Page 294 -
P. 294

§8. Tate’s Description of Homomorphisms  271


                                                           u 0
        Proof. In an extension of Q   diagonalize the action of π E to  on V   (E) and
                                                           0 v

                         u  0                                   2
        the action of π E to   , where u,v are the conjugate roots of T f E (1/T ) and
                          0 v

                               ab
                 2



        u ,v of T f E (1/T ).For     to be in Hom (Fr) (V   (E), V   (E )) it is necessary

                               cd
        and sufficient that

            au  bv      a   b    u  0      u   0    a   b      au   bu
                     =                 =                   =             .
            cu   cu     c   d    0  v      0   u    c   d      cv    dv

                      ab
        If one entry of    is nonzero, then one eigenvalue of π E is equal to one eigen-
                      cd


        value of π E and since they are conjugate pairs, the sets {u,v} and {u ,v } are equal.

        When u = v, the supersingular case, we see that Hom (Fr) consists of all homo-
        morphisms and has rank 4. When the eigenvalues are distinct, we cannot have both




        a  = 0and c  = 0 otherwise u = u and u = v contradicting u = v and, sim-
        ilarly, we cannot have both b  = 0and d  = 0 or it would contradict u  = v. Thus

                                         a 0         0 b
        Hom (Fr) (V   (E), V   (E )) consists of all  or all  and hence has rank 2.

                                         0 d         c 0
        This proves the proposition.
        (8.3) Remark. For two isogenies u : E → E and v : E → E the composite




        vu : E → E is an isogeny, in particular, nonzero. This means that the pairings


                        End k (E) × Hom k (E, E ) → Hom k (E, E )
        are nondegenerate. We know that rank End k (E) = 2 for E ordinary (not su-
        persingular) and rank End k (E) = 4 for E supersingular. Hence we see that for



        Hom k (E, E )  = 0, it follows that rank Hom k (E, E ) = rank Hom (Fr) (T   (E), T   (E )),

        and this proves Theorem (8.1) in the case where Hom k (E, E )  = 0. In other words,


                     T   :Hom k (E, E ) ⊗ Z   → Hom (Fr) (T   (E), T   (E ))
        is a monomorphism between two modules of the same rank with torsion free coker-
        nel, and, therefore, T   is an isomorphism.
        (8.4) Theorem. Let E and E be two elliptic curves over a finite field k=F q . Then

        the following are equivalent:

         (1) E and E are isogenous over k.
         (2) V   (E) and V   (E ) are Gal(k/k)-isomorphic modules.

         (3) #E(k) = #E (k), i.e., the two curves have the same number of elements over k.

         (4) ζ E (s) = ζ E (s), the two curves have the same zeta functions.

        Proof. The equivalence of (1) and (2) follows by Tate’s theorem, (8.1), since

                   Hom(E, E ) ⊗ Q    and  Hom      (V   (E), V   (E ))
                                             Gal( ¯ k/k)
   289   290   291   292   293   294   295   296   297   298   299