Page 157 -
P. 157

134    6. Proof of Mordell’s Finite Generation Theorem

        Proof. By (5.5) the function q is quasiquadraticif and only if the weak parallelogram
        law holds q(x + y) + q(x − y) ∼ 2q(x) + 2q(y). Setting x = y we have q(2x) ∼
                                                                      n
        4q(x),thatis, |q(2x) − 4q(x)|≤ A for positive constant A. Replacing x by 2 x,we
        obtain the inequality
                                                n
                        |2 −2(n+1) q(2 n+1 x) − 2 −2n q(2 x)|= 2 −2n  A
        which leads to the following estimate for all n and p
                                                          4A
                                                      −2n
                        −2(n+p)
                                 n+p
                                               n
                                         −2n
                      |2      q(2   x) − 2  q(2 x)|= 2   ·   .
                                                           3
                               ∗
        Thus the sequence defining q (x) is Cauchy, and, therefore, it is convergent. The last
                                                                    ∗
        assertion follows now from the previous proposition since the condition q (2x) =
                                                           ∗
          ∗
                                                                   ∗
        4q (x) follows from the defining limit and the conditions that q (x) = q (−x) and
        q (x) be quasiquadraticare preserved in the limit. This proves the theorem.
         ∗
           Another important situation where an equivalence can be modified to become an
        equality is contained in the next proposition.
        (5.8) Proposition. Let f : X → X be a function, let d > 1, and let h : X → R
                                                                       n
        be a function such that h f ∼ d · h. Then the limit h f (x) = lim n→∞ d −n  · h( f (x))
        exists uniformly on X and:
                                                 2
                                                         2
         (1) h f ( f (x)) = dh f (x) and |h f (x) − h(x)|≤ cd /(d − 1) where
                                 |h( f (x)) − d · h(x)|≤ c
            for all x ∈ X.
         (2) If h is proper, then h f is proper.



         (3) If g : X → X with hg ∼ d · h for a constant d ,thenh f g ∼ d · h f holds on X.
        Proof. From the inequality |h( f (x) − d · h(x)|≤ c,wehave
                                              n
                       |d −n−1 h( f  n+1 (x) − d −n h( f (x))|≤ cd −n−1 ,
        and, therefore, we have the estimate
                                                                n

                                                       c     1
                            n+p
                    −(n+p)
                                      −n
                                            n
                  |d     h( f  (x)) − d  h( f (x))|≤             .
                                                   1 − (1/d) d
                   n
        Thus d −n h( f (x)) is a Cauchy sequence, and it converges uniformly to h f .
           As for assertion (1), the relation h f ( f (x)) = d · h f (x) follows since the two
        sides are rearrangements of the same limit. The second part results from the above
        estimate by letting p →∞ and taking n = 0. Assertion (2) follows from (1) and
        (5.3). As for (3), we calculate the difference

           h f (g(x)) − d · h f (x)


               = [h f (g(x)) − h(g(x))] + [h(g(x)) − d · h(x)] + [d · h(x) − h f (x)],
        and see that each of the three terms is bounded. This proves the proposition.
   152   153   154   155   156   157   158   159   160   161   162