A. We prove that a very general elliptic surface E → P 1 over the complex numbers with a section and with geometric genus p g ≥ 2 contains no rational curves other than the section and components of singular fibers. Equivalently, if E/C(t) is a very general elliptic curve of height d ≥ 3 and if L is a finite extension of C(t) with L ∼ = C(u), then the Mordell-Weil group E(L) = 0.
IFix a field k and consider elliptic curves E over K = k(t). When k is a finite field, we showed in [Ulm07] that there are often finite extensions L of K which are themselves rational function fields (i.e., L ∼ = k(u)) such that the rank of E(L) is as large as desired. Indeed, under a mild parity hypothesis on the conductor of E (which should hold roughly speaking in one half of all cases), we may take extensions of the form L = k(t 1/d ) with d varying through integers prime to p. More generally, for any elliptic curve E over K with j(E) ∈ k there is a finite extension of K of the form k ′ (u) with k ′ a finite extension of k such that E obtains unbounded rank in the layers of the tower k ′ (u 1/d ). Our aim in this note is to show that the situation is completely different when k is the field of complex numbers.From now on we take k = C. If E is an elliptic curve over C(t), the height of E is the smallest non-negative integer d such that E has a Weierstrass equationwhere a(t) and b(t) are polynomials of degree ≤ 4d and ≤ 6d respectively. Our results concern elliptic curves of height d ≥ 3.1.1. Theorem. A very general elliptic curve E over C(t) of height d ≥ 3 has the following property: For every finite rational extension L ∼ = C(u) of C(t), the Mordell-Weil group E(L) = 0.Here "very general" means that in the relevant moduli space, the statement holds on the complement of a countable union of proper closed subsets. See Subsection 3.4 below for more details.The theorem shows in particular that there is no hope of producing elliptic curves of large rank over C(u) by starting with a general curve over C(t) and iteratively making rational field extensions. What can be done with special elliptic curves remains a very interesting open question about which we make a few speculations in the last section. Now we connect the theorem with the title of the paper. Attached to E/C(t) is a unique elliptic surface π : E → P 1 with the properties that E is smooth over C, and π is proper and relatively minimal with generic fiber isomorphic to E/C(t). Conversely, given a relatively minimal elliptic surface π : E → P 1 , its generic fiber is an elliptic curve over C(t). The height of E is then equal