Page 145 -
P. 145

122    5. Reduction mod p and Torsion Points

        (7.5) Definition. An elliptic curve E over K has potential good reduction provided
        there exists a finite extension L and an extension w of v to L such that E over L has
        good reduction at the valuation w.

           Be extending the ground field from K to L, the minimal equation of E can change
        to one with good reduction, or, in other words, E can become isomorphic over L to


        a curve E with good reduction at w.Inthe above case E was the Legendre or
        Hessian form of the equation. In general the discriminant   of E over K is different


        from the discriminant   of E over L and in this case of potential good reduction


        v( ) > 0and w(  ) = 0 while j(E) = j(E ). Since any curve E with good

        reduction has v( j(E )) ≥ 0, it follows immediately that if E has potential good

        reduction, then j(E) is a local integer, i.e., v( j(E)) ≥ 0. The above argument with
        Legendre or Hessian forms of the equations shows that the converse holds, and this
        is the theorem of Deuring.
        (7.6) Theorem. An elliptic curve E defined over K has potential good reduction if
        and only if j(E) is a local integer, i.e., v( j(E)) ≥ 0.
        §8. Tate’s Theorem on Good Reduction over the Rational
            Numbers

        Tate’s theorem says that over the rational numbers there are no elliptic curves with
        good reduction everywhere. We give a proof using formulas for coefficients follow-
        ing Ogg [1966].
        (8.1) Theorem. Every elliptic curve E over the rational numbers Q has bad reduc-
        tion at some prime, i.e.,   cannot be equal to ±1.
        Proof. Assume that   =±1. Recall from 3(3.1) that

                                 2       3     2
                             =−b b 8 − 8b − 27b − 9b 2 b 4 b 6
                                 2       4     6
        and
                                2
                          b 2 = a + 4a 2 ,  b 4 = a 1 a 3 + 2a 4 .
                                1
        If a 1 is even, then 4|b 2 and 2|b 4 so that b 6 will have to be odd, and in fact, ±1 =
               2
          ≡ 5b (mod 8). Since any square modulo 8 is conguent to 0, 1, 4 (mod 8), this is
               6
        impossible.
           Now consider the case where a 1 is odd, and hence b 2 is also odd. Then the co-
                      2
        efficient c 4 = b − 24b 4 ≡ 1 (mod 8). We write c 4 = x ± 12 from the relation
                     2
                            3
         3
                   3
             2
        c − c = 12   =±12 , and thus
         4   6
                                               2
                                         3
                            2
                      2
                     c = x(x ± 36x + 3 · 12 ) ≡ x (x + 4)  (mod 8).
                      6
        This means that x ≡ 5(mod8). Now3|x, for otherwise any p|x with p > 3, it would
                   2
        follow that p |x and ±x would be a square. This would contradict x ≡ 5(mod8).
   140   141   142   143   144   145   146   147   148   149   150