Exercise 11.13

(continuation) If p ≡ 1 ( mod 4 ) , calculate the zeta function of y 2 = x 3 − Dx over F in terms of π and χ ( D ) , where π = − J ( χ , χ 2 ) . This calculation in somewhat sharpened form is contained in [23]. The result has played a key role in recent empirical work of B.J.Birch and H.P.F. Swinnerton-Dyer on elliptic curves.

Answers

Proof. Here p ≡ 1 ( mod 4 ) , thus p s ≡ 1 ( mod 4 ) . We consider here the two fields F = 𝔽 p and F s = 𝔽 p s , where | F | = p and | F s | = p s .

Let ρ ′ = ρ ∘ N F s ∕ F , and χ ′ = χ ∘ N F s ∕ F . The results of ¤3 show that the map ξ ↦ ξ ′ = ξ ∘ N F s ∕ F induces a group isomorphism between the group cyclic C n of characters on F whose order divides n on the group cyclic C n ′ of characters on F s whose order divides n (see Exercise 16). Thus the order of ρ ′ is 2 and the order of χ ′ is 4 , and χ ′ 2 = ρ ′ .

Replacing χ , ρ by χ ′ , ρ ′ , and p by p s , we obtain by the same reasoning that the number of projective point of C 1 in H ¯ f ( F s ) is

N s = p s + 1 + χ ′ ( − 4 ) [ χ ′ ( D ) ¯ J ( χ ′ , χ ′ 2 ) + χ ′ ( D ) J ( χ ′ , χ ′ 2 ) ¯ ] .

To compute χ ′ ( − 4 ) and χ ′ ( D ) we use the property (c) of ¤3. Since − 4 and D are in F ,

χ ′ ( − 4 ) = χ ( − 4 ) s = 1 , χ ′ ( D ) = χ ( D ) s .

Therefore

N s = p s + 1 + χ ( D ) ¯ s J ( χ ′ , χ ′ 2 ) + χ ( D ) s J ( χ ′ , χ ′ 2 ) ¯ .

It remains to compute J ( χ ′ , χ ′ 2 ) . Since χ ′ ≠ 𝜀 , χ ′ 2 ≠ 𝜀 , χ ′ 3 ≠ 𝜀 ,

J ( χ ′ , χ ′ 2 ) = g ( χ ′ ) g ( χ ′ 2 ) g ( χ ′ 3 ) .

The Hasse-Davenport relation gives g ( χ ′ k ) = − ( − g ( χ k ) ) s , thus

J ( χ ′ , χ ′ 2 ) = − [ − g ( χ ) g ( χ 2 ) g ( χ 3 ) ] s = − ( − J ( χ , χ 2 ) ) s = − π s ,

where π = − J ( χ , χ 2 ) ∈ ℤ [ i ] . To conclude,

N s = p s + 1 − χ ( D ) ¯ s π s − χ ( D ) s π ¯ s , π = − J ( χ , χ 2 ) .

Then Exercise 2 gives

Z f ( u ) = ( 1 − χ ( D ) ¯ πu ) ( 1 − χ ( D ) π ¯ u ) ( 1 − u ) ( 1 − pu ) , π = − J ( χ , χ 2 ) .

Since | π | 2 = | J ( χ , χ 2 ) | 2 = p (corollary of Theorem 1, chapter 8), expanding the numerator, we obtain

Z f ( u ) = 1 + au + p u 2 ( 1 − u ) ( 1 − pu ) , a = − tr ( χ ( D ) ¯ π ) ∈ ℤ , π = − J ( χ , χ 2 ) ∈ ℤ [ i ] .

Note: Since Z f ( u ) = exp ⁡ ( N 1 u + ⋯ ) = 1 + N 1 u + ⋯ , and

Z f ( u ) = ( 1 + au + p u 2 ) ( 1 + u + u 2 + ⋯ ) ( 1 + pu + p 2 u 2 + ⋯ ) = 1 + ( a + p + 1 ) u + ⋯ ,

the comparison of the coefficient of u in the two power series gives

a = N 1 − p − 1 , where  N 1 = p + 1 − χ ( D ) ¯ π − χ ( D ) π ¯ , π = − J ( χ , χ 2 ) .

This gives anew a = − tr ( χ ( D ) ¯ π ) . □

User profile picture
2022-07-19 00:00
Comments