Page 131 -
P. 131

108    5. Reduction mod p and Torsion Points


                                 2   3    c 4    c 6
                                y = x −       −    .
                                          48    864
                                           6
                               4
        By the above remark, if 48p | c 4 and 864p | c 6 , then the equation is not minimal.
                                              4
                                                           5
                                                              3
        Since the equation is minimal and since 48 = 2 · 3 and 864 = 2 · 3 , it follows that
                v( ) + v( j) = 3v(c 4 )< 12 + 3v(48) = 12 + 12v(2) + 3v(3),
        or
                           3
             v( ) + v( j − 12 ) = 6v(c 6 )< 12 + 2v(864) = 12 + 10v(2) + 6v(3).
                                                          3
        Since v( ) + min{v( j), 0}≤ v( ) + v( j) or v( ) + v( j − 12 ), we obtain the first
        inequality.
           For the second statement observe that for v(2) = v(3) = 0, the minimal model
        satisfies v( ) + min{v( j), 0} < 12. The converse holds by Remark (2.4) since 0 <
        v( ) + min{v( j), 0} and the relation between two valuations of the discriminants
        given above (2.3). This proves the proposition.

           Now we return to the notations of (1.1) where R is a factorial ring with field
        of fractions k. Two normal forms of equations for E with coefficients a j in R are
                                                              2
        related by an admissible change of variables 3(2.3) where x = u ¯x + r and y =
                2
         3
        u ¯y + su ¯x + t. The discriminants are related by u   =   by 3(3.2). For an
                                                    12 ¯
        irreducible element p in R we have
                            ord p ( ) = 12 ord p (u) + ord p ( ¯  ).
        This leads to the global version of (2.2) since by a change of variable we can always
        choose an equation where ord p ( ) is minimal for all irreducibles p in R.
        (2.6) Definition. Let k be the field of fractions of a factorial ring R, and let E be
        an elliptic curve over k. A minimal normal form for E is a normal form with all
        a j in R such that all ord p ( ) is minimal among all equations in normal form with
        coefficients a j in R.
           By the above discussion a minimal normal form for an elliptic curve E always
        exists.
           Unfortunately we are interested in rings R, namely the ring of integers in a num-
        ber field k, which are not always factorial. In the case there is only a minimal model
        of an elliptic curve E locally at valuations v of k with valuation ring R (v) associ-
        ated with prime ideals of R. For questions of reduction mod v it is good enough to
        work locally with coefficients in R (v) . We return later to the more general case of a
        Dedekind ring R to define the conductor of an elliptic curve. In any case the present
        theory includes curves defined over Q.

        Exercise

         1. Over the rational numbers Q under what conditions on the constant a are the normal
                 2
                     3
                                   3
                               2
            forms y = x + ax and y = x + a minimal.
   126   127   128   129   130   131   132   133   134   135   136