Page 320 - DMTH401_REAL ANALYSIS
P. 320

Real Analysis




                    Notes


                                     Notes Above Lemma 4 is false if the condition m(F ) < ¥ is missing.
                                                                                 1
                                   Now, we are ready to prove Littlewood’s 2nd Principle.
                                   Proof: Let  > 0 be given. Our proof is divided into two steps.
                                   Step I: Assume m  f  M for some m, M  .
                                   We divide [m, M] into n subintervals such that the length of each  subinterval is less than  .
                                   Symbolically, we take the partition points as follows:
                                                  m = y  < y  <  < y  = M  with y  – y  <for 1  i  n.
                                                      0   1      n            i   i–1
                                   Let E  = {x  E : m  f(x)  y }, E  = {x  E : y  < f(x)  y  },..., E  = {x  E : y  < f(x)  M }. Now, take
                                       1                1  2        1       2     n        n–1
                                    = y  + y  +  + y  . Since E , E ,..., E are all measurable (why?),  is simple and satisfies
                                       1 E 1    2 E 2   n E n  1  2  n
                                   the inequality |f – | <  with no exceptions.
                                   Step II: General case.
                                   We let
                                                            F  = {x  E : |f(x)|  n}.
                                                             n
                                   Then F   F   . Note that m(F )  m(E) < ¥ and m(F ) = 0 by assumption, apply Lemma 4 there
                                        1   2               1                ¥
                                   exists N   such that
                                                                     m(F ) < .
                                                                        N
                                   Now, let f* = (–N  f)  N, then f = f* on E except on a set of measure less than . From the result
                                   of Step I, there is a simple function  such that |f* – | <  on E. Hence
                                                   |f – | <  on E except on a set of measure less than .

                                   Corollary: There is a  sequence of  simple  functions     such that      f pointwisely  almost
                                                                               n          n
                                   everywhere on E. If E = [a, b], there are also sequence of step functions and sequence of continuous
                                   functions converging to f pointwisely almost everywhere on [a,b].
                                                                             n
                                   Proof: Applying Littlewood’s 2nd Principle to  = 1/2 , there are simple functions   and sets A
                                                                                                     n        n
                                                n
                                   with m(A ) < 1/2  such that
                                          n
                                                                1
                                                        |f –  |<      on E\A .
                                                            n    n           n
                                                                2
                                   Let A = lim A  :=  ¥ k 1 ( ¥ n k A ), then m(A) = 0 (why?). The proof is completed by noting that    f
                                             n    =   =  n                                                 n
                                   pointwisely on E\A.

                                     Notes In fact, the sequence   can be chosen so that    f pointwisely everywhere on E.
                                                             n                  n
                                                                                     2
                                     For example, we can first divide the interval [–n, n] into 2n  subintervals such that each
                                     subinterval has length 1/n, i.e. choose
                                                           –n = y  < y  <  < y  2 = n
                                                                0   1      2n
                                     such that y  – y  = 1/n for all i. Then let
                                              i  i–1
                                                                ì y i  if y   f(x) <  y i 1  for some i
                                                                                +
                                                                       i
                                                                ï
                                                           (x) = í n  if f(x)   n
                                                           n
                                                                ï - n if f(x) < - n
                                                                î
          314                               LOVELY PROFESSIONAL UNIVERSITY
   315   316   317   318   319   320   321   322   323   324   325