Page 227 - INTRODUCTION TO THE CALCULUS OF VARIATIONS
P. 227

214                                            Solutions to the Exercises

                       we find that
                                              ϕ xx  = f, ϕ xy  = g, ϕ yy  = h
                       and hence that
                                                 ϕ ϕ    − ϕ 2 xy  =1 .
                                                   xx yy
                       The fact that ϕ is convex follows from the above identity, ϕ  > 0, ϕ  > 0 and
                                                                           xx      yy
                       Theorem 1.50.

                       7.6    Chapter 6: Isoperimetric inequality


                       7.6.1   The case of dimension 2
                       Exercise 6.2.1. One can consult Hardy-Littlewood-Polya [55], page 185, for
                       more details. Let u ∈ X where
                                     ½                                  Z  1     ¾
                                X =   u ∈ W  1,2  (−1, 1) : u (−1) = u (1) with  u =0 .
                                                                         −1
                       Define
                                               z (x)= u (x +1) − u (x)

                       and note that z (−1) = −z (0),since u (−1) = u (1). Wededucethatwecan find
                       α ∈ (−1, 0] so that z (α)=0, which means that u (α +1) = u (α).We denote
                       this common value by a (i.e. u (α +1) = u (α)= a). Since u ∈ W 1,2  (−1, 1) it
                                                                    2
                       is easy to see that the function v (x)= (u (x) − a) cot (π (x − α)) vanishes at
                       x = α and x = α +1 (this follows from Hölder inequality, see Exercise 1.4.3).
                       We therefore have (recalling that u (−1) = u (1))

                                   Z
                                     1 n             2                           2  o
                                         02
                                              2
                                        u − π (u − a) − (u − π (u − a)cot π (x − α))  dx
                                                          0
                                    −1
                                    h                       i 1
                                              2
                               = π (u (x) − a) cot (π (x − α))  =0 .
                                                             −1
                            R  1
                       Since    u =0,weget from theabove identity that
                             −1
                            Z                          Z
                              1                          1
                                ¡  02  2 2 ¢      2 2                               2
                                                             0
                                 u − π u   dx =2π a +      (u − π (u − a)cot π (x − α)) dx
                             −1                          −1
                       and hence Wirtinger inequality follows. Moreover we have equality in Wirtinger
                       inequality if and only if a =0 and,c denoting a constant,
                                       0
                                      u = πu cot π (x − α) ⇔ u = c sin π (x − α) .
   222   223   224   225   226   227   228   229   230   231   232