Page 57 -
P. 57

k ∈ B,weget n − k ∈ S. These are the two elements of S which
                        sum to n.                            V. The Waring Problem     51
                                                                                 j
                           Repeating Lemma 1 j times, then, leads to a summing of 2 copies
                        of S and a density of 1 − (1 − αð  2 j  or more. Since this latter quantity,
                        for large enough j, will become bigger than  1  , Lemma 2 tells us
                                                                     2
                        that 2 j+1  copies of S give us all the integers, just as Schnirelmann’s
                        theorem claims.                                           Q.E.D.


                        C. Evaluation of Weyl Sums. Let b ∈ Z, b  ø       0 and k ≤ N,
                        P (n) be a polynomial of degree k with real coefficients and leading
                        coefficient integral and prime to b, and let I be an interval of length
                        ≤ N. Then

                                                 P (n)                1−k
                                             e            N   1+o(1) −2
                                                                   b
                                                  b
                                         n∈I
                        where the bound depends on k.


                           Here – as usual – we denote eàx)   e 2πix .

                           We proceed by induction on k, which represents the degree of
                        P (n). It is clearly true for k   1, and generally we may write


                                                           P(n)
                                               S        e
                                                             b
                                                    n∈I
                        and may assume w.l.o.g. that I  {1, 2, 3,...,N}. Thereby

                                        N−1
                                                              P(n) − Pàn − jð
                                  2
                               |S|                        e                       .
                                                                      b
                                      j −N+1    n∈{1,2,...,N}
                                              n∈{j+1,j+2,...,j+N}
                        This inner sum involves a polynomial of degree (k − 1) but has a
                        leading coefficient which varies with j. If we count those j which
                        produce a denominator of d, which of course must divide b, then we
                        observe that this must appear roughly d times in an interval of length
                        b. So this number of j in the full interval of length 2N + 1 is roughly
                         (2N+1)  d.
                           b
   52   53   54   55   56   57   58   59   60   61   62