Page 210 - INTRODUCTION TO THE CALCULUS OF VARIATIONS
P. 210

Chapter 3: Direct methods                                         197

                Invoking Poincaré inequality (cf. Theorem 1.47) we can find γ ,γ ,γ ,sothat
                                                                      3
                                                                            5
                                                                         4
                                         p           p            q
                          m +1 ≥ γ ku ν k W 1,p − γ ku 0 k W 1,p − γ ku ν k W 1,p − γ 5
                                   3
                                                             1
                                                4
                and hence, γ being a constant,
                           6
                                              p            q
                                m +1 ≥ γ ku ν k  1,p − γ ku ν k  1,p − γ .
                                         3    W       1    W       6
                Since 1 ≤ q< p,wecan find γ ,γ so that
                                          7
                                             8
                                                     p
                                       m +1 ≥ γ ku ν k W 1,p − γ 8
                                                7
                which, combined with the fact that m< ∞, leads to the claim, namely
                                            ku ν k W 1,p ≤ γ .
                                                       9
                Exercise 3.3.2. This time it is the lower semicontinuity step in Theorem 3.3
                that has to be changed. Let therefore u ν   u in W 1,p  and let
                                         I (u)= I 1 (u)+ I 2 (u)

                where            Z                         Z
                          I 1 (u)=  h (x, ∇u (x)) dx , I 2 (u)=  g (x, u (x)) dx .
                                   Ω                        Ω
                It is clear that, by the proof of the theorem, lim inf I 1 (u ν ) ≥ I 1 (u).The result
                will thereforebeprovedifwecan show

                                          lim I 2 (u ν )= I 2 (u) .
                                         i→∞
                   Case 1: p> n. From Rellich theorem (Theorem 1.43) we obtain that u ν → u
                    ∞
                in L ; in particular there exists R> 0 such that ku ν k  ∞ , kuk  ∞ ≤ R.The
                                                                 L      L
                result then follows since
                                     Z
                    |I 2 (u ν ) − I 2 (u)| ≤  |g (x, u ν ) − g (x, u)| dx ≤ γ ku ν − uk  ∞ meas Ω.
                                                                       L
                                      Ω
                   Case 2: p = n. The estimate
                                           Z
                                             ³                  ´
                                                     q−1     q−1
                         |I 2 (u ν ) − I 2 (u)| ≤ γ  1+ |u ν |  + |u|  |u ν − u| dx ,
                                            Ω
                combined with Hölder inequality gives us
                                    µZ                       q ¶ q−1 µZ        ¶ 1
                                        ³                 ´      q               q
                                                q−1    q−1  q−1               q
                  |I 2 (u ν ) − I 2 (u)| ≤ γ  1+ |u ν |  + |u|          |u ν − u|  .
                                      Ω                               Ω
   205   206   207   208   209   210   211   212   213   214   215