Page 66 - INTRODUCTION TO THE CALCULUS OF VARIATIONS
P. 66

Euler-Lagrange equation                                            53

                           ©     1                        ª
                where X = u ∈ C ([0, 1]) : u (0) = 1,u (1) = 0 . We will show that (P) has
                                   1
                    1
                no C or piecewise C solution (not even in any Sobolev space). The Euler-
                Lagrange equation is
                                            c
                              0 0
                                        0
                           (xu ) =0 ⇒ u =     ⇒ u (x)= c log x + d, x ∈ (0, 1)
                                            x
                where c and d are constants. Observe first that such a u cannot satisfy simulta-
                neously u (0) = 1 and u (1) = 0.
                   We associate to (P) the following problem
                                         ½       Z               ¾
                                                   1
                                                          0
                           (P piec )  inf  I (u)=   f (x, u (x)) dx  = m piec
                                   u∈X pie c
                                                  0
                                    ©     1                          ª
                             X piec = u ∈ C  ([0, 1]) : u (0) = 1,u (1) = 0 .
                                          piec
                   We next prove that neither (P) nor (P piec ) have a minimizer. For both cases
                it is sufficient to establish that m piec = m =0. Let us postpone for a moment the
                                                                            1
                proof of these facts and show the claim. If there exists a piecewise C function
                v satisfying I (v)= 0, this would imply that v =0 almost everywhere in (0, 1).
                                                        0
                Since the function v ∈ X piec , it should be continuous and v (1) should be equal
                to 0, we would then deduce that v ≡ 0, which does not verify the other boundary
                condition, namely v (0) = 1. Hence neither (P) nor (P piec ) have a minimizer.
                   We first prove that m piec =0.Let n ∈ N and consider the sequence
                                           ⎧              £  1  ¤
                                                1    if x ∈ 0,
                                           ⎨                 n
                                   u n (x)=
                                           ⎩ − log x      ¡  1  ¤
                                                     if x ∈  , 1 .
                                               log n       n
                                        1
                Note that u n is piecewise C , u n (0) = 1, u n (1) = 0 and
                                              1
                                     I (u n )=    → 0,as n →∞,
                                             log n
                hence m piec =0.
                   We finally prove that m =0. This can be done in two different ways. A more
                sophisticated argument is given in Exercise 2.2.6 and it provides an interesting
                continuity argument. A possible approach is to consider the following sequence
                                     ⎧
                                        −n 2  2   n             £  1  ¤
                                     ⎪      x +     x +1 if x ∈ 0,
                                     ⎨  log n   log n              n
                             u n (x)=
                                     ⎪        − log x           ¡ 1  ¤
                                     ⎩                     if x ∈  , 1 .
                                               log n             n
                We easily have u n ∈ X and since
                                        ⎧                     £    ¤
                                            n  (1 − 2nx) if x ∈ 0,  1
                                        ⎨ log n                  n
                                 0
                                u (x)=
                                 n
                                        ⎩       −1            ¡  1  ¤
                                              x log n    if x ∈  n  , 1
   61   62   63   64   65   66   67   68   69   70   71