other hand, Y + X  [0,2]  and therefore  m  * L,d  n i 1 (Y+x )] < 2 .
                                                          d
                                                                            =
                                                                                 i
             Note  The construction above is due to Vitali, and hence the set Y is called a Vitali set.
          30.3 About Functions Behaving Nicely Outside a Small Set
          There are a  few classical results in Analysis with conclusion of the following form: “…  the
          function has nice behavior outside a small set”. We will consider some such results here.
          We know that a function that is the pointwise limit of a sequence of continuous functions may
          not be continuous. For instance, f : [0, 1]   given by f(1) = 1 and f(x) = 0 for x < 1 is the pointwise
          limit of (f ), where f : [0, 1]   is f (x) = x .
                                            n
                  n       n            n
          Definition: Let X, Y be metric spaces and let f: X  Y be a function. Then the oscillation (f, x) of
          f at a point x  X is defined as (f, x) = lim   diam[f(B(x, ))]. Clearly, f is continuous at x iff 
                                            0+
          (f, x) = 0.



              Task Let X,Y be metric spaces and let f: X  Y be a function. Then the set {x  X: f is
             continuous at x} is a G  subset of X. [Hint: The given set is equal to   ¥  U , where U  = {x  X:
                                                                  =
                                                                n 1  n      n
             (f, x) < 1/n}, and U  is open.]
                             n
          Let (X,  ) be a complete metric space, (Y,   ) be an arbitrary  metric space, and let (f ) be a
                 1                             2                                 n
          sequence of continuous functions from X to Y, converging pointwise to a function f: X  Y. Then
          the set {x  X : f is continuous at x} is a dense G  subset of X.
                                                
          Proof: Let  > 0 and D  = {x  X : (f, x) > }. We know that D is a closed set. We claim that D  is
                                                                                  
          nowhere dense in X. Let U  X be a nonempty open set. We have to find a nonempty open set V
           U such that D   V = Ø.
                       
          For n  , let K  = {x  X:  (f (x), fj(x))  /8 for every j  n}. Then K  is a closed set and X =    ¥  K .
                                                                                   =
                      n        2  n                           n                   n 1  n
          The continuity of the distance function   implies that  (f (x), f(x))  /8 for every x  K . Let
                                           2            2  n                      n
          U   X be a nonempty open set with  U  U. Since ( U ,  ) is a complete metric space, there is
            1                             1           1  1
          n  N such that U := int[K    U ]  Ø. Let b  U  and V  U be an open set with diam[f (V)]
                         2      n    1            2         2                      n
           /8. For any x, y  V, we have  (f (x), f(y))   (f(x), f  (x)) +  (f (x), f (b)) +  (f (b), f (y)) +
                                      2           2     n      2  n  n     2  n  n
           (f (y), f(y))  /8 + /8 + /8 + /8 = /2. Hence diam[f(V)]  /2 and therefore (f, x)  /2 for
           2  n
          every x  V. This shows D  V = Ø, proving our claim.
                                
          The claim implies that D : =    ¥ n 1 D /n is an F  set of first category in X. This completes the proof
                                   =
                                      1
                                              
          since {x  X : f is continuous at x} = X\D, and X is a complete metric space.
          We know that the derivative of a differentiable real function need not be continuous. However,
          we can say the following.
          Let f :    be differentiable. Then there exists a sequence (g ) of continuous functions from
                                                             n
           to  converging to f’ pointwise. Consequently, {x  : f is continuous at x} is a dense G  subset
                                                                                 
          of .

                                           LOVELY PROFESSIONAL UNIVERSITY                                   357
   358   359   360   361   362   363   364   365   366   367   368