Back to the theory contents

Elliptic geometry of square classes · 4.1

Genus-One Curves, Jacobians, and Elliptic Surfaces

This chapter fixes the common language used later: how a genus-one quartic produces an elliptic curve, what changes when a fiber is replaced by a surface, and exactly which ranks are bounded by the Shioda–Tate formula.

1. Genus-one curves and elliptic curves

A smooth projective curve of genus one is not yet an elliptic curve: one must choose a rational point O to serve as the identity of the group law. With such a point, the curve is isomorphic to its Jacobian. Without a rational point, the Jacobian still exists, but the original curve is a torsor under it and may have no rational points.

What is done with a quartic

In the 6/9 articles, parametrizing two quadratic conditions usually leaves an equation v²=Q₄(u). When the discriminant is nonzero, its smooth projective completion has genus one. A chosen rational point permits an exact birational transformation to a Weierstrass model

y2=x3+Ax2+Bx+C. y^2=x^3+A x^2+B x+C.

Such a transformation describes an open chart: omitted points and degenerate parameter values must be checked separately. A parametrization on one chart therefore does not automatically classify every rational point of the original surface.

2. Fibers, sections, and the total surface

An elliptic surface is a surface 𝓔 equipped with a morphism π:𝓔→C to a base curve C whose generic fiber is a smooth genus-one curve, together with a zero section. In our models C is usually ℙ¹, with coordinate p, q, t, or k.

π:EC,Eη/Q(C). \pi:\mathcal E\longrightarrow C,\qquad E_\eta/\mathbb Q(C).

A rational section C→𝓔 is the same as a point of the generic fiber Eη over the function field ℚ(C). Substituting a particular rational parameter gives a specialized fiber, an individual elliptic curve over ℚ. The rank of the generic fiber and the rank of a specialization are different quantities.

NotationMeaning
rankE(Q(t))\operatorname{rank}E(\mathbb Q(t))arithmetic rank of rational sections
rankE(Q(t))\operatorname{rank}E(\overline{\mathbb Q}(t))geometric rank after extending the constant field
rankEt0(Q)\operatorname{rank}E_{t_0}(\mathbb Q)arithmetic rank of one nonsingular fiber

One always has rank E(ℚ(t))≤rank E(ℚ̄(t)). The specialization theorem preserves the independence of specified sections for all rational parameters outside a thin exceptional set, but it does not assert that every fiber has the same total rank.

3. Weierstrass models and singular fibers

A Weierstrass equation over ℚ(t) first defines the generic fiber. To study the surface, one extends it over the whole base, resolves singularities, and passes to a relatively minimal model. Zeros of the discriminant mark possible singular fibers; their exact type is determined from a minimal local model, not by the discriminant multiplicity alone.

Kodaira typeRoot latticeEuler number
InAn−1n
I0*D46
IVA24
IV*E68

In these articles, “split Jacobian” means that all three nonzero 2-torsion points are visible over the stated function field, for example in a model y²=x(x−a)(x−b). It does not mean that the surface itself is a product of curves.

4. Rational elliptic surfaces and K3 surfaces

For a relatively minimal elliptic surface with a section over ℙ¹ and without multiple fibers, the fundamental line bundle has degree χ(𝒪𝓔), the Euler numbers of the singular fibers sum to 12χ(𝒪𝓔), and the canonical-bundle formula is

KEπ ⁣(KP1L),degL=χ(OE). K_{\mathcal E}\cong \pi^*\!\left(K_{\mathbb P^1}\otimes\mathcal L\right), \qquad \deg\mathcal L=\chi(\mathcal O_{\mathcal E}).

When χ=1, one obtains a rational elliptic surface and Euler sum 12. When χ=2, the canonical bundle is trivial, the sum is 24, and under the standard smoothness and relative-minimality hypotheses the surface is an elliptic K3 surface. The numerical sum 24 is used only after these hypotheses have been checked.

There is another route to a K3 surface: a smooth intersection of three quadrics in ℙ⁵ has trivial canonical bundle by adjunction, while H¹(𝒪)=0 follows from the Lefschetz theorem or the Koszul complex. Its smooth minimal model is therefore K3. The Jacobian of a chosen fibration is additional structure on that surface.

5. The Shioda–Tate formula

Let ρ(𝓔) be the geometric Picard number and let Rᵥ be the root lattice formed by components of a singular fiber that do not meet the zero section. Then

ρ(E)=2+vrankRv+rankE(Q(t)). \rho(\mathcal E)= 2+\sum_v\operatorname{rank}R_v+ \operatorname{rank}E(\overline{\mathbb Q}(t)).

A rational elliptic surface has ρ=10. A complex K3 surface has ρ≤20. Hence the singular-fiber passport gives an upper bound on the geometric rank. A lower bound comes from explicit independent non-torsion sections. Matching bounds determine the exact geometric rank; otherwise an article records only the proved interval.

6. What an explicit section proves

The coordinate formulas of a section must satisfy the equation identically in the function field. A few numerical checks do not prove non-torsion: one uses an exact specialization with certified non-torsion, a height computation, descent, or another certificate. Independence of several sections requires a separate exact argument.

A non-torsion section gives infinitely many points on the generic fiber and, after excluding degenerate parameters, infinitely many rational specializations. It does not prove that the chosen chart covers every rational point of the surface or that every specialization is positive and has exactly the required 6/9 type.

7. Two uses of the name Legendre

The Legendre symbol (a/p) in the chapter on residues records whether a is a quadratic residue modulo an odd prime p. Legendre normal form is the family of elliptic curves

Eλ:y2=x(x1)(xλ),λ0,1. E_\lambda:\qquad y^2=x(x-1)(x-\lambda), \qquad \lambda\ne0,1.

These are different notions that merely share a name. The ABEFGH article uses the normal form of a family of curves, not the quadratic-residue symbol.

8. References