Page 42 - INTRODUCTION TO THE CALCULUS OF VARIATIONS
P. 42

Sobolev spaces                                                     29

                   We will now give a simple characterization of W 1,p  which will turn out to be
                particularly helpful when dealing with regularity problems (Chapter 4).
                                                                   p
                                       n
                Theorem 1.36 Let Ω ⊂ R be open, 1 <p ≤∞ and u ∈ L (Ω). The following
                properties are then equivalent.
                   (i) u ∈ W  1,p  (Ω);

                   (ii) there exists a constant c = c (u, Ω,p) so that

                       ¯Z               ¯
                       ¯       ∂ϕ       ¯
                                        ¯
                       ¯  u (x)   (x) dx ≤ c kϕk L p 0 , ∀ϕ ∈ C 0 ∞  (Ω) , ∀i =1, 2, ..., n
                       ¯                ¯
                         Ω     ∂x i
                (recalling that 1/p +1/p =1);
                                     0
                   (iii) there exists a constant c = c (u, Ω,p) so that for every open set ω ⊂
                                                                             c
                                                         n
                ω ⊂ Ω,with ω compact, and for every h ∈ R with |h| < dist (ω, Ω ) (where
                      n
                 c
                Ω = R \ Ω), then
                                                   ¶ 1
                            µZ
                                                     p
                                                p
                                |u (x + h) − u (x)| dx  ≤ c |h| if 1 <p < ∞
                               ω
                         |u (x + h) − u (x)| ≤ c |h| for almost every x ∈ ω if p = ∞ .
                   Furthermore one can choose c = k∇uk L p in (ii) and (iii).
                Remark 1.37 (i)Asa consequenceofthe theorem, it caneasilybeprovedthat
                if Ω is bounded and open then
                                             ¡ ¢
                                           0,1
                                         C    Ω ⊂ W  1,∞  (Ω)
                         ¡ ¢
                where C 0,1  Ω has been defined in Section 1.2, and the inclusion is, in general,
                strict. If, however, the set Ω is also convex (or sufficiently regular, see Theorem
                5.8.4 in Evans [43]), then these two sets coincide (as usual, up to the choice of a
                representative in W  1,∞  (Ω)). In other words we can say that the set of Lipschitz
                functions over Ω can be identified, if Ω is convex, with the space W  1,∞  (Ω).
                   (ii) The theorem is false when p =1. Wethenonlyhave(i) ⇒ (ii) ⇔
                (iii). The functions satisfying (ii) or (iii) are then called functions of bounded
                variations.
                   Proof. We will prove the theorem only when n =1 and Ω =(a, b).For the
                more general case see, for example, Proposition IX.3 in Brézis [14] or Theorem
                5.8.3 and 5.8.4 in Evans [43].
   37   38   39   40   41   42   43   44   45   46   47