MORDELL’S THEOREM: THE ‘IRREDUCIBLE’ CASE. IAN KIMING The following exposition is based mainly on [1], pp. 66–70. 1. Let E : y 2 = f (x) be an elliptic curve over Q, f (x) = x3 +. . .. To finish the proof of Mordell’s theorem that E(Q) is a finitely generated abelian group it is enough to prove the ‘Weak Mordell’ theorem: Theorem 1. The abelian group E(Q)/2E(Q) is finite. We have already proved this theorem in case that E has a rational point of order 2, i.e., in case the polynomial f has a rational root. So we assume now that this is not the case, i.e., that f is irreducible. Via a simple linear change of variables we can and will assume that f has form x3 + bx + c with b, c ∈ Z, i.e., that E is given by a Weierstraß equation of form: E : y 2 = f (x) := x3 + bx + c , b, c ∈ Z , with f (x) irreducible. Denote by θ = θ1 , θ2 , θ3 the roots of f . Consider the field extension K/Q obtained by adjoining the root θ of f (x) to Q: K := Q(θ) ∼ = Q[x]/(f (x)) . So ‘physically’ we have K = fuθ2 + vθ + w j u, v, w ∈ Qg, — in particular a vector space of dimension 3 over Q. Definition 1. Let α ∈ K. The map ξ 7→ αξ is a linear map of the Q-vector space K into itself. We define the norm of α, denoted by N (α), as the determinant of this linear map. So we have that N (α) is a rational number which is non-zero if α is, and that N (αβ) = N (α)N (β) for α, β ∈ K. Lemma 1. If q ∈ Q then N (q − θ) = f (q). Proof. Consider 1, θ, θ2 as a Q-basis for K. As (q − θ) · 1 = q − θ , (q − θ) · θ = qθ − θ2 , (q − θ) · θ2 = c + bθ + qθ2 1 2 IAN KIMING (where we used that −θ3 = bθ + c) we see that the determinant of the linear map ξ 7→ (q − θ) · ξ is: q −1 0 det 0 q −1 = q 3 + bq + c = f (q) . c b q We will now define an abelian group A and a map µ : E(Q) → A that will turn out to be a homomorphism. The abelian group A is the following subgroup of K × /(K × )2 : A := fα ∈ K × /(K × )2 j N (α) ∈ (Q× )2 g ; that A is a subgroup is a consequence of the property N (αβ) = N (α)N (β) of the norm map. We define a map µ : E(Q) → A as follows: µ(O) = 1 mod (K × )2 , and for (x0 , y0 ) ∈ E(Q): µ(x0 , y0 ) := (x0 − θ) mod (K × )2 . Notice that µ(x0 , y0 ) actually lands in A if (x0 , y0 ) ∈ E(Q): For according to Lemma 1 we have: N (x0 − θ) = f (x0 ) = y02 , so (x0 − θ) mod (K × )2 is really in A. Theorem 1 will follow when we prove below that µ is a homomorphism with kernel 2E(Q), and that µ has finite image. Note: For the proof of these assertions it is not really necessary to know explicitly that µ has its values in the subgroup A of K × /(K × )2 . I.e., we could work with µ simply as a homomorphism to K × /(K × )2 . However, this property of µ will appear implicitly in the arguments via the relation (∗) below. Also, if one wants to work concretely with some examples, it is of value to know that µ lands in the smaller subgroup A. Lemma 2. µ is a homomorphism. Proof. We have µ(O) = 1. Let P ∈ E(Q). From the definition of µ we clearly have µ(P ) = µ(−P ). Notice also that η 2 = 1 for any element η ∈ A; so µ(−P ) = µ(P ) = µ(P )−1 . Thus, to show that µ is a homomorphism it suffices to show: P1 + P2 + P3 = O ⇒ µ(P1 )µ(P2 )µ(P3 ) = 1 for points P1 , P2 , P3 ∈ E(Q). So let P1 , P2 , P3 ∈ E(Q) with P1 + P2 + P3 = O. If at least 2 of the Pi are O then so is the 3’rd, and so trivially µ(P1 )µ(P2 )µ(P3 ) = 1. If exactly 1 of the Pi , say P3 , is O, then P2 = −P1 , and again the claim trivially follows. So we may assume Pi 6= O. Write then Pi = (xi , yi ). Since P1 + P2 + P3 = O the points Pi are precisely the points of intersection between E and some rational line MORDELL’S THEOREM: THE ‘IRREDUCIBLE’ CASE. 3 `, counting multiplicities. For the line ` we have an equation αx + βy + γz = 0, and in this equation we must have β 6= 0: For otherwise the line would apparently pass through the point O = (0, 1, 0). But then the point O would have to be one of the points Pi , contrary to assumption. For the line ` we may thus an equation of form: ` : y = λx + ν . Then we have that x1 , x2 , x3 are the roots of the polynomial f (x) − (λx + ν)2 , i.e., f (x) − (λx + ν)2 = (x − x1 )(x − x2 )(x − x3 ) . Then: (x1 − θ)(x2 − θ)(x3 − θ) = (λθ + ν)2 − f (θ) = (λθ + ν)2 ∈ (K × )2 , because f (θ) = 0. Thus, µ(P1 )µ(P2 )µ(P3 ) = 1. Lemma 3. The kernel of µ is 2E(Q). Proof. Since η 2 = 1 for any η ∈ A (A ‘has exponent 2’) it is clear that 2E(Q) is contained in the kernel of µ. Suppose that P = (x0 , y0 ) ∈ E(Q) is in the kernel of µ, i.e., that x0 − θ ∈ (K × )2 . There are then u, v, w ∈ Q such that x0 − θ = (uθ2 + vθ + w)2 . Here we must have u 6= 0 since this equation would otherwise imply either that θ were rational or root of a polynomial of degree 2 with rational coefficients (which is not possible since θ is root of f , and f is irreducible of degree 3). If we put: v2 vw v , s := + ub − w , t := + uc , u u u then r, s, t ∈ Q, and we check that r := (r − θ)(uθ2 + vθ + w) = sθ + t (use that −θ3 = bθ + c). Then: (sθ + t)2 = (r − θ)2 (uθ2 + vθ + w)2 = (r − θ)2 (x0 − θ) . So if we consider the polynomial g(x) := (sx + t)2 − (r − x)2 (x0 − x) we have g(θ) = 0. Since g is a monic polynomial of degree 3 with rational coefficients it follows then that g(x) = f (x) . Let us now look at the intersection between E and the line y = sx + t. The x-coordinates of the 3 intersection points are the roots of the polynomial f (x) − (sx + t)2 = g(x) − (sx + t)2 = −(r − x)2 (x0 − x) , 4 IAN KIMING i.e., there is 1 point with x-coordinate x0 , and 2 points with x-coordinate r. The point with x-coordinate x0 must be (x0 , ±y0 ) = ±P . The 2 points with x-coordinate r must be equal, since the y-coordinate is uniquely determined by the x-coordinate (because the points are on the line y = sx + t). So we have ±P + 2 · Q = O for some point Q ∈ E(Q). Thus P ∈ 2E(Q), as desired. We can now begin the proof of Theorem 1. Since we now know that µ : E(Q) → A is a homomorphism with kernel 2E(Q), Theorem 1 follows if we can show that the image of µ is finite. Showing this is the real core of the matter. Let P = (x0 , y0 ) ∈ E(Q). We know that we can write n m x0 = 2 , y0 = 3 e e with integers m, n, e, where e ∈ N, gcd(m, e) = gcd(n, e) = 1. Since (x0 , y0 ) is on E we have then n2 /e6 = f (m/e2 ) = (m/e2 − θ1 )(m/e2 − θ2 )(m/e2 − θ3 ) , i.e., n2 = (m − θ1 e2 )(m − θ2 e2 )(m − θ3 e2 ) , (∗) and we have µ(x0 , y0 ) = (m − θ1 e2 ) mod (K × )2 . So one way or the other we will have to use the information contained in (∗) to deduce that there are only a finite number of possibilities for the quantity (m−θ1 e2 ) mod (K × )2 . We will use algebraic number theory in the algebraic number field L := Q(θ1 , θ2 , θ3 ) to do that. 2. Interlude: Some results from algebraic number theory. An algebraic number field is a field obtained by adjoining to Q a finite number of roots of polynomials with rational coefficients. So the fields K and L of the previous section are algebraic number fields. Our application of the review below will be exclusively to the field L, but they hold generally. So in this section we let L be an arbitrary algebraic number field. We will not prove any of the statements in this section. Definition 2. The set of algebraic integers in L, denoted by OL , is the set of elements of L that are roots of some monic polynomial with integral coefficients (the polynomial depending of course on the element). One can show that OL is a ring with respect to addition and multiplication in the field L. Example: L = Q, OL = Z. Another example: L = Q(i), OL = Z[i]. For every number λ ∈ L× there is an integer z 6= 0 such that z · λ ∈ OL . In particular, L is then the field of fractions of OL . MORDELL’S THEOREM: THE ‘IRREDUCIBLE’ CASE. 5 Since OL is a ring we can consider ideals in it. One can show that non-zero ideals can be written uniquely as (finite) products of prime ideals (OL is a so-called Dedekind ring). One obtains from this the possibility of computing gcd (greatest common divisor) of ideals: If Y a Y b a= pi i , b= pi i i i with the pi prime ideals, then (]) gcd(a, b) := a + b = Y minfai ,bi g pi . i In particular on has ([) b ⊆ a ⇒ (bi ≥ ai , ∀i) . If λ ∈ L× we can consider the set: (λ) := fν · λ j ν ∈ OL g . This is visibly a set closed under addition and under multiplication by elements of OL . So, if λ ∈ OL then obviously (λ) is an ideal of OL : The principal ideal generated by λ. If λ 6∈ OL then (λ) is not an ideal, but still there may exist an ideal a such that the set a · (λ) := fν · λ j ν ∈ ag is an ideal; in fact, there will exist such an a (why?). The so-called ‘Theorem on finiteness of class number’ implies the following statement: There exist finitely many ideals v1 , . . . vτ of OL such that: For any ideal a of OL there exist j ∈ f1, . . . , τ g and an element λ ∈ L× such that: a = vj · (λ) The invertible elements of OL (i.e. elements u ∈ OL s.t. u 6= 0 and u−1 also belongs to OL ) are called units of L (abuse of notation). Denote the set of units by UL . Then UL is obviously an abelian group with respect to multiplication. Theorem: (Dirichlet). UL is finitely generated. In fact, one can also give the rank of UL , and there is an algorithm that will produce a system of generators for this group. 6 IAN KIMING 3. End of proof of Theorem 1 Let us continue the reasoning from the end of section 1. Retain the notation: We have P = (x0 , y0 ) ∈ E(Q), and write x0 = em2 , y0 = en3 with integers m, n, e, where e ∈ N, gcd(m, e) = gcd(n, e) = 1. So we have the equation: (∗) n2 = (m − θ1 e2 )(m − θ2 e2 )(m − θ3 e2 ) , and mod (K × )2 . µ(x0 , y0 ) = (m − θ1 e2 ) We will work in the algebraic number field L := Q(θ1 , θ2 , θ3 ) . The numbers θj −θk and m−θj e2 are all in OL . We shall work with the principal ideals generated by these. Recalling the facts mentioned in section 2 we may choose the following: • A finite number of prime ideals p1 , . . . , pσ such that: For j 6= k the ideal (θj − θk ) is a power product of the pi . • Ideals v1 , . . . , vτ as in section 2 above. • Numbers δ1 , . . . , δζ ∈ L× such that: Any one of the finitely many products Y a pi i with ai ∈ f0, 1g i equals vr · (δs ) for some r, s. • Numbers η1 , . . . ηξ ∈ OL such that: If α, β ∈ f1, . . . , τ g, and if vα v2β is principal then for some γ we have vα v2β = (ηγ ) . • Units u1 , . . . uπ ∈ UL that are a system of representatives for the finite quotient UL /UL2 . Now, for each j we can write: ! 2 (m − θj e ) = Y a qi ij · a2j , i where the qi are prime ideals, aij ∈ f0, 1g, and aj is just some ideal. We may and will assume for each i that aij = 1 for at least one j (since, if this is not the case we have aij = 0 for this i and j = 1, 2, 3; but then in the above we can just leave out the ideal qi for this particular i). Now, the relation (∗): (∗) n2 = (m − θ1 e2 )(m − θ2 e2 )(m − θ3 e2 ) , MORDELL’S THEOREM: THE ‘IRREDUCIBLE’ CASE. 7 together with the fact that we have unique factorization of ideals as products of prime ideals implies: ai1 + ai2 + ai3 ≡ 0 (\) (2) , for every i. Claim: qi ∈ fp1 , . . . , pσ g for each i. Proof of the claim: Fix an i. We first notice that there are j, k ∈ f1, 2, 3g with j 6= k and such that aij = aik = 1 . For because of our assumption there is j such that aij = 1; because of (\) (and since the a’s are all 0 or 1) there is then a k 6= j such that aik = 1. Consider then the exponent of qi in the factorization into prime ideals of the ideal (m − θj e2 ) + (m − θk e2 ): According to (]) this exponent is the minimum of two (non-negative) integers congruent mod 2 to aij = 1 and aik = 1, respectively. So, the exponent is odd, and hence in particular non-zero. I.e., we have that qi occurs with a non-zero exponent in the prime factorization of the ideal (m − θj e2 ) + (m − θk e2 ) . Now we notice that this ideal contains the (non-zero) ideal (θj − θk ): For, the equality of numbers: −(m − θj e2 ) + (m − θk e2 ) = (θj − θk )e2 , shows that the ideal (m − θj e2 ) + (m − θk e2 ) contains the number (θj − θk )e2 ; but it also contains the number n2 , and hence the number (θj − θk )n2 , because of (∗); as (n, e) = 1 the ideal contains the number θj − θk , and the statement follows. Now ([) implies that qi occurs with non-zero exponent in the prime factorization of the ideal (θj −θk ); by definition however, the only prime ideals with that property are p1 , . . . , pσ . So, qi must be one of these which was the claim. By the choices we made above, we can now conclude that (m − θe2 ) = (m − θ1 e2 ) = vr · (δs ) · a21 , for some r, s. We can also write: a1 = vt · (λ) for some t and some λ ∈ L× . Consequently, (m − θe2 ) = (δs λ2 ) · vr v2t . From this equality we may conclude that vr v2t is a principal ideal. So according to the choices we have made, we may write vr v2t = (ηγ ) for some γ, and accordingly (m − θe2 ) = (δs ηγ λ2 ) . 8 IAN KIMING This equality of ideals implies that δs ηγ λ2 ∈ OL , and that m − θe2 = δs ηγ λ2 · u for some unit u ∈ UL . There is some ρ such that u ≡ uρ mod UL2 . But then we have: m − θe2 ≡ δs ηγ uρ mod (L× )2 for some m ∈ f1, . . . ζg, γ ∈ f1, . . . ξg, ρ ∈ f1, . . . πg. We have now shown that there are only finitely many possibilities for µ(x0 , y0 ) mod (L× )2 = m − θe2 mod (L× )2 , where obviously µ(x0 , y0 ) mod (L× )2 means the image of µ(x0 , y0 ) under the homomorphism K × /(K × )2 −→ L× /(L× )2 induced by the injection K × ,→ L× . The proof of finiteness of µ(E(Q)), and hence of Theorem 1, may now be concluded by the following lemma (which is essentially what is classically called ‘Kummer theory’): Lemma 4. Let L/K be a finite extension of (algebraic number) fields. Then there is a 1–1 correspondence between the kernel of the natural homomorphism K × /(K × )2 −→ L× /(L× )2 coming from the injection K × ,→ L× , and the set of quadratic extensions of K contained in L. In particular, this kernel is finite. Proof. The kernel in question is seen to be (†) K × ∩ (L× )2 /(K × )2 . √ Attach to any element ξ of the group (†) the quadratic extension K( ξ); this extension is contained in L, as ξ ∈ (L× )2 . Conversely, any quadratic extension √ of K has the form K( ξ) for some ξ ∈ K × , and this extension is contained in L precisely if ξ ∈ (L× )2 . √ The claim now follows at once when we remind ourselves that we have K( ξ1 ) = √ K( ξ2 ) for 2 elements ξ1 , ξ2 ∈ K × , precisely if ξ1 and ξ2 differ by a square in K ×. References [1] J. W. S. Cassels: ‘Lectures on elliptic curves.’ London Mathematical Society Student Texts 24, Cambridge University Press 1991. Department of Mathematics, University of Copenhagen, Universitetsparken 5, DK2100 Copenhagen Ø, Denmark. E-mail address: [email protected]
© Copyright 2026 Paperzz