Page 44 - INTRODUCTION TO THE CALCULUS OF VARIATIONS
P. 44

Sobolev spaces                                                     31

                Since a similar argument holds for h< 0, we deduce from (1.8) and (1.9) that,
                if 1 <p ≤∞,
                              ¯                         ¯
                               Z  b
                              ¯                         ¯
                              ¯                         ¯
                                  u (x)[ϕ (x − h) − ϕ (x)] dx¯ ≤ c |h|kϕk L p 0 .
                              ¯
                              ¯  a                      ¯
                Letting |h| tend to zero, we get
                                 ¯        ¯
                                  Z  b
                                 ¯        ¯
                                 ¯      0  ¯                 ∞
                                     uϕ dx¯ ≤ c kϕk L p 0 , ∀ϕ ∈ C  (a, b)
                                 ¯                           0
                                 ¯  a     ¯
                which is exactly (ii).
                   (i) ⇒ (iii). From Lemma 1.38 below, we have for every x ∈ ω
                                           Z  x+h          Z  1
                          u (x + h) − u (x)=     u (t) dt = h  u (x + sh) ds
                                                  0
                                                                0
                                            x               0
                and hence
                                                    Z  1
                                                         0
                               |u (x + h) − u (x)| ≤ |h|  |u (x + sh)| ds .
                                                     0
                Let 1 <p< ∞ (the conclusion is obvious if p = ∞), we have from Hölder
                inequality that
                                                       1
                                                     Z
                                              p     p              p
                                                          0
                               |u (x + h) − u (x)| ≤ |h|  |u (x + sh)| ds
                                                      0
                and hence after integration
                        Z                             Z Z  1
                                          p          p                p
                                                              0
                          |u (x + h) − u (x)| dx ≤ |h|      |u (x + sh)| dsdx
                         ω                             ω  0
                                                      Z  1  Z
                                                     p                p
                                                              0
                                               = |h|        |u (x + sh)| dxds .
                                                       0   ω
                Since ω + sh ⊂ (a, b),we find
                             Z                  Z
                                          p                 p       0 p
                                                       0
                                |u (x + sh)| dx =     |u (y)| dy ≤ ku k L p
                                  0
                              ω                  ω+sh
                and hence
                                                       ¶ 1
                                µZ
                                                         p
                                                   p
                                                              0
                                    |u (x + h) − u (x)| dx  ≤ ku k  p |h|
                                                                L
                                  ω
                which is the claim.
                   In the proof of Theorem 1.36, we have used a result that, roughly speaking,
                                                                                   p
                says that functions in W  1,p  are continuous and are primitives of functions in L .
   39   40   41   42   43   44   45   46   47   48   49