• 沒有找到結果。

The Proof of the Main Theorem

在文檔中 橢圓曲線的二次扭變 (頁 39-48)

In this chapter, we complete the proof of our Main Theorem I. According to Proposition IV.1, if the local symbols map s is injective but not surjective, then the theorem is proved. But this condition does not always hold, as shown in the case of congruent curve. Our strategy is to find a prime number q 6∈ T0, so that for the q-twist Eq, the associated local symbols map,

s(q) : Sel2(Eq/Q) −→ f(q),

is injective but not surjective. Then we can complete the proof of the Main Theorem I, by applying Proposition IV.1 again, taking the ±pq twist of Eq, which is the same as E±p. Thus, it is enough for us to prove the following:

Proposition V.1. Let E/Q be an elliptic curve defined by the Weierstrass equation

y2 = x(x − a)(x − b),

with a, b ∈ Z, 0 < a < b and (a, b) = 1 or 2. Then there exists an odd prime number q, relatively prime to ab(b − a), so that the local symbols map s(q) is injective but not surjective.

For the rest of this Chapter, we assume that E is defined as in the proposition.

1. The Injectivity of s(q)

By Corollary IV.1, the kernel of s(q)is contained in Hom(Gal(Q(√

2)/Q), Eq[2]).

On the other hand, since Eq has potential good additive reduction at q, an element ϕ ∈ Sel2(Eq/Q) has local symbol {ϕ}q= 0 if and only if Pϕq ∈ 2Eq(Qq) (see Section IV,4), or equivalently, the extension K(ϕ)/Q splits at q. However, the extension Q(

√2)/Q only splits at odd primes that are congruent to ±1 modulo 8. Therefore, we have proved the following:

Lemma V.1. If q is a prime number relatively prime to ab(b − a) with q ≡ ±3 (mod 8),

then the map s(q) is injective.

2. The Non-surjectivity of s(q)

Let q be congruent to ±3 modulo 8 so that by Lemma V.1, the map s(q) is injective. Write T00(q) = T00∪ {q} and f(q) =Q

v∈T00(q)f(q)v .

Suppose that s(q) is surjective and let Λ ⊂ Sel2(Eq/Q) denote the pre-image of f(q)q considered as a subspace of f(q). Then by Lemma IV.5, we have Λ ' Eq[2], and

39

hence |Λ| = 4. Lemma IV.7 implies that each element in Λ is unramified outside field of the decomposition subgroup < [v] > of Gal(Q(√

2,√ q)/Q).

Lemma V.2. Let the notation be as above. Then there exists a proper subgroup D ⊂ E[2] such that for each ϕ ∈ Λ, we have ϕ([v]) ∈ D.

Proof. Since Eq has multiplicative reduction at v, by Lemma IV.2 and Lemma IV.4, for each ϕ ∈ Λ, the point ϕ([v]) has smooth reduction at v. Let D be the subgroup of Eq[2] consisting of points with smooth reduction. Then |D| = 2.  Since |Λ| = 4 and |D| = 2, there exists a non-trivial element ϕ ∈ Λ such that

Since E has good reduction at q, for the reduction curve ¯E, we have

| ¯E(Fq)| = 1 + q − α, for some integer α of absolute value not greater than 2√

q. Let ¯E0/Fq denote the quadratic twist ¯E. Then we have

| ¯E0(Fq)| = 1 + q + α.

Consider the composition

jq: Hom(Gal(K/Q), Eq[2]) ,→ H1(Q, Eq[2]) −→ H1(Q, Eq) −→ H1(Qq, Eq), where the first arrow is the natural embedding, the second is induced from the inclusion Eq[2] ⊂ Eq( ¯Q) and the last is the localization map.

Lemma V.3. Let the notation be as above. Then the following are true:

(a) If K = Q(√

2), then jq is injective.

(b) If K = Q(√

q) and | ¯E(Fq)| ≡ 4 mod 8, then jq is injective.

(c) If K = Q(√

2q) and | ¯E0(Fq)| ≡ 4 mod 8, then jq is injective.

Proof. For a local field K of residual characteristic q, the topological group Eq(K) = A × B where A is a compact q-group and B is a finite discrete group. We will simply call the Sylow 2-subgroup of B the 2-part of Eq(K).

Suppose K = Q(√

2). Then the extension K/Q is inert at q. Therefore, Eq/Kq has additive reduction and the the 2-part of Eq(Kq) is Eq[2] (Lemma IV.5). In par-ticular, the natural map Eq(Qq)/2Eq(Qq) −→ Eq(Kq)/2Eq(Kq) is an isomorphism.

This implies the injectivity of jq. Suppose K = Q(√

q). Then the extension K/Q is totally ramified at q and Eq/Kq has good reduction and as groups Eq(Kq) and E(Kq) are isomorphic. Thus, the 2-parts of Eq(Kq), E(Kq) and ¯E(Fq) are isomorphic. The condition of the lemma implies that the 2-part of Eq(Kq) equals Eq[2]. Then we also deduce that the natural

2. THE NON-SURJECTIVITY OF s(q) 41

map Eq(Qq)/2Eq(Qq) −→ Eq(Kq)/2Eq(Kq) is an isomorphism and hence the map jq is injective.

Finally, if K = Q(√

2q), then the reduction of Eq/Kq equals ¯E0. A similar

argument also leads to the injectivity of jq. 

Lemma V.4. Let q be a prime number congruent to ±3 modulo 8. Then the map s(q) is not surjective, if there is a prime number l | ab(a − b) satisfying one of the following conditions: as in Lemma V.2. Then K/Q splits at v and hence the residue class of l modulo 8 determines K, since K is an intermediate field of the extension Q(√

2,√ q)/Q.

Suppose (a) holds. Then K = Q(√

2) and Lemma V.3 implies that ϕ can not be contained in Sel2(Eq/Q), a contradiction. If either (b) or (c) holds, then K = Ki = Q(Qi). Consider the following three mutual exclusive cases:

(1) There exists some i ∈ {0, a, b} so that √

(3) There exists some Ki(√

−1) that does not contains √

2 and every such Ki(√

−1) contains √

l for all odd prime divisor l of ab(b − a).

Suppose (1) holds. Then there exists an element λ ∈ Gal(Ki(√

l)) is a non-trivial automorphism of the field. By the density theorem, there is a prime number q so that the Frobenius element [q] ∈ Gal(Ki(√

−1,√ 2,√

l)/Q) equals λ. Then q splits completely over Ki(√ has at least an solution in the residue field of each Ki. Therefore, none of these residue field equals Fq, and consequently, | ¯E(Fq)| ≡ 4 (mod 8). We see that the condition (d) of Lemma V.4 holds.

Suppose the condition (3) holds. Note that Ki = Q(√ α,√

β) where α, β ∈ {±a, ±b, ±(a − b)}. Therefore, we must have

{a, b, b − a} = {2n0, ln11, l2n2}, n1 > 0.

Furthermore, we know that n0 must be odd, for otherwise we shall deduce the contradiction that neither √

First, let us assume that n2 > 0. We shall show that it is enough to choose q so that

q)/Q) is the just generator, denoted as h, of the group Gal((Q(√ the equality holds, since both groups are of order 4. For simplicity, write

G = Gal(Qq(√ identified with each other. Via this identification, we deduce that for an element ψ ∈ Hom(G, Eq[2]), the local class [ψv] ∈ H1(G, Eq(Qq(√

q))) is trivial if and only

3. THE EXAMPLES OF THE CONGRUENT CURVE 43

the reduction ψ(g) (considered as a element of ¯E(Fq)) is contained in NG( ¯E(Fq)).

Since G acts trivially on ¯E(Fq), we have NG( ¯E(Fq)) = 2 ¯E(Fq). Also, the condition (49) implies that Fq contains no primitive 4’th root of one, and hence

E[4] * ¯¯ E(Fq). (50)

This shows that the natural map

E[2] −→ ¯E[2] −→ ¯E(Fq)/ NG( ¯E(Fq))

is not the trivial map. In particular, there is an element ψ ∈ Λ = Hom(G, Eq[2]) such that [ψv] 6= 0. This contradicts to the fact that Λ ⊂ Sel2(Eq/Q).

At last, we consider the case where (a, b) = (1, 3), or (a, b) = (2, 3). We shall show that if q ≡ 5 (mod 8), then the map s(q) is not surjective, by showing that there is no ψ ∈ Sel2(Eq/Q) so that the local symbols {ψ} 6= 0 and {ψ}v = 0, for v 6= ∞.

If such ψ were to exist, it must be unramified outside 2 and ∞. Therefore, ψ ∈ Hom(Gal(Q(√

−1,√

2)/Q), Eq[2]). But, since ψq is trivial (as Eq has potential good additive reduction at q), we have ψ ∈ Hom(Gal(Q(√

−1)/Q), Eq[2]). Denote G = Gal(Q(√

−1)/Q) =< g >. Then we have ψ(g) = P0 ∈ Eq[2] with P0 = (0, 0).

However, since g is the Frobenius at 3 and the reduction of P0 at 3 is not a smooth point on ¯Eq, we know that either [ψ]3 6= 0 or {ψ}3 6= 0, a contradiction.

 3. The examples of the Congruent Curve

In this section, we complete the proof of Proposition V.1. We shall deal with the cases (a, b) = (1, 2), (2, 4) by direct computation. Doing so, we also get a proof that primes congruent to 3 modulo 8 are not congruent numbers.

3.1. Consider that (a, b) = (1, 2). Then the elliptic curve is the congruent curve:

E : y2 = x(x − 1)(x − 2).

Let q = 3 and denote the 3-twist by E3. Then

T0(3) = {∞, 2, 3} and E3[2] = {O, P0, P3, P6},

where Pi = (i, 0) for i = 0, 3 and 6. Let σv, τv, δ−1 and δ3 be as in the section IV.7.

As in Section II.4, we choose

AT0 := {ρ(pv,P0), ρ(pv,P3)| v ∈ T0(3)}.

Since E3 has potential good additive reduction at 3, by the Lemma IV.5, we know that B3 consists of ρP0 and ρP3. By the method described in the Section II.2, we compute B2, B3 and list them in the following tables.

Table V.1 B3

No. σ3 τ3

ρP0 P0 P6 ρP3 P3 P0

Table V.2 B2

No. σ2 δ−1 δ3 2.1 P3 P0 P0

2.2 O P0 P3

2.3 P3 P6 P6

Now, we use the Proposition II.2 to compute the values of the pairings and list it in

be the solutions to the system of linear equations X

ϕ∈AT0

< ϕv, ψv >Qv ·xϕ = 0, ψv ∈ Bv, v ∈ T0.

A basis for V and the corresponding element in Sel2(E3/Q) are listed in V.4 and V.5 Table V.4

Now, we need to determine the local symbols for the two generators listed above.

Since E3 has potential good additive reduction at 3, by the tables above we get the table below.

Table V.6

No. ∞ 3

1 1 P0

2 1 P3

Thus we get that the local symbols map s(3) is injective but not surjective. Therefore the Proposition V.1 holds for (a, b) = (1, 2) when we take q = 3.

The computation above implies that a prime p congruent 3 modulo 8 is not a congruent number, and a prime p congruent to 5 modulo 8 is a congruent number by the Propositon IV.1. Since Sel2(E3/Q) contains the zero map only and T00 consists of ∞, we may put that Θ0 consists of ∞. Thus the condition (46) holds for d = 3p with 3p congruent to 1 modulo 8. Thus the 3p-twist of E3 has rank 0 for the primes

3. THE EXAMPLES OF THE CONGRUENT CURVE 45

p satisfying p congruent to 3 modulo 8. Since the 3p-twist of E3 is equivalent to the p-twist of E, the primes congruent to 3 modulo 8 are not congruent numbers. We take Θ1 to be the empty set. Then the condition (47) holds for d = −3p with (−3p) congruent to 1 modulo 8. Since the (−3p)-twist of E3 is equivalent to the p-twist of E, the primes congruent to 5 modulo 8 are congruent numbers.

3.2. Consider that (a, b) = (2, 4). Then the elliptic curve is given by E2 : y2 = x(x − 2)(x − 4),

which is the 2-twist of the congruent curve E. Let q = 5 and denote the 5-twist by E10. Then

T0(5) = {∞, 2, 5} and E10[2] = {O, P0, P10, P20},

where Pi = (i, 0) for i = 0, 10 and 20. Let σv, τv, δ−1 and δ3 be defined as in the section IV.7. As in Section II.4, we choose that

AT0 := {ρ(pv,P0), ρ(pv,P10) | v ∈ T0(5)}.

Similar to the above case, B5 consists of ρP0 and ρP10. By the method described in the Section II.2, we compute B2, B5 and list them in the following tables.

Table V.7 B5

Now, we use the Proposition II.2 to compute the values of the pairings and list it in the following table.

be the solutions to the system of linear equations X

ϕ∈AT0

< ϕv, ψv >Qv ·xϕ = 0, ψv ∈ Bv, v ∈ T0.

A basis for V and the corresponding element in Sel2(E10/Q) are listed in the follow-ing.

Table V.10

No. 1 2

ρ(−1,P0) 1 1 ρ(−1,P10) 0 0 ρ(2,P0) 1 0 ρ(2,P

10) 1 1

ρ(5,P0) 1 0 ρ(5,P

10) 0 1

Table V.11

No. 1 2

σ P0 P0 σ2 P20 P10 σ5 P0 P10

Now, we need to determine the local symbols for the two generators listed above.

Since E10 has potential good additive reduction at 5, by the tables above we get the table below.

Table V.12

No. ∞ 5

1 1 P0

2 1 P10

Thus we get that the local symbols map s(5) is injective but not surjective. Therefore the Proposition V.1 holds for (a, b) = (2, 4) when we take q = 5.

The computation above implies that a number 2p with the prime p congruent 5 modulo 8 is not a congruent number, and a number 2p with the prime p congruent to 3 modulo 8 is a congruent number by the Propositon IV.1. Since Sel2(E10/Q) contains the zero map only and T00 consists of ∞, we may put that Θ0 consists of

∞. Thus the condition (46) holds for d = 5p with 5p congruent to 1 modulo 8. Thus the 5p-twist of E10 has rank 0 for the primes p satisfying p congruent to 5 modulo 8. Since the 5p-twist of E10 is equivalent to the 2p-twist of E, a number 2p with the prime p congruent to 5 modulo 8 is not a congruent number. We take Θ1 to be the empty set. Then the condition (47) holds for d = −5p with (−5p) congruent to 1 modulo 8. Since the (−5p)-twist of E10 is equivalent to the 2p-twist of E, a number 2p with the prime p congruent to 3 modulo 8 is a congruent number.

Bibliography

[Bir69] B. J. Birch, Diophantine analysis and modular functions, in Algebraic Geometry (Internat.

Colloq., Tata Inst. Fund. Res., Bombay, 1968), 35-42, Oxford Univ. Press, London, 1969.

[Bir70] -, Elliptic curves and modular functions, In Symposia Mathematica, Vol. IV (INDAM, Rome, 1968/69), 27-32 Academic Press, London, 1970.

[CaE56] H. Cartan, S. Eilenberg Homological Algebra, Princenton University Press, Princeton, New Jersey, 1956.

[CLT05] W.-C. Chi, K.F. Lai and K.-S. Tan, Integral points on elliptic curves, Pac. J. Math. 222, No. 2 (2005), 237-252

[CKRS] J.B. Conrey, J.P. Keating, M.O. Rubinstein, N.C. Snaith On the frequency of vanishing of quadratic twists of modular L-functions, Number theory for the millennium, I (Urbana, IL, 2000), 301V315, A K Peters, Natick, MA, 2002.

[Elk94] N. Elkies, Heegner point computations, Algorithmic Number Theory (ANTS-1), Lect. Notes in Comp. Sci. 877, Springer-Verlag, Berlin, 1994, 122-133.

[ElkWeb] -, http://www.math.harvard.edu/ elkies/compnt.html

[Gol79] D. Goldfeld, Conjectures on elliptic curves over quadratic fields, Lecture Notes in Math.

751, Springer-Verlag, 1979, 108-119.

[GouM91] F. Gouvˆea, B. Mazur, The square-free sieve and the rank of elliptic curves J. Amer.

Math. Soc. 4, 1991, no.1, 1-23.

[Hea94] D.R. Heath-Brown, The size of Selmer groups for the congruent number problem. II Invent.

Math. 118, 1994, no.2, 331-370.

[Hee52] K. Heegner, Diophantische Analysis und Modulfunktionen, Math. Z. 56,1952, 227-253.

[Mil86] J.S. Milne, Arithmetic Duality Theorems, Academic Press, New York, 1986. , 1952, 227-253.

[M-M94 ] Mai, L. and Murty, R., A note on quadratic twists of an elliptic curve, CRM Proceedings and Lecture Notes, 4, 1994, 121-124.

[Mon90] Paul Monsky. Mock heegner points and congruent numbers Mathematische Zeitschrift, 1990, 204:45-67.

[Kob84] Neal Koblitz, Introduction to elliptic curves and modular forms, Graudate Texts in Math.

97, Springer, New York, 1984.

[Ono97] K. Ono, Twists of Elliptic Curves, Compositio Mathematica 106, 1997, 349-360.

[OS98] K. Ono, C. Skinner, Non-vanishing of quadratic twists of modular L-functions Invent. Math.

134, 1998, 651-660.

[Rei75] Reiner, I. Maximal Orders, Academic Press, New York, 1975.

[RubS01] K. Rubin, A. Silverberg, Rank frequencies for quadratic twists of elliptic curves Exper.

Math. 10, 2001, no.4, 559-569.

[RubS02] -, Ranks of elliptic curves Bull. Amer. Math Soc. 39. 2002, 455-474.

[Ser79] J.-P. Serre, Local Fields, Spronger-Verlag, New York, 1979.

[Sht72] S. Shatz, Profinite groups, Arithmetic, and Geometry, Annals of Math. Studies 67, Prince-ton University Press, 1972.

[Sil86] J. H. Silverman, The Arithmetic of Elliptic Curves, Graduate Texts in Math. 106, Springer, New York, 1986.

[Sil94] -, Advanced Topics in the Arithmetic of Elliptic Curves, Graduate Texts in Math. 151, Springer, New York, 1994.

47

[SteT95] C.L. Stewart, J. Top, On ranks of twists of elliptic curves and power-free values of binary forms, J. Amer. Math. Soc. 8, 1967, no.4, 943-973.

[Tat57/58] J. Tate, WC-groups over p-adic fields, S´eminaire Bourbaki, Expos´e 156, 13pp.

[Tat62] J. Tate, Duality theorems in Galois cohomology over number fields, Proc. Intern. Congress Math. Stockholm, 234-241.

[Tat67] J. Tate, Global class field theory, in Algebraic Number Theory, J.W.S. Cassels and A.

Fr¨ohlich, eds., Acdemic Press, 1967, 162-203. Proc. Intern. Congress Math. Stockholm, 234-241.

[Tun83] J.B. Tunnell. A classical diophantine problem and modular forms of weight 3/2 Inventiones Mathematicae, 1983, 72:323-334.

在文檔中 橢圓曲線的二次扭變 (頁 39-48)

相關文件