For points on a smooth plane cubic, draw the line through them, using the tangent at when , and let its third intersection with the cubic be . The chord-and-tangent group law defines as the reflection of in the -axis; the point at infinity is the identity and .
If , their chord or tangent has rational coefficients. Substitution into the cubic gives a polynomial with rational coefficients for which two intersection roots are rational, so the third is rational as well. The identity and inverse of every rational point are rational, and closure under addition follows. Assuming the elliptic-curve law is a group law, is therefore a subgroup of .
For , the addition formulas use
with for doubling. They give
and
Let be rational. If the 3-adic valuation satisfies , first cannot be negative: otherwise is the unique term of least valuation in , giving . If , the right side has valuation zero. If , its three terms have valuations at least , at least , and exactly , so . In every case .
The Lutz–Nagell theorem says that if
has nonzero discriminant and is a torsion point, then and either or
For integrality, fix a prime . If a rational point has nonintegral coordinates, its primitive projective coordinates reduce to , so it belongs to the kernel of reduction. The parameter identifies this kernel with the formal group of an elliptic curve over . The formal logarithm, with the standard separate first-step argument at , shows that this group has no nonzero rational torsion. A rational torsion point therefore has nonnegative -adic valuations in both coordinates for every prime , hence integral coordinates.
Suppose now that . The point is again a nonzero torsion point and hence integral. Its -coordinate is , where
Thus is an integer. A rational number whose square is integral is integral, so and in particular . The curve equation and the identity
then prove . This is the divisibility proof in the Nagell–Lutz theorem.
Here
By the Lutz–Nagell theorem, a nonzero rational torsion point has integral coordinates and either or . Part b gives , so . The cubic has no integral zero. Substitution of gives exactly
This proves the required inclusion.
Since , the point has order three. The point is not torsion because is nonintegral, contradicting Nagell–Lutz. If or were torsion, adding the torsion point would make torsion; their negatives are excluded in the same way. Hence
The degree of an isogeny is the degree of the induced finite extension of function fields. Saying that degree is a quadratic form on means
and that
is bilinear, equivalently
The trace of an elliptic-curve endomorphism is
The relation implies
the trace of the square of an elliptic-curve endomorphism.
The Hasse theorem for elliptic curves states that for ,
Let be the Frobenius isogeny of an elliptic curve, put , and note that and
For integers , quadraticity gives
This binary quadratic form cannot have positive discriminant, since rational numbers are dense, so . Substitution proves the bound. This is the degree-form proof of the Hasse bound.
Both endpoints occur. The curve over is supersingular with trace zero. Over , its Frobenius is , so
Its nontrivial quadratic twist over has the opposite trace and therefore has
points.
For a short Weierstrass equation in characteristic different from two, the invariant differential on an elliptic curve
is nonzero and regular, including at .
Write . Since , it is an even rational function and therefore belongs to ; write with coprime polynomials . The pullback of a nonzero invariant differential is invariant, so for some ,
Since
we obtain
Thus
Take . Because and have degrees , their leading terms dominate, giving
The polynomial has degree with nonzero leading coefficient, the same degree as . Their quotient consequently has valuation zero at , so
The isogeny sends to , hence .
Use the alternative affine coordinates
so that and . The equation becomes
Recursive coefficient comparison gives a unique series . Since and hence , , this gives the formal coordinates on a short Weierstrass curve identification
A one-dimensional commutative formal group law over a ring is a power series satisfying , commutativity, and associativity. A morphism is a series satisfying
Express the isogeny of part b in the formal coordinates of part d and set
Part c shows that points approaching map to points approaching , so . Since is a group homomorphism, applying the parameter to gives
Thus is the formal-group morphism induced by an isogeny .
Let . If becomes an th power in , choose with . Then
is a cocycle with values in the finite group . Changing changes it by a coboundary, and its class is trivial exactly when was already an th power in . Thus the kernel injects into the finite group and is finite. This is the multiplicative case of the finite-extension kernel of a Kummer map.
Assume and put . All roots of are with , so is its splitting field and is a Finite Galois extension. The map
is an injective group homomorphism.
For a finite Galois extension , the kernel of
is finite. Indeed, if for , then is a cocycle in the finite -module , and the resulting map from the kernel to is injective.
The analogue of part b assumes . Given and with , every conjugate of is for some . Hence is Galois and
is an injective homomorphism. These are the two elliptic forms of the Kummer pairing.
Pass to the finite extension . Part c shows that the kernel of is finite, so it is enough to control the image after -torsion becomes rational. The Kummer map of an elliptic curve
associates to the finite extension generated by one -division point of .
Let contain the places over , all archimedean places, and all places of bad reduction. The local theory of elliptic curves shows that these Kummer classes are unramified outside . Because is finite and constant over , such classes are controlled by finitely many -unramified power classes. Finiteness of the class group and finite generation of the unit group make that power-class group finite. The Kummer image is therefore finite, proving the Weak Mordell-Weil theorem that is finite; this is the Kummer-theoretic proof of the weak Mordell-Weil theorem.
Let a rational right triangle have legs , hypotenuse , and area . After interchanging the legs if necessary, set
Using and gives
so lies on the congruent number elliptic curve . The triangle is nondegenerate, so and .
Conversely the three displayed lengths in the question satisfy
and their area is
Thus every such triangle is .
The change of variables , identifies with
A full two-isogeny descent, equivalently the standard full 2-descent for a cubic with three rational roots, gives
The local conditions at , , and infinity leave exactly these eight square-class combinations; primes outside have even valuations and contribute none. Hence the quotient has dimension three. Since the rational 2-torsion has dimension two, the rank is one.
The Lutz–Nagell theorem on the integral model, or reduction at two good primes, excludes odd torsion and torsion of order greater than two. Therefore
and generates the free part. This is the Mordell-Weil group of the congruent number curve for five.
For in lowest terms, take the logarithmic naive height . Its required properties are
and
with constants depending only on the curve. Define the canonical height of an elliptic curve by
The first bounded-error relation makes this a convergent telescoping correction to . Apply the second relation to , divide by , and let to obtain
Also , and the parallelogram identity then gives for every integer . Thus is a quadratic form.
By part b, write
The canonical height of an elliptic curve vanishes on torsion and satisfies . If , their nonzero integer coefficients therefore have equal squares, so
for some .
Negation leaves the -coordinate unchanged. Addition by the four 2-torsion points changes among
Substitution in the three side formulas for , using , only changes signs and permutes the three values. Hence and are the same right triangle up to reordering their sides.

Articles by others on the same topic (0)

There are currently no matching articles.