Page 196 - INTRODUCTION TO THE CALCULUS OF VARIATIONS
P. 196

Chapter 1: Preliminaries                                          183

                Indeed, f being convex, we have

                                              0             0
                              f (0) ≥ f (ξ) − ξf (ξ) ⇒  lim f (ξ)= ±∞ .
                                                     ξ→±∞
                Moreover, since f  00  > 0, wehavethat f is strictly increasing and therefore
                                                     0
                invertible and we hence obtain
                                            ξ = f  0−1  (v) .
                                                              1
                The hypotheses on f clearly imply that ξ = f 0−1  is C (R).
                   Step 3. We now conclude that
                                             f  ∗0  = f 0−1 .                   (7.13)

                Indeed, we have from (7.12) that
                                                                0
                                                0
                              ∗0
                             f (v)= ξ (v)+ ξ (v) v − f (ξ (v)) ξ (v)
                                                        0
                                                0
                                                         0
                                     = ξ (v)+ ξ (v)[v − f (ξ (v))] = ξ (v) .
                                           1
                                                                             00
                                      ∗
                Wethereforehavethat f is C (R) and (7.13). Furthermore, since f > 0,we
                                                                   2
                            ∗
                deduce that f is as regular as f (so, in particular f is C ).
                                                             ∗
                Exercise 1.5.7. (i) Let h> 0. Use the convexity of f and (1.14) to write
                                                                p            p
                         0
                      hf (x) ≤ f (x + h) − f (x) ≤ α 1 (1 + |x + h| )+ α 1 (1 + |x| )
                                                                p            p
                     −hf (x) ≤ f (x − h) − f (x) ≤ α 1 (1 + |x − h| )+ α 1 (1 + |x| ) .
                         0
                We can therefore find eα 1 > 0,so that
                                                       p     p
                                              e α 1 (1 + |x| + |h| )
                                       0
                                     |f (x)| ≤                 .
                                                     h
                Choosing h =1 + |x|,wecan surely find α 2 > 0 so that (1.15) is satisfied, i.e.,
                                              ³         ´
                                                     p−1
                                     0
                                   |f (x)| ≤ α 2 1+ |x|  , ∀x ∈ R .
                The inequality (1.16) is then a consequence of (1.15) and of the mean value
                theorem.
                   (ii) Note that the convexity of f is essential in the above argument. Indeed,
                                                2
                taking, for example, f (x)= x+sin x ,we find that f satisfies (1.14) with p =1,
                but it does not verify (1.15).
                   (iii) Of course if f satisfies (1.15), we have by straight integration
                                  0
                                                      x
                                                    Z
                                      f (x)= f (0) +   f (s) ds
                                                        0
                                                     0
                that f verifies (1.14), even if f is not convex.
   191   192   193   194   195   196   197   198   199   200   201