Number Theory

L-functions and the BSD Conjecture

Цели урока

  • Understand the Euler product construction of L(E,s) from Frobenius traces
  • State weak and strong forms of the BSD conjecture
  • Know Kolyvagin's theorem: BSD for rank 0 and the Euler system method
  • Understand the strong BSD formula and the five invariants it involves

Предварительные знания

  • Modular Forms
  • Elliptic Curves
  • Modular Forms
  • Elliptic Curves

Why does knowing a single complex number L(E,1) determine whether an elliptic curve has finitely or infinitely many rational points?

  • Cryptography: rank-0 curves guarantee ECDLP hardness without infinite-order rational points
  • LMFDB: 3 million curves with BSD data, used daily by researchers worldwide
  • Algorithms: Cremona's algorithm computes analytic rank via L-function evaluation
  • Prize problem: one of 7 Millennium Problems; solution carries USD 1,000,000 reward

EDSAC 2 to a million-dollar prize

In 1958-1965, Bryan Birch and Peter Swinnerton-Dyer computed products of N_p/p for hundreds of curves on the EDSAC 2 computer in Cambridge. They noticed: curves with more rational points had products tending to 0. In 1965, they formulated the conjecture connecting the zero of L(E,s) to the rank. In 2000, the Clay Mathematics Institute named BSD one of 7 Millennium Prize Problems with a USD 1,000,000 award. In 60 years since, only rank 0 has been proved.

L-functions of Elliptic Curves

The Birch and Swinnerton-Dyer conjecture is one of 7 Millennium Prize Problems at the Clay Institute, carrying a reward of 1,000,000 USD. It links the analytic behavior of L(E,s) at s=1 to the algebraic rank of E(Q). The conjecture has been verified numerically for millions of curves, but no complete proof exists.

The root number epsilon = +/-1 predicts the parity of the analytic rank: epsilon = -1 forces the analytic rank to be odd (at least 1). This can be computed from the prime factorization of the conductor N without knowing any rational points.

What does the weak form of the BSD conjecture state?

Weak BSD: ord_{s=1} L(E,s) = rk E(Q). This is proved for rank 0 (Kolyvagin 1990) and partially for rank 1 (Gross-Zagier + Kolyvagin). For rank >= 2, BSD is completely open.

Strong BSD and the Leading Coefficient Formula

Strong BSD goes beyond predicting the order of vanishing. It gives an exact formula for the leading Taylor coefficient of L(E,s) at s=1, involving five independent arithmetic invariants of the curve.

Finiteness of Sha is known only for specific curves (rank 0 via Kolyvagin, rank 1 via Gross-Zagier + Kolyvagin). In general, it is an open problem - a consequence of BSD that has not been proved independently.

InvariantSymbolStatus
PeriodOmegaComputable from the curve equation
RegulatorRComputable from generators of E(Q)
Tamagawa numbersc_pComputable from local reduction data
Torsion order|tors|^2Computable (Mazur's theorem)
Sha order|Sha|Conjectured finite; not always computable

Which invariant in the strong BSD formula is conjecturally finite but not known to be so in general?

Sha = Shafarevich-Tate group: BSD predicts |Sha| is finite. This is proved for rank 0 (Kolyvagin 1990) and rank 1 (Gross-Zagier + Kolyvagin), but is an open problem for rank >= 2.

Partial Results and Evidence for BSD

The BSD conjecture has been verified for millions of curves numerically, and partial theorems exist for rank 0 and rank 1. The key tools are Kolyvagin's Euler systems and the Gross-Zagier theorem connecting Heegner points to L-function derivatives.

RankBSD statusKey result
0Proved (given modularity)Kolyvagin 1990
1Proved when L'(E,1) is nonzeroGross-Zagier + Kolyvagin
2OpenNumerical evidence only
3+OpenConjectural only

For rank >= 2, BSD is completely open. Not a single curve of rank 2 has had BSD proved for it in full generality. The structure of Sha, the regulator, and their product remain mysterious.

What did Kolyvagin prove in 1990 regarding BSD?

Kolyvagin 1990: L(E,1) nonzero implies rk E(Q) = 0 and |Sha| < infinity. Method: Euler systems built from Heegner points. This is BSD for rank 0, conditional on modularity.

Numerical Verification and the LMFDB

The L-functions and Modular Forms Database (LMFDB) contains BSD data for over 3 million elliptic curves over Q. For each curve, the analytic rank, Sha order, period, regulator, and Tamagawa numbers are tabulated - all consistent with BSD to machine precision.

Birch and Swinnerton-Dyer ran their computations on the EDSAC 2 computer in Cambridge in the early 1960s. They computed the partial products for hundreds of curves and noticed the pattern: curves with more rational points had products tending to 0. This empirical observation led to the conjecture.

What does the product prod(N_p/p) over all good primes estimate for an elliptic curve?

The partial product prod_p N_p/p approximates L(E,1) * C. When L(E,1) is nonzero (rank 0), it converges to a positive number. When L(E,1) = 0 (rank >= 1), it tends to 0. This was the original experimental evidence for BSD.

Connections to other topics

BSD sits at the intersection of analytic number theory, algebraic geometry, and representation theory.

  • Etale cohomology — Related topic
  • Iwasawa theory — Related topic
  • Algebraic K-theory — Related topic

Итоги

  • L(E,s) = Euler product with factors (1 - a_p p^{-s} + p^{1-2s})^{-1} for good primes p
  • Weak BSD: ord_{s=1} L(E,s) = rk E(Q)
  • Strong BSD: leading coefficient = (|Sha| * Omega * R * prod c_p) / |tors|^2
  • Kolyvagin 1990: L(E,1) nonzero implies rank = 0 and Sha finite - BSD proved for rank 0
  • For rank >= 2, BSD is completely open; only numerical evidence exists

Вопросы для размышления

  • Why does the root number epsilon = -1 force the analytic rank to be at least 1, even without computing any rational points?
  • What prevents Kolyvagin's Euler system method from working for curves of rank >= 2?
  • In what sense is the BSD conjecture a special case of the general Beilinson-Bloch conjecture about special values of L-functions?

Связанные уроки

  • nt-22 — Modular forms provide the analytic framework for elliptic curve L-functions
  • nt-24-class-field-theory — Class field theory explains the L-function structure via Galois groups
  • nt-29 — Full BSD arithmetic including descent and Heegner points
L-functions and the BSD Conjecture

0

1

Sign In