Page 172 - INTRODUCTION TO THE CALCULUS OF VARIATIONS
P. 172

Thecaseofdimension 2                                              159

                We, however, know from Corollary 6.3 that
                         Z  1                  Z  1
                             ¡  02  02 ¢                           1,2
                                                     0
                              ϕ + ψ    dx ≥ 2π    ϕψ dx, ∀ϕ, ψ ∈ W per  (−1, 1)
                          −1                    −1
                which implies the claim
                               2                          1,1        1
                        [L (u, v)] − 4πM (u, v) ≥ 0, ∀u, v ∈ W  (a, b) ∩ C ([a, b]) .
                                                          per
                The uniqueness in the equality case follows also from the corresponding one in
                Corollary 6.3.
                                                                          1
                   Step 2. Wenow removethehypothesis u, v ∈ W  1,1  (a, b) ∩ C ([a, b]).As
                                                               per
                                                                            1
                before, given u, v ∈ W  1,1  (a, b),we can find u ν ,v ν ∈ W 1,1  (a, b) ∩ C ([a, b]) so
                                   per                           per
                that
                                 u ν ,v ν → u, v in W 1,1  (a, b) ∩ L ∞  (a, b) .
                Therefore, for every  > 0,wecan find ν sufficiently large so that
                                2            2
                        [L (u, v)] ≥ [L (u ν ,v ν )] −   and M (u ν ,v ν ) ≥ M (u, v) −
                and hence, combining these inequalities with Step 1, we get
                       2                       2
                [L (u, v)] −4πM (u, v) ≥ [L (u ν ,v ν )] −4πM (u ν ,v ν )−(1 + 4π)   ≥− (1 + 4π)  .
                Since   is arbitrary, we have indeed obtained the inequality.
                   We now briefly discuss the geometrical meaning of the inequality obtained
                in Theorem 6.4. Any bounded open set A, whose boundary ∂A is a closed curve
                                                      1,1
                which possesses a parametrization u, v ∈ W per  (a, b) so that its length and area
                are given by
                                              Z
                                                b p
                                                         02
                                                    02
                          L (∂A)= L (u, v)=        u + v dx
                                               a
                                                 Z  b              Z  b
                                               1
                                                                        0
                                                            0
                                                       0
                          M (A)= M (u, v)=          (uv − vu ) dx =   uv dx
                                               2  a                 a
                will therefore satisfy the isoperimetric inequality
                                              2
                                       [L (∂A)] − 4πM (A) ≥ 0 .
                This is, of course, the case for any simple closed smooth curve, whose interior is
                A.
                   One should also note that very wild sets A can be allowed. Indeed sets A
                that can be approximated by sets A ν that satisfy the isoperimetric inequality
                and which are so that
                          L (∂A ν ) → L (∂A) and M (A ν ) → M (A) , as ν →∞
                also verify the inequality.
   167   168   169   170   171   172   173   174   175   176   177