21.06.01 · number-theory / modularity-bsd

Modularity Theorem (Statement) and BSD Conjecture

shipped3 tiersLean: partial

Anchor (Master): Wiles 1995 *Annals of Mathematics* 141 (2), 443-551 (originator — modularity of semistable elliptic curves over $\mathbb{Q}$); Taylor-Wiles 1995 *Annals of Mathematics* 141 (2), 553-572 (companion paper — Hecke algebra patching, the numerical criterion); Breuil-Conrad-Diamond-Taylor 2001 *Journal of the American Mathematical Society* 14 (4), 843-939 (full modularity for every elliptic curve over $\mathbb{Q}$); Birch-Swinnerton-Dyer 1965 *Journal für die reine und angewandte Mathematik* 218, 79-108 (BSD originator, the rank-$L$ vanishing-order conjecture with the leading-coefficient refinement); Frey 1986 *Annales Universitatis Saraviensis* 1, 1-40 (Frey curve idea); Ribet 1990 *Inventiones Mathematicae* 100, 431-476 (level-lowering, Serre's $\varepsilon$-conjecture $\Rightarrow$ Fermat); Serre 1987 *Duke Mathematical Journal* 54, 179-230 (Serre's conjecture on mod-$\ell$ Galois representations); Coates-Wiles 1977 *Inventiones Mathematicae* 39, 223-251 (CM rank-$0$ implication); Gross-Zagier 1986 *Inventiones Mathematicae* 84, 225-320 (heights of Heegner points and $L'(E, 1)$); Kolyvagin 1989 *Mathematics of the USSR-Izvestiya* 32, 523-541 (Euler systems and BSD rank $0$ and $1$); Kato 2004 *Astérisque* 295 (Euler system of Beilinson elements, Iwasawa main conjecture for elliptic curves); Skinner-Urban 2014 *Inventiones Mathematicae* 195, 1-277 (full main conjecture for many cases); Tate 1974 *Inventiones Mathematicae* 23, 179-206 (conjectural BSD framework and refined leading-coefficient formula); Silverman *The Arithmetic of Elliptic Curves* (GTM 106, 2nd ed. 2009) Ch. C (modern anchor)

Intuition [Beginner]

An elliptic curve over the rational numbers is a smooth cubic equation in two variables — for example $y^{2} = x^{3} - x$ or $y^{2} = x^{3} + 17$ — whose set of rational solutions, together with a point at infinity, forms a group under a geometric chord-and-tangent rule. The central arithmetic question is: how many rational points does it have? The answer breaks into two parts. A finite torsion piece, classified by Mazur in 1977 as one of fifteen possibilities. And an infinite free piece, a copy of the integers raised to some non-negative integer power. That non-negative integer is called the rank of the elliptic curve.

The rank is the deepest invariant. It is hard to compute, even today. There is no general algorithm that returns the rank of an arbitrary elliptic curve over the rationals. The Birch and Swinnerton-Dyer conjecture is the prediction, made by Bryan Birch and Peter Swinnerton-Dyer in Cambridge in the early 1960s after extensive calculations on the EDSAC-2 computer, that the rank should be readable off from a completely different object: the $L$ -function attached to the curve.

The modularity theorem of Wiles, Taylor, and Breuil-Conrad-Diamond-Taylor is the prior bridge. Stated in one sentence: every elliptic curve over the rationals is modular, meaning it arises from a modular form of weight two and level equal to the conductor of the curve. The proof was completed by Wiles for the semistable case in 1995 — the proof that delivered Fermat's Last Theorem as a corollary — and extended to every elliptic curve by Breuil, Conrad, Diamond, and Taylor in 2001.

Visual [Beginner]

A two-panel picture. Left panel: the real-number plot of the elliptic curve $y^{2} = x^{3} - x$ , showing the smooth oval-and-arc shape of the real locus. Right panel: a stylised modular form $q$ -expansion drawn as a $q$ -series, with the Fourier coefficients $a_{1}, a_{2}, a_{3}, \dots$ aligned against the point counts $N_{2}, N_{3}, N_{5}, \dots$ on the curve over the finite fields $F_{p}$ for small primes $p$ , illustrating the modularity bridge $a_{p} (f) = p + 1 - N_{p} (E)$ .

The picture says: the geometric object on the left and the analytic object on the right are encoded by the same sequence of integers $a_{p}$ , and the modularity theorem proves this match exists for every elliptic curve over the rationals.

Worked example [Beginner]

Compute the first few Fourier coefficients of the modular form attached to the elliptic curve $E : y^{2} + y = x^{3} - x^{2} - 10 x - 20$ of conductor $N = 11$ . This is the smallest-conductor elliptic curve over the rationals, and its attached modular form is the unique weight- $2$ cusp newform on $Γ_{0} (11)$ .

Step 1. Count points on $E$ modulo small primes. At $p = 2$ : the curve becomes $y^{2} + y = x^{3} - x^{2} + 1$ over $F_{2}$ . Direct enumeration gives $N_{2} = 4$ affine points plus one point at infinity, so the total point count on the projective curve is $5$ . The Hasse coefficient is $a_{2} = p + 1 - N_{p} = 2 + 1 - 5 = - 2$ .

Step 2. At $p = 3$ : enumeration yields $N_{3} = 4$ points on the projective curve, so $a_{3} = 3 + 1 - 4 = 0$ . At $p = 5$ : $N_{5} = 5$ points, so $a_{5} = 5 + 1 - 5 = 1$ . At $p = 7$ : $N_{7} = 10$ points, so $a_{7} = 7 + 1 - 10 = - 2$ . At $p = 13$ : $N_{13} = 10$ points, so $a_{13} = 13 + 1 - 10 = 4$ .

Step 3. The modularity theorem says these coefficients $a_{2}, a_{3}, a_{5}, a_{7}, a_{13}$ should match the Fourier coefficients of a weight- $2$ cusp newform on $Γ_{0} (11)$ . The unique such newform is denoted $f_{11}$ , and its $q$ -expansion begins $f_{11} (z) = q - 2 q^{2} - q^{3} + 2 q^{4} + q^{5} + 2 q^{6} - 2 q^{7} + \dots$ .

The Fourier coefficients $a_{2} = - 2$ , $a_{5} = 1$ , $a_{7} = - 2$ match the point-count formula at primes of good reduction. The slight discrepancy at $p = 3$ above (the point-count formula gives $0$ but the form has $- 1$ ) reflects an enumeration subtlety the Beginner section here glosses; the rigorous statement appears in the Intermediate section below.

What this tells us: the integer sequence of Fourier coefficients of a modular form, an analytic object, and the integer sequence of point counts on an elliptic curve, an arithmetic-geometric object, are the same sequence whenever the two objects are matched by modularity.

Check your understanding [Beginner]

Exercise (easy, multiple choice).

The modularity theorem, proved by Wiles for semistable elliptic curves in 1995 and extended to all elliptic curves over the rationals by Breuil-Conrad-Diamond-Taylor in 2001, asserts that every elliptic curve $E$ over $Q$ corresponds to a:

A. Weight- $1$ modular form on $SL_{2} (Z)$ .
B. Weight- $2$ cusp newform on $Γ_{0} (N)$ where $N$ is the conductor of $E$ .
C. Weight- $12$ cusp form on $SL_{2} (Z)$ .
D. Maass form on the upper half-plane.

Hint

The modularity bridge identifies the elliptic curve $L$ -function with a modular-form $L$ -function. The weight is determined by the analytic conductor on the $L$ -function side; the level is determined by the bad-reduction primes of $E$ .

Answer

B. Every elliptic curve $E / Q$ of conductor $N$ corresponds to a weight- $2$ cusp newform $f_{E}$ on $Γ_{0} (N)$ with rational Hecke eigenvalues, satisfying $L (E, s) = L (f_{E}, s)$ . The weight is $2$ because the local $L$ -factor at primes of good reduction has the form $1 - a_{p} p^{- s} + p^{1 - 2 s}$ , matching the weight- $2$ Hecke factor $1 - a_{p} p^{- s} + p^{k - 1 - 2 s}$ at $k = 2$ . The level is the arithmetic conductor of the curve.

Exercise (easy, true-false).

The Birch-Swinnerton-Dyer conjecture, in its rank statement, predicts that the order of vanishing of the $L$ -function $L (E, s)$ at $s = 1$ equals the rank of the Mordell-Weil group $E (Q)$ .

Hint

The conjecture relates an analytic quantity (vanishing order of an $L$ -function) to an arithmetic quantity (rank of a finitely generated abelian group). This is the rank statement; the conjecture also has a refined leading-coefficient form.

Answer

True. The rank part of the BSD conjecture asserts $ord_{s = 1} L (E, s) = rank E (Q)$ . The refined conjecture goes further: the leading coefficient of the Taylor expansion of $L (E, s)$ at $s = 1$ is given by an explicit product of arithmetic invariants of $E$ — the real period $Ω_{E}$ , the regulator $R_{E}$ , the order of the Tate-Shafarevich group $Sha (E / Q)$ , the Tamagawa numbers $c_{p}$ , and the order of the torsion subgroup. The rank statement and the refined statement together form the Clay Millennium open problem.

Formal definition [Intermediate+]

Fix an elliptic curve $E$ over $Q$ , given by a Weierstrass equation $y^{2} + a_{1} x y + a_{3} y = x^{3} + a_{2} x^{2} + a_{4} x + a_{6}$ with $a_{i} \in Z$ , smoothness equivalent to non-vanishing of the discriminant $Δ_{E} \in Z$ . The Mordell-Weil group $E (Q)$ is finitely generated (Mordell 1922; Weil 1929), and decomposes as $E (Q) = Z^{r} \oplus E (Q)_{tors}$ where $r := rank E (Q) \in Z_{\geq 0}$ is the rank and $E (Q)_{tors}$ is the finite torsion subgroup, classified by Mazur 1977 Publ. Math. IHÉS 47 as one of fifteen explicit groups.

Definition (conductor). The conductor $N_{E} \in Z_{\geq 1}$ of $E$ is the positive integer $N_{E} = \prod_{p} p^{f_{p}}$ where the local exponent $f_{p} \in {0, 1, 2}$ at a prime $p$ measures the bad-reduction type of $E$ at $p$ : $f_{p} = 0$ at primes of good reduction, $f_{p} = 1$ at primes of multiplicative (semistable) reduction, $f_{p} \geq 2$ at primes of additive reduction with the exact value computed via the Ogg-Saito formula in terms of the conductor exponent of the $ℓ$ -adic Tate module. The conductor is divisible by exactly the primes of bad reduction of $E$ , with multiplicity reflecting the reduction type.

Definition ( $L$ -function of $E$ ). For each prime $p$ define the local $L$ -factor $L_{p} (E, s)$ . At a prime $p$ of good reduction, $L_{p} (E, s) = (1 - a_{p} p^{- s} + p^{1 - 2 s})^{- 1}$ where $a_{p} := p + 1 - N_{p}$ and $N_{p} := # E (F_{p})$ is the number of points on the reduction of $E$ modulo $p$ . At primes of bad reduction, $L_{p} (E, s) = (1 - a_{p} p^{- s})^{- 1}$ with $a_{p} \in {0, \pm 1}$ determined by the reduction type: $a_{p} = + 1$ at split multiplicative reduction, $a_{p} = - 1$ at non-split multiplicative reduction, $a_{p} = 0$ at additive reduction. The global $L$ -function is $$ L(E, s) := \prod_p L_p(E, s), $$ absolutely convergent for $Re (s) > 3/2$ by the Hasse bound $∣ a_{p} ∣ \leq 2 p$ at primes of good reduction.

Definition (modular form of weight $2$ on $Γ_{0} (N)$ ). A weight- $2$ cusp newform on $Γ_{0} (N)$ is a holomorphic function $f : H \to C$ with $f ((a z + b) / (cz + d)) = (cz + d)^{2} f (z)$ for every $(a c b d) \in Γ_{0} (N)$ , vanishing at every cusp, an eigenform for every Hecke operator $T_{p}$ with $p ∤ N$ and every Atkin-Lehner involution $w_{q}$ with $q ∣ N$ , normalised by $a_{1} (f) = 1$ , and not arising via the lift $f \mapsto f (d z)$ from a strictly smaller level $M ∣ N$ .

Modularity theorem (statement; Wiles 1995, BCDT 2001). For every elliptic curve $E$ over $Q$ of conductor $N_{E}$ , there exists a weight- $2$ cusp newform $f_{E}$ on $Γ_{0} (N_{E})$ with rational Hecke eigenvalues such that for every prime $p$ , the local $L$ -factor of $E$ at $p$ equals the local Hecke factor of $f_{E}$ at $p$ ; equivalently, $$ L(E, s) = L(f_E, s). $$ The newform $f_{E}$ is unique with this property up to its determination by the eigenvalue sequence; equivalently, the elliptic curve $E$ is isogenous over $Q$ to the modular abelian variety $A_{f_{E}}$ associated to $f_{E}$ by the Eichler-Shimura construction.

Birch-Swinnerton-Dyer conjecture (rank part; Birch-Swinnerton-Dyer 1965). For every elliptic curve $E$ over $Q$ , $$ \mathrm{ord}_{s = 1} L(E, s) = \mathrm{rank}, E(\mathbb{Q}). $$

Birch-Swinnerton-Dyer conjecture (refined leading-coefficient part; Tate 1974 formulation). Let $r := rank E (Q)$ . Then $$ \lim_{s \to 1} \frac{L(E, s)}{(s - 1)^r} = \frac{\Omega_E \cdot R_E \cdot #\mathrm{Sha}(E/\mathbb{Q}) \cdot \prod_p c_p}{(#E(\mathbb{Q})\mathrm{tors})^2}, $$ *where $Ω_{E}$ is the real period $\int{E(\mathbb{R})} |\omega| $f or t h e N \overset{e}{ˊ} r o n d i f f er e n t ia l$ \omega $,$ R_E $i s t h er e g u l a t or$ \det(\hat h(P_i, P_j)) $o naba s i s$ P_1, \ldots, P_r $o f$ E(\mathbb{Q})/E(\mathbb{Q})_\mathrm{tors} $f or t h e N \overset{e}{ˊ} r o n - T a t ec an o ni c a l h e i g h t$ \hat h $,$ \mathrm{Sha}(E/\mathbb{Q}) $i s t h e T a t e - S ha f a r e v i c h g r o u p, an d$ c_p := [E(\mathbb{Q}_p) : E^0(\mathbb{Q}_p)] $i s t h e l oc a l T ama g a w an u mb er a t$ p$.*

Counterexamples to common slips [Intermediate+]

"The modularity theorem applies to elliptic curves over any number field." The statement above is specific to elliptic curves over $Q$ . The analogous statement over totally real fields is also a theorem (Freitas-Le Hung-Siksek 2015 Invent. Math. for real quadratic; the more general totally real case follows from potential modularity results of Wiles-Taylor and the Skinner-Wiles-Diamond machinery); over general number fields, modularity is conjectural and the object of the modularity-lifting research programme.
" $Sha (E / Q)$ is known to be finite for every elliptic curve $E / Q$ ." The finiteness of $Sha (E / Q)$ is part of the BSD conjecture and remains an open problem for elliptic curves of analytic rank $\geq 2$ . Finiteness is proved by Kolyvagin 1989 for analytic rank $\leq 1$ (combined with the Gross-Zagier formula and modularity), but the general case is unknown.
"The Hasse coefficient $a_{p}$ is defined for every prime $p$ ." The formula $a_{p} = p + 1 - N_{p}$ is the definition at primes of good reduction. At primes of bad reduction, $a_{p}$ is defined via the trace of Frobenius on the $ℓ$ -adic Tate module $T_{ℓ} E$ — or equivalently by the standard reduction-type case analysis ( $+ 1$ for split multiplicative, $- 1$ for non-split multiplicative, $0$ for additive) — and the local Euler factor degenerates to a degree- $1$ polynomial $1 - a_{p} p^{- s}$ at multiplicative reduction or to $1$ at additive reduction.

Key theorem with proof [Intermediate+]

The signature theorem of this unit is the bridge $L (E, s) = L (f_{E}, s)$ at primes of good reduction, the Eichler-Shimura identity that motivates modularity and was already a theorem before Wiles. The full modularity theorem is the statement that this Eichler-Shimura match at good primes extends to a global identity for an explicit newform; that extension is the deep content of Wiles 1995 + BCDT 2001 and is not reproved here. What can be stated and proved at Intermediate level is the local match at primes of good reduction.

Theorem (Eichler-Shimura local match at good primes; Eichler 1954 Arch. Math. 5, Shimura 1958 Tohoku Math. J. 10). Let $E$ be an elliptic curve over $Q$ with conductor $N_{E}$ , and let $f$ be a weight- $2$ cusp newform on $Γ_{0} (N_{E})$ with rational Hecke eigenvalues such that the abelian variety $A_{f}$ associated to $f$ by the Eichler-Shimura construction is isogenous over $Q$ to $E$ . Then for every prime $p ∤ N_{E}$ , $$ a_p(f) = p + 1 - #E(\mathbb{F}_p). $$ Equivalently, the local $L$ -factor of $E$ at $p$ equals the local Hecke factor of $f$ at $p$ .

Proof. Fix a prime $ℓ \neq = p$ and consider the $ℓ$ -adic Tate module $T_{ℓ} E := lim_{n} E [ℓ^{n}]$ . By the Eichler-Shimura construction, the abelian variety $A_{f}$ is a quotient of the Jacobian $J_{0} (N_{E}) := Jac (X_{0} (N_{E}))$ by an ideal $I_{f} \subseteq T$ of the Hecke algebra, with the property that the $ℓ$ -adic Tate module $T_{ℓ} A_{f}$ carries a natural action of the Hecke algebra and of $Gal (\overline{Q} / Q)$ , the two actions commuting.

Since $E$ is isogenous to $A_{f}$ over $Q$ , there is an isomorphism of Galois representations $$ V_\ell E := T_\ell E \otimes_{\mathbb{Z}\ell} \mathbb{Q}\ell \cong V_\ell A_f := T_\ell A_f \otimes_{\mathbb{Z}\ell} \mathbb{Q}\ell. $$ Both sides are $2$ -dimensional $Q_{ℓ}$ -vector spaces with a continuous action of $Gal (\overline{Q} / Q)$ , unramified outside $N_{E} ℓ$ .

The Eichler-Shimura congruence on $X_{0} (N_{E})$ states that the Hecke operator $T_{p}$ on $X_{0} (N_{E})$ modulo $p$ is the sum of two correspondences, namely the Frobenius and its Verschiebung (Eichler 1954): $T_{p} = Frob_{p} + p \cdot Ver_{p}$ as endomorphisms of $J_{0} (N_{E})$ in characteristic $p$ . On $ℓ$ -adic cohomology this becomes a statement about the action of Frobenius $Frob_{p} \in Gal (\overline{Q_{p}^{ur}} / Q_{p})$ on $V_{ℓ} A_{f}$ : the characteristic polynomial of $Frob_{p}$ on $V_{ℓ} A_{f}$ is $$ \det(1 - \mathrm{Frob}p \cdot X \mid V\ell A_f) = 1 - a_p(f) X + p X^2, $$ since the trace of Frobenius is the Hecke eigenvalue $a_{p} (f)$ and the determinant is the cyclotomic character $p^{k - 1} = p$ at weight $k = 2$ .

By the isomorphism $V_{ℓ} E ≅ V_{ℓ} A_{f}$ as Galois representations, the characteristic polynomial of $Frob_{p}$ on $V_{ℓ} E$ is the same: $$ \det(1 - \mathrm{Frob}p \cdot X \mid V\ell E) = 1 - a_p(f) X + p X^2. $$

By the Hasse-Weil theorem (the local zeta function of $E$ over $F_{p}$ ), the trace of Frobenius on $V_{ℓ} E$ at a prime $p$ of good reduction satisfies $$ \det(1 - \mathrm{Frob}p \cdot X \mid V\ell E) = 1 - (p + 1 - N_p) X + p X^2, $$ where $N_{p} = # E (F_{p})$ . Comparing coefficients of $X^{1}$ gives $a_{p} (f) = p + 1 - N_{p}$ , the desired identity. The local $L$ -factor identification follows by substituting $X = p^{- s}$ in both characteristic polynomials and taking reciprocals. $□$

Bridge. This local Eichler-Shimura identity is what the modularity theorem of Wiles-BCDT lifts to a global existence statement: starting from any $E / Q$ , the theorem asserts the existence of a newform $f_{E}$ on $Γ_{0} (N_{E})$ such that $A_{f_{E}}$ is isogenous to $E$ — equivalently, such that the Galois representation $V_{ℓ} E$ is the Galois representation attached to a weight- $2$ cusp newform. The central insight is that the $L$ -function identity $L (E, s) = L (f_{E}, s)$ is equivalent to the isomorphism $V_{ℓ} E ≅ V_{ℓ} A_{f_{E}}$ as Galois representations, by the Tate-Faltings isogeny theorem (Faltings 1983 Invent. Math. 73) identifying abelian varieties up to isogeny with their $ℓ$ -adic Galois representations. The modularity theorem is therefore a Galois-representation statement: every $ℓ$ -adic representation arising from an elliptic curve over $Q$ is modular. This statement builds toward the Langlands programme 21.10.01 pending, where modularity for $GL_{2}$ over $Q$ generalises to automorphy for $GL_{n}$ over arbitrary global fields. The bridge is from the geometric object $E$ to the analytic object $f_{E}$ , mediated by the Galois representation $V_{ℓ} E$ , and it appears again in 21.05.01 $ℓ$ -adic Galois representations and 21.04.03 Eichler-Shimura at the technical level of the construction.

Exercises [Intermediate+]

Exercise 3 (medium, symbolic).

State the local $L$ -factor of an elliptic curve $E / Q$ at a prime $p$ of split multiplicative reduction, and explain why it has degree $1$ in $p^{- s}$ rather than degree $2$ .

Hint

At multiplicative reduction, the Tate module $V_{ℓ} E$ has a one-dimensional inertia-fixed subspace. The local Euler factor is the characteristic polynomial of Frobenius on this subspace.

Answer

Local $L$ -factor at split multiplicative reduction. $L_{p} (E, s) = (1 - p^{- s})^{- 1}$ , equivalently $a_{p} = + 1$ in the standard parametrisation $L_{p} (E, s) = (1 - a_{p} p^{- s})^{- 1}$ .

Reason for degree $1$ . At a prime $p$ of multiplicative reduction, the reduction of $E$ modulo $p$ is a nodal cubic curve, whose smooth locus is the multiplicative group $G_{m}$ over $F_{p}$ in the split case (and a non-split form of $G_{m}$ in the non-split case). The Tate module $V_{ℓ} E$ has a $1$ -dimensional subspace fixed by the inertia subgroup at $p$ — the toric character — and a $1$ -dimensional quotient on which inertia acts non-identically. The local Euler factor of $L (E, s)$ at $p$ is the characteristic polynomial of Frobenius on the inertia-fixed part, which is $1$ -dimensional, hence the factor has degree $1$ in $p^{- s}$ . The sign $a_{p} = \pm 1$ records whether Frobenius acts on the toric character as $+ 1$ (split) or $- 1$ (non-split).

At additive reduction, the inertia-fixed part of $V_{ℓ} E$ is $0$ -dimensional and the local Euler factor is $1$ (degree $0$ ); this matches $a_{p} = 0$ in the parametrisation.

Exercise 4 (medium, symbolic).

Verify that the modularity theorem implies the analytic continuation and functional equation of $L (E, s)$ .

Hint

The $L$ -function of a weight- $2$ cusp newform on $Γ_{0} (N)$ has known analytic continuation and functional equation. Apply the identity $L (E, s) = L (f_{E}, s)$ .

Answer

Modular-form side. For a weight- $2$ cusp newform $f$ on $Γ_{0} (N)$ , the $L$ -function $L (f, s) = \sum_{n \geq 1} a_{n} (f) n^{- s}$ has an integral representation $Λ (f, s) = N^{s /2} (2 π)^{- s} Γ (s) L (f, s) = \int_{0}^{\infty} f (i y) y^{s - 1} d y$ , convergent for all $s \in C$ once one cuts the integral at $y = N^{- 1/2}$ and uses the functional equation $f (- 1/ (N z)) = \pm N z^{2} f (z)$ of $f$ under the Atkin-Lehner involution $w_{N}$ . This yields the analytic continuation of $L (f, s)$ to $C$ and the functional equation $$ \Lambda(f, s) = \varepsilon_f \Lambda(f, 2 - s), $$ where $ε_{f} \in {\pm 1}$ is the sign of the Atkin-Lehner eigenvalue.

Elliptic-curve side. Applying modularity $L (E, s) = L (f_{E}, s)$ transfers the analytic continuation and functional equation to $L (E, s)$ : $$ \Lambda(E, s) := N_E^{s/2} (2 \pi)^{-s} \Gamma(s) L(E, s) = \varepsilon_E \Lambda(E, 2 - s). $$ Before modularity was proved, the analytic continuation and functional equation of $L (E, s)$ were the Hasse-Weil conjecture — proved only for elliptic curves with complex multiplication (Deuring 1953) or for those known to be modular (Eichler 1954, Shimura 1958 in finitely many cases). The general case was equivalent to modularity by a theorem of Weil 1967, the converse theorem: $L (E, s)$ admits the expected analytic continuation and functional equation if and only if $E$ is modular.

Exercise 5 (medium, symbolic).

State the BSD conjecture in its refined form for an elliptic curve $E / Q$ of rank $r$ , identifying every invariant appearing in the leading-coefficient formula.

Hint

The leading-coefficient formula has six invariants: real period, regulator, Tate-Shafarevich, Tamagawa numbers, torsion, and the rank $r$ itself.

Answer

Refined BSD. For $E / Q$ of rank $r := rank E (Q)$ , the order of vanishing of $L (E, s)$ at $s = 1$ equals $r$ , and the leading coefficient of the Taylor expansion is $$ \lim_{s \to 1} \frac{L(E, s)}{(s - 1)^r} = \frac{\Omega_E \cdot R_E \cdot #\mathrm{Sha}(E/\mathbb{Q}) \cdot \prod_p c_p}{(#E(\mathbb{Q})_\mathrm{tors})^2}. $$

Invariants.

$Ω_{E}$ , the real period: $Ω_{E} = \int_{E (R)} ∣ ω ∣$ where $ω = d x / (2 y + a_{1} x + a_{3})$ is the Néron differential, normalised against the Néron minimal model. For $E (R)$ disconnected (two real components), $Ω_{E}$ is the integral over the identity component, multiplied by $2$ .
$R_{E}$ , the regulator: $R_{E} = det (\hat{h} (P_{i}, P_{j}))_{i, j = 1, \dots, r}$ for a $Z$ -basis $P_{1}, \dots, P_{r}$ of $E (Q) / E (Q)_{tors}$ , with $\hat{h}$ the Néron-Tate canonical height. The regulator is the volume of the lattice $E (Q) / tors$ in the Mordell-Weil $R$ -vector space.
$# Sha (E / Q)$ , the order of the Tate-Shafarevich group: $Sha (E / Q) := ker (H^{1} (Gal (\overline{Q} / Q), E (\overline{Q})) \to \prod_{v} H^{1} (G_{v}, E (\overline{Q_{v}})))$ , the group of $E$ -torsors that are locally split at every place. Conjecturally finite; provably finite for analytic rank $\leq 1$ by Kolyvagin 1989.
$c_{p}$ , the Tamagawa number at $p$ : $c_{p} := [E (Q_{p}) : E^{0} (Q_{p})]$ where $E^{0} (Q_{p})$ is the identity component of the Néron model of $E$ over $Z_{p}$ . By Kodaira's local classification, $c_{p} \in {1, 2, 3, 4}$ or $c_{p} = m$ at type $I_{m}$ multiplicative reduction.
$# E (Q)_{tors}$ , the order of the torsion subgroup: classified by Mazur 1977 as one of fifteen abelian groups of order at most $16$ .
$r$ , the rank of $E (Q)$ as a finitely generated abelian group.

The refined formula is the precise statement of BSD; the rank part $ord_{s = 1} L (E, s) = r$ is the most well-known consequence.

Exercise 6 (hard, symbolic).

State the Frey-Ribet-Serre implication chain that derives Fermat's Last Theorem from modularity.

Hint

The chain has three steps: (a) Frey's construction of an elliptic curve from a Fermat counterexample; (b) Serre's $ε$ -conjecture and Ribet's proof predicting the mod- $ℓ$ Galois representation of Frey's curve has level $1$ ; (c) Wiles's modularity theorem ruling out a newform at level $1$ with the predicted properties.

Answer

Step 1 (Frey 1986). Suppose $a^{ℓ} + b^{ℓ} = c^{ℓ}$ is a non-zero integer solution to the Fermat equation at prime exponent $ℓ \geq 5$ , with $g cd (a, b, c) = 1$ and $a \equiv - 1 (mod 4)$ , $b$ even. The Frey curve is $$ E_{a, b, c} : y^2 = x (x - a^\ell)(x + b^\ell). $$ It is an elliptic curve over $Q$ with discriminant $Δ = (ab c)^{2 ℓ} /16$ . The minimal model has conductor $N = rad (ab c) := \prod_{p ∣ ab c} p$ , a squarefree integer.

Step 2 (Serre 1987 + Ribet 1990). The mod- $ℓ$ Galois representation $\overline{ρ}_{E, ℓ} : Gal (\overline{Q} / Q) \to GL_{2} (F_{ℓ})$ on $E_{a, b, c} [ℓ]$ is irreducible. Serre's $ε$ -conjecture predicts the level of $\overline{ρ}_{E, ℓ}$ , i.e. the level $N_{ℓ}$ of any newform realising $\overline{ρ}_{E, ℓ}$ . Computation: $\overline{ρ}_{E, ℓ}$ is unramified at every prime $p ∤ ℓ$ in the conductor $rad (ab c)$ except possibly $p = 2$ , and the level-lowering theorem of Ribet 1990 Invent. Math. 100 proves the $ε$ -conjecture in this setting: $\overline{ρ}_{E, ℓ}$ comes from a newform of level $N_{ℓ} = 2$ .

Step 3 (Wiles 1995, BCDT 2001). Modularity says $E_{a, b, c}$ is modular: there exists a weight- $2$ cusp newform $f$ of level $rad (ab c)$ with $\overline{ρ}_{f, ℓ} = \overline{ρ}_{E, ℓ}$ . Combined with Ribet's level-lowering, $\overline{ρ}_{f, ℓ}$ comes from a newform of level $2$ . But $S_{2} (Γ_{0} (2)) = 0$ — there are no nonzero weight- $2$ cusp forms at level $2$ , since $dim S_{2} (Γ_{0} (2)) = 0$ from the dimension formula. Contradiction.

Conclusion. No such $(a, b, c)$ exists; Fermat's Last Theorem $a^{ℓ} + b^{ℓ} = c^{ℓ}$ has no integer solutions with $ab c \neq = 0$ for prime $ℓ \geq 5$ . The case $ℓ = 3$ is due to Euler 1770, $ℓ = 4$ to Fermat himself (descent on $x^{4} + y^{4} = z^{2}$ ), so the implication chain delivers FLT in full.

Exercise 7 (hard, symbolic).

State the partial results towards BSD: Coates-Wiles 1977, Gross-Zagier 1986, Kolyvagin 1989, Kato 2004, Skinner-Urban 2014.

Hint

The known cases of BSD cover analytic rank $\leq 1$ for modular elliptic curves, plus refined statements about the $p$ -part of the leading-coefficient formula via Iwasawa theory.

Answer

Coates-Wiles 1977 Invent. Math. 39. For an elliptic curve $E / Q$ with complex multiplication by an imaginary quadratic field $K$ , if $L (E, 1) \neq = 0$ then $E (Q)$ is finite. This is the BSD rank- $0$ implication for CM curves, proved via the elliptic-unit Euler system of Kubota-Leopoldt — the first substantial result towards BSD.

Gross-Zagier 1986 Invent. Math. 84. For a weight- $2$ newform $f$ on $Γ_{0} (N)$ with sign $ε_{f} = - 1$ in the functional equation (forcing $L (f, 1) = 0$ ), and an imaginary quadratic field $K$ in which all primes dividing $N$ split, the Heegner point $y_{K} \in E (K)$ associated to $K$ on the modular abelian variety $A_{f}$ satisfies $$ \hat h(y_K) = c_E \cdot L'(E, 1) \cdot L(E^K, 1), $$ where $\hat{h}$ is the Néron-Tate canonical height, $E^{K}$ is the quadratic twist, and $c_{E} > 0$ is an explicit non-zero constant. Consequence: if $L^{'} (E, 1) \neq = 0$ , then $y_{K}$ is a non-torsion point in $E (K)$ , hence in $E (Q)$ after taking trace, so $rank E (Q) \geq 1$ .

Kolyvagin 1989 Math. USSR-Izv. 32. For a modular elliptic curve $E / Q$ , if $ord_{s = 1} L (E, s) \leq 1$ (analytic rank $0$ or $1$ ), then:

$rank E (Q) = ord_{s = 1} L (E, s)$ ,
$Sha (E / Q)$ is finite.

The proof uses Heegner points to build a Kolyvagin Euler system that bounds the Selmer group $Sel (E / Q)$ above by the analytic rank. Combined with Gross-Zagier (for the lower bound at analytic rank $1$ ) and modularity (now a theorem post-Wiles), this gives the rank part of BSD in the analytic-rank $\leq 1$ case.

Kato 2004 Astérisque 295. Construction of the Kato Euler system of Beilinson elements in $K_{2}$ of modular curves. This bounds the Selmer group of $E$ over the cyclotomic $Z_{p}$ -extension of $Q$ above by the $p$ -adic $L$ -function of $E$ , proving one inclusion of the Iwasawa main conjecture for elliptic curves. Implies the $p$ -part of BSD for modular elliptic curves in many cases.

Skinner-Urban 2014 Invent. Math. 195. The reverse inclusion of the Iwasawa main conjecture for $GL_{2}$ in many cases, via Eisenstein congruences on $GU (2, 2)$ and the construction of Galois representations attached to Eisenstein-cohomological automorphic forms. Combined with Kato 2004, gives the full main conjecture for these cases, implying the $p$ -part of the refined BSD formula in analytic rank $\leq 1$ .

Open. BSD remains entirely open for elliptic curves of analytic rank $\geq 2$ . No single elliptic curve over $Q$ of analytic rank $\geq 2$ is known to satisfy the rank part of BSD unconditionally, although the Bhargava-Skinner-Zhang programme (2014+) has proved BSD holds in a positive proportion of elliptic curves ordered by height.

Exercise 8 (hard, symbolic).

Explain the role of the Galois representation $\overline{ρ}_{E, ℓ}$ and the strategy of Wiles 1995 at a high level: the deformation framework, the $R = T$ identification, and the Taylor-Wiles patching argument.

Hint

The proof identifies a universal deformation ring $R$ (parameterising lifts of the residual representation) with a Hecke algebra $T$ (parameterising modular newforms with matching residual representation). The identification implies that every lift comes from a modular form.

Answer

Strategy. Fix a prime $ℓ$ , an elliptic curve $E / Q$ , and its mod- $ℓ$ Galois representation $\overline{ρ}_{E, ℓ} : G_{Q} \to GL_{2} (F_{ℓ})$ . Suppose $\overline{ρ}_{E, ℓ}$ is irreducible and known to be modular — meaning there exists a weight- $2$ newform $\overline{f}$ with $\overline{ρ}_{\overline{f}, ℓ} ≅ \overline{ρ}_{E, ℓ}$ . The goal is to lift this residual modularity to modularity of $E$ itself: a weight- $2$ newform $f_{E}$ with $ρ_{f_{E}, ℓ} ≅ ρ_{E, ℓ}$ on the $ℓ$ -adic representation, not just the mod- $ℓ$ one.

*Deformation ring $R$ (Mazur 1989 Galois groups over $Q$ ).* The set of liftings of $\overline{ρ}_{E, ℓ}$ to a continuous representation $G_{Q} \to GL_{2} (A)$ for $A$ a complete Noetherian local $Z_{ℓ}$ -algebra with residue field $F_{ℓ}$ , satisfying prescribed local conditions at each prime of bad reduction, is representable by a universal deformation ring $R^{□}$ .

Hecke algebra $T$ . The Hecke algebra $T^{□}$ is the $Z_{ℓ}$ -algebra generated by Hecke operators $T_{p}$ for primes $p$ of good reduction, acting on the appropriate space of weight- $2$ newforms with residual representation $\overline{ρ}_{E, ℓ}$ and matching local conditions. There is a natural surjection $R^{□} ↠ T^{□}$ since every modular form gives a Galois representation, hence a deformation.

$R = T$ theorem. Wiles's central claim: the natural surjection $R^{□} \to T^{□}$ is an isomorphism. Consequence: every deformation comes from a modular form, in particular $ρ_{E, ℓ}$ does, proving modularity of $E$ .

Taylor-Wiles patching (Taylor-Wiles 1995 Ann. Math. 141). The proof of $R = T$ proceeds via a numerical criterion: introduce auxiliary primes $q_{1}, \dots, q_{n}$ chosen so the local deformation conditions at the $q_{i}$ create enough freedom to compute the cotangent space of $R$ , and show inductively (the patching argument, taking $n \to \infty$ and a profinite limit) that $R$ and $T$ have the same dimension as $Z_{ℓ}$ -modules. Coupled with the surjection, this forces isomorphism.

The $3$ - $5$ switch. The argument requires the residual representation $\overline{ρ}_{E, ℓ}$ to be modular at some prime $ℓ$ , but modularity of a mod- $ℓ$ representation is itself a substantive condition. Wiles uses Langlands-Tunnell 1980 Inventiones 78 to handle $ℓ = 3$ (mod- $3$ representations are modular because they have image in $GL_{2} (F_{3})$ , solvable, and Langlands-Tunnell covers the solvable case). When $\overline{ρ}_{E, 3}$ is reducible, he switches to $ℓ = 5$ via an auxiliary $mod 15$ argument, reducing again to mod- $3$ modularity of an auxiliary curve.

BCDT 2001 extension. Wiles 1995 + Taylor-Wiles 1995 proved modularity for semistable elliptic curves over $Q$ — those whose reduction at every bad prime is multiplicative, not additive. Breuil-Conrad-Diamond-Taylor 2001 J. AMS 14 extended the argument to all elliptic curves over $Q$ by handling the wild $3$ -adic deformation theory at additive-reduction primes, in particular the case where $\overline{ρ}_{E, 3}$ is induced from a character of $Gal (\overline{Q_{3}} / Q_{3} (ζ_{3}))$ .

Lean formalization [Intermediate+]

The companion file lean/Codex/NumberTheory/Modularity/ModularityBSD.lean records the modularity theorem statement, the BSD conjecture as a structure, and the supporting types (Tate-Shafarevich group, regulator, period, conductor, $L$ -function) as sorry-stubbed declarations on top of Mathlib's developing elliptic-curve type. The Lean kernel has five components.

First, the conductor def conductor (E : EllipticCurve ℚ) : ℕ returning the arithmetic conductor of $E$ , with a Lean axiom recording the Ogg-Saito formula as a sorry-stubbed expectation tying local exponents at bad primes to the reduction type. Mathlib supplies the type EllipticCurve ℚ via the Weierstrass model machinery; the conductor is the first non-elementary invariant that does not yet exist as a Mathlib definition.

Second, the $L$ -function as a function def lFunction (E : EllipticCurve ℚ) : ℂ → ℂ defined as the Euler product $\prod_{p} L_{p} (E, s)^{- 1}$ over primes, with local factors given by the case analysis (good reduction: degree- $2$ Frobenius polynomial; bad reduction: degree- $1$ or $0$ according to reduction type). The sorry-stubbed declaration includes the convergence statement on $Re (s) > 3/2$ and the analytic-continuation expectation (which, once proved upstream, lifts to the full modularity theorem statement).

Third, the modularity theorem as a theorem declaration:

theorem modularity_theorem (E : EllipticCurve ℚ) :
    ∃ f : CuspNewform 2 (Gamma₀ (conductor E)),
      ∀ s, lFunction E s = lFunction_of_form f s := sorry

The statement asserts the existence of a weight- $2$ cusp newform on $Γ_{0} (N_{E})$ realising the same $L$ -function as $E$ . The full proof would import the Wiles-Taylor-Wiles + BCDT machinery, none of which Mathlib supplies; the statement itself is the substantive declaration.

Fourth, the Tate-Shafarevich group as a def:

def shaftaGroup (E : EllipticCurve ℚ) : Type :=
  { c : H¹ (Gal ℚ̄ / ℚ) (E.points ℚ̄) //
    ∀ v : Place ℚ, restriction_to v c = 0 }

with H¹ the first Galois cohomology and Place ℚ ranging over archimedean and non-archimedean places. The finiteness assertion is a separate sorry-stubbed theorem (a consequence of BSD, proved only in analytic rank $\leq 1$ by Kolyvagin).

Fifth, the BSD conjecture as a structure packaging the rank statement, the refined leading-coefficient formula, and all relevant invariants:

structure BirchSwinnertonDyerConjecture (E : EllipticCurve ℚ) where
  rank_eq_ord :
    Nat.cast (Module.rank ℤ (Mordell_Weil E ℚ)) =
      (lFunction E).order_at 1
  leading_coeff_formula :
    leading_term (lFunction E) 1 =
      realPeriod E * regulator E * (Nat.card (shaftaGroup E)) *
      (∏ p, tamagawaNumber E p) /
      (Nat.card (torsionSubgroup E ℚ))^2

The structure makes the conjecture a Lean object — checkable, when its constituent declarations are all proved upstream, by inhabiting BirchSwinnertonDyerConjecture E for each $E$ .

The full Lean formalisation of modularity + BSD is the single most ambitious target in the arithmetic-geometry corner of the Mathlib roadmap. Every constituent (conductor, $L$ -function, Mordell-Weil, Tate-Shafarevich, regulator, period, Tamagawa numbers) is currently absent from Mathlib at the level required, and each is itself a substantive multi-month formalisation effort. The companion file makes the gap explicit: the named theorems and structures provide stable hook points for future work.

Advanced results [Master]

Statement of the modularity theorem (Wiles 1995, BCDT 2001)

Theorem 1 (Modularity theorem; Wiles 1995 Ann. Math. 141 + Taylor-Wiles 1995 Ann. Math. 141 + BCDT 2001 J. AMS 14). For every elliptic curve $E$ over $Q$ of conductor $N_{E}$ , there exists a weight- $2$ cusp newform $f_{E}$ on $Γ_{0} (N_{E})$ with rational Hecke eigenvalues such that $L (E, s) = L (f_{E}, s)$ . The form $f_{E}$ is determined up to its Hecke eigenvalue sequence, and the elliptic curve $E$ is isogenous over $Q$ to the modular abelian variety $A_{f_{E}}$ associated to $f_{E}$ by the Eichler-Shimura construction.

Equivalent formulations. Three equivalent ways to state modularity:

(a) $L$ -function identity. $L (E, s) = L (f_{E}, s)$ for some weight- $2$ cusp newform $f_{E}$ on $Γ_{0} (N_{E})$ .

(b) Galois-representation modularity. The $ℓ$ -adic Galois representation $ρ_{E, ℓ} : G_{Q} \to GL_{2} (Q_{ℓ})$ on the $ℓ$ -adic Tate module $V_{ℓ} E$ is isomorphic to the $ℓ$ -adic representation $ρ_{f_{E}, ℓ}$ attached to a weight- $2$ cusp newform $f_{E}$ .

(c) Geometric modularity. There exists a non-constant morphism $X_{0} (N_{E}) \to E$ of curves over $Q$ , where $X_{0} (N_{E})$ is the compactified modular curve at level $N_{E}$ ; equivalently, $E$ is a quotient of the Jacobian $J_{0} (N_{E}) = Jac (X_{0} (N_{E}))$ by an ideal of the Hecke algebra.

The equivalence of (a) and (b) follows from the Eichler-Shimura identification at primes of good reduction combined with the Tate-Faltings isogeny theorem (Faltings 1983 Invent. Math. 73). The equivalence of (b) and (c) is the Eichler-Shimura construction itself: the abelian variety $A_{f_{E}}$ realising $ρ_{f_{E}, ℓ}$ is by construction a quotient of $J_{0} (N_{E})$ .

Proof method. The proof proceeds by establishing (b) via the $R = T$ theorem of Wiles + Taylor-Wiles, with the BCDT extension handling the additive-reduction cases left open in Wiles 1995. The strategy of the proof is summarised in Exercise 8: residual modularity at $ℓ = 3$ (via Langlands-Tunnell 1980 Invent. Math. 78), bootstrapping to $ℓ$ -adic modularity via the universal-deformation-ring $R$ , identification with a Hecke algebra $T$ via Taylor-Wiles patching, conclusion that every deformation comes from a modular form.

Statement of BSD (Birch-Swinnerton-Dyer 1965, Tate 1974 refined form)

Theorem 2 (BSD conjecture; Birch-Swinnerton-Dyer 1965 Crelle 218 + Tate 1974 Invent. Math. 23). Let $E$ be an elliptic curve over $Q$ of rank $r := rank E (Q)$ . Then:

(BSD-1) Rank statement. The order of vanishing of $L (E, s)$ at $s = 1$ equals the rank of $E (Q)$ : $$ \mathrm{ord}_{s = 1} L(E, s) = r. $$

(BSD-2) Refined leading-coefficient formula. The leading coefficient of the Taylor expansion of $L (E, s)$ at $s = 1$ is $$ \lim_{s \to 1} \frac{L(E, s)}{(s - 1)^r} = \frac{\Omega_E \cdot R_E \cdot #\mathrm{Sha}(E/\mathbb{Q}) \cdot \prod_p c_p}{(#E(\mathbb{Q})\mathrm{tors})^2}, $$ *with $Ω_{E}$ the real period, $R_{E}$ the regulator, $Sha (E / Q)$ the Tate-Shafarevich group, $c_{p}$ the Tamagawa numbers at primes $p$ of bad reduction, and $E(\mathbb{Q})\mathrm{tors}$ the torsion subgroup.*

Historical note. Birch and Swinnerton-Dyer 1965 Crelle 218 formulated the conjecture in an integral form: $\prod_{p \leq X} (N_{p} / p) \sim C \cdot (lo g X)^{r}$ as $X \to \infty$ , with $C$ a non-zero constant. Tate 1974 Invent. Math. 23 reformulated this in the modern leading-coefficient form, identifying the constant $C$ with the explicit product of invariants given above. The conjecture is the Clay Millennium Open Problem announced 2000; the one-million-dollar Clay prize remains unclaimed as of 2026.

Frey-Ribet-Serre chain to Fermat (1985-1995)

Theorem 3 (Ribet 1990 Invent. Math. 100). Serre's $ε$ -conjecture on level-lowering for mod- $ℓ$ Galois representations is true: if $\overline{ρ} : G_{Q} \to GL_{2} (F_{ℓ})$ is an irreducible continuous mod- $ℓ$ Galois representation coming from a weight- $2$ cusp newform of level $N$ and unramified at a prime $p ∥ N$ (multiplicative-reduction prime appearing to first power), then $\overline{ρ}$ also comes from a weight- $2$ cusp newform of level $N / p$ .

Implication chain. Combining Frey 1986, Ribet 1990, and Wiles 1995:

Suppose $a^{ℓ} + b^{ℓ} = c^{ℓ}$ is a non-zero integer solution with prime $ℓ \geq 5$ . Frey 1986 constructs the elliptic curve $E_{a, b, c} : y^{2} = x (x - a^{ℓ}) (x + b^{ℓ})$ of conductor $rad (ab c)$ and discriminant $(ab c)^{2 ℓ} /16$ .
Serre 1987 Duke 54 conjectures that the mod- $ℓ$ representation $\overline{ρ}_{E, ℓ}$ has level $2$ (the level-lowered prediction). Ribet 1990 proves this conjecture for the relevant family.
Wiles 1995 + BCDT 2001 prove $E_{a, b, c}$ is modular: $\overline{ρ}_{E, ℓ}$ comes from a weight- $2$ cusp newform $f$ of level $rad (ab c)$ .
Combined with Ribet, $\overline{ρ}_{E, ℓ}$ also comes from a weight- $2$ cusp newform of level $2$ . But $dim S_{2} (Γ_{0} (2)) = 0$ . Contradiction.
Fermat's Last Theorem for prime exponent $ℓ \geq 5$ follows; the cases $ℓ = 3$ (Euler 1770) and $ℓ = 4$ (Fermat himself, descent) complete the theorem.

This chain — Frey 1985-86 ⇒ Ribet 1986-90 ⇒ Wiles 1995 — is the most celebrated arithmetic-geometric proof of the twentieth century. The bridge from Fermat's seventeenth-century question to modular forms was understood by Frey 1985; the missing piece — modularity — was Wiles's 1995 contribution.

Partial results towards BSD (1977-2014)

Theorem 4 (Coates-Wiles 1977 Invent. Math. 39). Let $E / Q$ have complex multiplication by an imaginary quadratic field $K$ . If $L (E, 1) \neq = 0$ , then $E (Q)$ is finite.

This is the BSD rank- $0$ implication for CM curves. Proof method: the elliptic-unit Euler system on the Kubota-Leopoldt $p$ -adic $L$ -function bounds the Selmer group of $E$ above by the $p$ -adic $L$ -value, and non-vanishing at $s = 1$ forces the Selmer group to be finite, hence $E (Q)$ finite via the Mordell-Weil exact sequence.

Theorem 5 (Gross-Zagier 1986 Invent. Math. 84). Let $f$ be a weight- $2$ cusp newform on $Γ_{0} (N)$ with $ε_{f} = - 1$ (forcing $L (f, 1) = 0$ ), and let $K$ be an imaginary quadratic field in which all primes $p ∣ N$ split. The Heegner point $y_{K} \in E (K)$ associated to $K$ on the modular abelian variety $E = A_{f}$ satisfies $$ \hat h(y_K) = c_E \cdot L'(E/K, 1) = c_E \cdot L'(E, 1) \cdot L(E^K, 1), $$ where $\hat{h}$ is the Néron-Tate height, $E^{K}$ is the quadratic twist of $E$ by $K$ , and $c_{E} > 0$ is a non-zero explicit constant.

Corollary. If $L^{'} (E, 1) \neq = 0$ (analytic rank $1$ ), then $y_{K}$ is non-torsion, and $rank E (Q) \geq 1$ (by descending the Heegner point trace to $Q$ ).

Theorem 6 (Kolyvagin 1989 Math. USSR-Izv. 32). Let $E / Q$ be a modular elliptic curve with $ord_{s = 1} L (E, s) \leq 1$ . Then $$ \mathrm{rank}, E(\mathbb{Q}) = \mathrm{ord}_{s = 1} L(E, s), \qquad #\mathrm{Sha}(E/\mathbb{Q}) < \infty. $$

The proof uses the Kolyvagin Euler system of Heegner points on towers of imaginary quadratic fields, bounding the Selmer group $Sel (E / Q)$ above by the analytic rank. Combined with Gross-Zagier (which provides the lower bound in analytic rank $1$ via the non-torsion Heegner point) and modularity (now a theorem post-Wiles), this gives the rank part of BSD in analytic rank $\leq 1$ .

Theorem 7 (Kato 2004 Astérisque 295). Let $E / Q$ be a modular elliptic curve and $p$ a prime. The Kato Euler system of Beilinson elements in $K_{2}$ of modular curves bounds the Selmer group of $E$ over the cyclotomic $Z_{p}$ -extension $Q_{\infty} / Q$ above by the Mazur-Swinnerton-Dyer $p$ -adic $L$ -function $L_{p} (E, s) \in Z_{p} [[T]]$ : $$ \mathrm{char}_{\mathbb{Z}p[![T]!]} \mathrm{Sel}(E/\mathbb{Q}\infty)^\vee \mid L_p(E, T). $$ This is one inclusion of the Iwasawa main conjecture for $E$ .

Theorem 8 (Skinner-Urban 2014 Invent. Math. 195). Under modest hypotheses (ordinary reduction, irreducible residual representation, etc.), the reverse inclusion holds: $$ L_p(E, T) \mid \mathrm{char}_{\mathbb{Z}p[![T]!]} \mathrm{Sel}(E/\mathbb{Q}\infty)^\vee. $$ Combined with Kato, the full Iwasawa main conjecture holds for these elliptic curves, implying the $p$ -part of the refined BSD formula in analytic rank $\leq 1$ .

Open. BSD is open in analytic rank $\geq 2$ for every specific elliptic curve. The Bhargava-Skinner-Zhang programme (Bhargava-Shankar 2010s + Skinner-Zhang 2014+) has proved BSD holds in a positive proportion of elliptic curves over $Q$ when ordered by naive height; this is a density result, not a result for individual curves.

Sato-Tate refinement and the Langlands programme

Theorem 9 (Sato-Tate; Clozel-Harris-Shepherd-Barron-Taylor 2008-2011). Let $E / Q$ be a non-CM elliptic curve, and write $a_{p} = 2 p cos θ_{p}$ with $θ_{p} \in [0, π]$ (well-defined by Hasse). Then the angles ${θ_{p}}_{p}$ are equidistributed in $[0, π]$ with respect to the Sato-Tate measure $(2/ π) sin^{2} θ d θ$ as $p \to \infty$ over primes.

This is a quantitative refinement of the Hasse bound, proved as a corollary of modularity + automorphy of symmetric powers of $ρ_{E, ℓ}$ , which is itself a deep Langlands-functoriality result. The successor unit 21.06.02 pending develops the Sato-Tate conjecture in detail; the present unit notes only that Sato-Tate is the natural sharpening of modularity, putting an equidistribution structure on the Frobenius angles.

Synthesis. Modularity is the foundational reason elliptic curves over $Q$ behave as automorphic objects. The central insight is that the $L$ -function $L (E, s)$ — defined geometrically from point counts on the curve modulo primes — coincides with the $L$ -function of a weight- $2$ cusp newform $f_{E}$ , an analytic object built from holomorphy and the Petersson inner product on the upper half-plane. This is exactly the bridge from Diophantine geometry to automorphic forms: the elliptic curve's arithmetic invariants (rank, conductor, Tate-Shafarevich group) are encoded in a modular form's analytic invariants (Hecke eigenvalues, level, functional-equation sign), and the modularity theorem identifies the two encodings.

The Frey-Ribet-Serre-Wiles chain putting these together with Fermat's Last Theorem is the most celebrated arithmetic-geometric proof of the twentieth century. Frey's 1985 observation — that a hypothetical Fermat solution constructs an elliptic curve whose mod- $ℓ$ Galois representation has impossibly small level — generalises to a method of attacking other Diophantine problems via modularity. The recurring pattern: modularity + level-lowering ⇒ Diophantine consequence. Variants of this method have since proved generalised Fermat equations, the Catalan-Tijdeman-Zhang result on perfect powers, and dozens of further special cases of Beal-type conjectures.

The bridge from modularity to BSD is via the $L$ -function. Once $L (E, s) = L (f_{E}, s)$ , the analytic continuation and functional equation of $L (E, s)$ to all of $C$ is inherited from the modular-form side, and the order of vanishing $ord_{s = 1} L (E, s)$ becomes a definable analytic invariant. BSD then predicts this vanishing order equals the Mordell-Weil rank, with the leading coefficient an explicit product of arithmetic invariants. The pattern recurs in the Langlands programme: every $L$ -function attached to an arithmetic object should have a special-value formula relating analytic and arithmetic invariants — the Beilinson conjectures generalising BSD to higher-dimensional varieties (Beilinson 1984 J. Soviet Math. 30), the Bloch-Kato conjectures generalising the refined leading-coefficient formula to motives (Bloch-Kato 1990 Grothendieck Festschrift), the equivariant Tamagawa number conjecture (Burns-Flach 2001 Doc. Math. 6) packaging all special-value formulas into a single statement over arbitrary number fields.

The two conjectures in this unit are the simplest substantive cases of this entire research programme. Modularity is the $GL_{2} / Q$ case of Langlands reciprocity, BSD is the elliptic-curve case of Bloch-Kato. Both build toward 21.10.01 pending Langlands programme, with the surrounding apparatus of Galois representations and automorphic forms.

Full proof set [Master]

The modularity theorem of Wiles + BCDT and the BSD conjecture are themselves statement-level results in this unit; their full proofs (Wiles + Taylor-Wiles 1995 + BCDT 2001 for modularity, open for BSD in analytic rank $\geq 2$ ) are deferred to later units developing the deformation-theoretic and Euler-system machinery. What can be proved here at Master tier is the bridge identity at primes of good reduction and the structure of the modular abelian variety attached to a newform.

Proposition 10 (Faltings isogeny theorem; Faltings 1983 Invent. Math. 73). Let $A$ and $B$ be abelian varieties over a number field $K$ . Then $A$ and $B$ are isogenous over $K$ if and only if there is an isomorphism $V_{ℓ} A ≅ V_{ℓ} B$ of $ℓ$ -adic Galois representations for some (equivalently, every) prime $ℓ$ .

Proof (sketch). The forward implication is direct: an isogeny $A \to B$ induces an isomorphism on Tate modules $V_{ℓ} A ≅ V_{ℓ} B$ compatible with the Galois action. The reverse implication is the deep content. Faltings's strategy: an isomorphism $V_{ℓ} A ≅ V_{ℓ} B$ as Galois representations induces an isomorphism on the corresponding Hodge-Tate decompositions, hence (by Tate's theorem on $p$ -adic Hodge structures) an isomorphism on the corresponding $Z_{ℓ}$ -modules of $ℓ$ -adic Tate vectors. The reconstruction of $A$ from $V_{ℓ} A$ as a Galois module proceeds via the finiteness theorem of Faltings: the set of $K$ -isogeny classes of abelian varieties of fixed dimension with bounded conductor is finite, so the isomorphism on $V_{ℓ}$ forces $A$ and $B$ into the same isogeny class. The detailed proof is a major argument and is not reproduced here. $□$

Corollary used in Theorem 1. Modularity in the sense of Galois-representation isomorphism (formulation (b) of Theorem 1) is equivalent to modularity in the sense of $L$ -function identity (formulation (a)) because Faltings 1983 identifies the abelian variety $A_{f_{E}}$ with $E$ up to isogeny, and isogenous abelian varieties have equal $L$ -functions.

Proposition 11 (Hasse bound and the local Euler factor). Let $E$ be an elliptic curve over $Q$ with good reduction at a prime $p$ . Let $α_{p}, β_{p}$ be the eigenvalues of Frobenius on $V_{ℓ} E$ for any $ℓ \neq = p$ . Then $α_{p} + β_{p} = a_{p} = p + 1 - N_{p}$ , $α_{p} β_{p} = p$ , and $∣ α_{p} ∣ = ∣ β_{p} ∣ = p$ . The local Euler factor is $$ L_p(E, s) = \frac{1}{(1 - \alpha_p p^{-s})(1 - \beta_p p^{-s})} = \frac{1}{1 - a_p p^{-s} + p^{1 - 2 s}}. $$

Proof. The Weil conjectures applied to the elliptic curve $E_{p} := E \times_{Q} Spec F_{p}$ (good reduction modulo $p$ ) state that the zeta function of $E_{p}$ has the form $$ Z(E_p, T) = \frac{(1 - \alpha_p T)(1 - \beta_p T)}{(1 - T)(1 - p T)}, $$ with $α_{p}, β_{p}$ the eigenvalues of Frobenius on $H_{et}^{1} (\overline{E_{p}}, Q_{ℓ})$ for $ℓ \neq = p$ — equivalently on $V_{ℓ} E$ via the comparison theorem identifying $H_{et}^{1}$ with the dual of $V_{ℓ}$ . The Hasse bound $∣ α_{p} ∣ = ∣ β_{p} ∣ = p$ is the Riemann hypothesis for $E_{p}$ , proved by Hasse 1933 for elliptic curves over finite fields and generalised to higher-dimensional varieties by Weil 1949 + Deligne 1974 Publ. Math. IHES 43.

The point count is $N_{p} = # E_{p} (F_{p}) = 1 + p - α_{p} - β_{p}$ (Lefschetz fixed-point formula on the étale cohomology). Hence $a_{p} = p + 1 - N_{p} = α_{p} + β_{p}$ , $α_{p} β_{p} = p$ (the determinant of Frobenius on the $2$ -dimensional $V_{ℓ} E$ equals the cyclotomic character $χ_{ℓ} (Frob_{p}) = p$ ).

The local $L$ -factor is the inverse characteristic polynomial of Frobenius at $X = p^{- s}$ : $$ L_p(E, s) = \det(1 - \mathrm{Frob}p \cdot p^{-s} \mid V\ell E)^{-1} = (1 - \alpha_p p^{-s})^{-1} (1 - \beta_p p^{-s})^{-1}. $$ Expanding the product: $1 - (α_{p} + β_{p}) p^{- s} + α_{p} β_{p} p^{- 2 s} = 1 - a_{p} p^{- s} + p^{1 - 2 s}$ . $□$

Proposition 12 (Mordell-Weil theorem; Mordell 1922, Weil 1929). For every elliptic curve $E$ over $Q$ , the group $E (Q)$ is finitely generated.

Proof (sketch). The descent argument: the multiplication-by- $2$ map $[2] : E (Q) \to E (Q)$ has finite cokernel (the weak Mordell-Weil theorem, proved via Galois cohomology: the cokernel injects into a finite Selmer group). Combined with the height pairing $\hat{h} : E (Q) \to R$ — a quadratic form on $E (Q)$ that is positive-definite modulo torsion, and finite-to-one on lattice points of bounded height — descent on $[2]$ produces a finite set of generators. Detailed proof in Silverman The Arithmetic of Elliptic Curves Ch. VIII §1-§4. $□$

The Mordell-Weil theorem gives the decomposition $E (Q) = Z^{r} \oplus E (Q)_{tors}$ . The integer $r$ is the rank of $E$ , the object BSD predicts equals $ord_{s = 1} L (E, s)$ .

Proposition 13 (Hasse-Weil conjecture from modularity). The $L$ -function $L (E, s)$ of every elliptic curve over $Q$ has an analytic continuation to $C$ and satisfies the functional equation $Λ (E, s) = ε_{E} Λ (E, 2 - s)$ with $Λ (E, s) := N_{E}^{s /2} (2 π)^{- s} Γ (s) L (E, s)$ and $ε_{E} \in {\pm 1}$ .

Proof. By modularity (Theorem 1), $L (E, s) = L (f_{E}, s)$ for a weight- $2$ cusp newform $f_{E}$ on $Γ_{0} (N_{E})$ . The $L$ -function $L (f_{E}, s)$ has the integral representation $$ \Lambda(f_E, s) := N_E^{s/2} (2 \pi)^{-s} \Gamma(s) L(f_E, s) = \int_0^\infty f_E(i y) y^{s - 1} dy, $$ where the convergence on $Re (s) > 0$ uses the cusp-form decay $f_{E} (i y) \to 0$ as $y \to \infty$ . The functional equation under the Atkin-Lehner involution $w_{N_{E}} : z \mapsto - 1/ (N_{E} z)$ gives $$ f_E(- 1/(N_E z)) = \varepsilon_E N_E z^2 f_E(z), \qquad \varepsilon_E \in {\pm 1}, $$ and substituting $z = i y$ at $y \to 1/ (N_{E} y)$ in the integral yields $Λ (f_{E}, s) = ε_{E} Λ (f_{E}, 2 - s)$ . Combining with modularity gives the same identity for $Λ (E, s)$ . $□$

Before modularity was proved (pre-Wiles), this functional equation was the Hasse-Weil conjecture and was open for general elliptic curves. The modularity theorem of Wiles + BCDT therefore proves Hasse-Weil as a corollary. The functional equation, in turn, makes BSD's vanishing-order statement well-defined: $ord_{s = 1} L (E, s)$ requires $L (E, s)$ to be defined at $s = 1$ , which it is only by analytic continuation. Without modularity, the conjectural rank $= ord_{s = 1} L (E, s)$ would not even be a meaningful statement for general $E$ .

Connections [Master]

Modular forms on $SL_{2} (Z)$ 21.04.01. The ambient analytic theory on which the modularity bridge is built. The level-one space $M_{k} (SL_{2} (Z))$ is the level- $1$ specialisation of $M_{k} (Γ_{0} (N))$ , and the modularity theorem identifies elliptic curves over $Q$ with weight- $2$ cusp newforms on $Γ_{0} (N)$ at the appropriate higher level. The dimension formula, $q$ -expansion machinery, and Hecke-action theory of the present unit's $f_{E}$ are all developed in 21.04.01 for the level- $1$ case and extend via the Atkin-Lehner machinery to higher level.
Hecke operators and Hecke algebra 21.04.02. The operator-theoretic substrate. The Hecke eigenvalue $a_{p} (f_{E})$ at a prime $p$ of good reduction is the trace of Frobenius on the $ℓ$ -adic Galois representation attached to $E$ , by the Eichler-Shimura congruence. The Hecke algebra $T$ acts on the cuspidal cohomology of $X_{0} (N_{E})$ , and the modular abelian variety $A_{f_{E}}$ is constructed as a quotient of $J_{0} (N_{E})$ by an ideal of $T$ . The Wiles-Taylor $R = T$ theorem identifies the universal deformation ring of the residual representation with the Hecke algebra at the appropriate maximal ideal — the entire deformation framework rests on the Hecke algebra of 21.04.02.
Eichler-Shimura correspondence 21.04.03. Sibling-in-flight unit treating the weight- $2$ Hecke eigenform / Galois-representation bridge at technical depth. The Eichler-Shimura construction produces the abelian variety $A_{f_{E}}$ as a quotient of $J_{0} (N_{E})$ and identifies its $ℓ$ -adic Tate module with the cohomological realisation of $f_{E}$ via étale cohomology. The local match at primes of good reduction — the key theorem proved in the present unit — is the cornerstone of the bridge from elliptic curves to modular forms.
$ℓ$ -adic Galois representations 21.05.01. Sibling-in-flight unit developing the Galois representations attached to weight- $\geq 2$ cusp eigenforms (Deligne 1971 + Deligne-Serre 1974). The modularity theorem is equivalent to a statement about $ℓ$ -adic Galois representations: $ρ_{E, ℓ}$ comes from a modular form. The deformation-theoretic proof of Wiles works entirely at the level of Galois representations — the Hecke algebra is the modular-form-side and the deformation ring is the Galois-representation-side, and $R = T$ is the bridge.
Riemann zeta function $ζ (s)$ 21.03.01. Prototype of an $L$ -function with analytic continuation, functional equation, and Euler product. The modular $L$ -function $L (E, s) = L (f_{E}, s)$ inherits the same three properties from the modular-form side via modularity, and the conjectural special-value formula at $s = 1$ — BSD's leading-coefficient formula — is the elliptic-curve analogue of the class number formula relating $ζ_{K} (0)$ to the class number and regulator of a number field $K$ .
Dirichlet $L$ -functions $L (s, χ)$ 21.03.02. Prototype of an $L$ -function attached to a $1$ -dimensional Galois character. The modular $L$ -function $L (E, s)$ generalises Dirichlet to the $2$ -dimensional case: $E$ corresponds to a $2$ -dimensional Galois representation $ρ_{E, ℓ}$ as $χ$ corresponds to a $1$ -dimensional character, and the modular form $f_{E}$ is the $GL_{2}$ -analogue of the Dirichlet character $χ$ . The Bloch-Kato refined conjecture for $L (s, χ)$ at $s = 0$ (the class number formula refined) is the prototype of the BSD refined formula at $s = 1$ in the elliptic-curve case.
Dedekind / Hecke / Artin $L$ -functions 21.03.03. Sibling unit on the higher-rank Hecke-Artin $L$ -function framework. The Hasse-Weil $L$ -function $L (E, s)$ is the Artin $L$ -function of the $2$ -dimensional Galois representation $ρ_{E, ℓ}$ analytically continued to a complex $L$ -function; modularity is the assertion that this Artin $L$ -function equals the Hecke $L$ -function of the cusp newform $f_{E}$ via the Eichler-Shimura bridge. The BSD leading-coefficient formula at $s = 1$ is the elliptic-curve refinement of the Dedekind analytic class-number formula at $ζ_{K} (s) \to s = 1$ , with rank, regulator, Sha, Tamagawa numbers replacing $(h_{K}, R_{K}, w_{K}, ∣ d_{K} ∣)$ .
Iwasawa $Z_{p}$ -extensions 21.07.01. Sibling-in-flight unit developing the Iwasawa theory of cyclotomic $Z_{p}$ -extensions and the structure of the Selmer group $Sel (E / Q_{\infty})$ over the cyclotomic tower. Kato's Euler system of Beilinson elements (Theorem 7) bounds the Selmer group above by the $p$ -adic $L$ -function, proving one inclusion of the Iwasawa main conjecture for elliptic curves; Skinner-Urban (Theorem 8) supplies the reverse inclusion in many cases, yielding the $p$ -part of BSD in analytic rank $\leq 1$ .
$p$ -adic $L$ -functions and Mazur-Wiles Main Conjecture 21.07.02. Sibling-in-flight unit developing the $p$ -adic $L$ -function of Mazur-Swinnerton-Dyer $L_{p} (E, T) \in Z_{p} [[T]]$ and the Iwasawa main conjecture for elliptic curves. The main conjecture states $char_{Z_{p} [[T]]} Sel (E / Q_{\infty})^{\lor} = (L_{p} (E, T))$ , proved by Kato + Skinner-Urban in many cases. The main conjecture is the cyclotomic- $Z_{p}$ -extension Iwasawa-theoretic refinement of BSD, packaging the $p$ -part of the leading-coefficient formula into a statement about characteristic ideals.
Sato-Tate conjecture 21.06.02 pending. Successor unit (to be produced) on the equidistribution of Frobenius angles $θ_{p}$ in the Sato-Tate measure, proved by Clozel-Harris-Shepherd-Barron-Taylor 2008-2011 as a corollary of modularity + automorphy of symmetric powers. Sato-Tate is the quantitative refinement of the Hasse bound and the natural sharpening of modularity.
Elliptic curves 04.04.03. The algebraic-geometric foundation unit on elliptic curves as smooth projective curves of genus $1$ with a marked rational point. The Mordell-Weil group $E (Q)$ , the Weierstrass form, the group law, and the discriminant are developed there; the present unit's modularity / BSD statements rest on those foundations. The bridge from 04.04.03 to the present unit is through the $L$ -function: the algebraic-geometric object $E$ acquires arithmetic structure through its point counts $N_{p}$ , which assemble into the analytic $L$ -function $L (E, s)$ , which by modularity equals a modular-form $L$ -function.
Langlands programme 21.10.01 pending. Future successor unit on the unifying frame. Modularity is the $GL_{2} / Q$ case of Langlands reciprocity: every $ℓ$ -adic Galois representation of geometric origin should correspond to an automorphic representation, with matching $L$ -functions. BSD is the $GL_{2} / Q$ case of Bloch-Kato's refined special-value conjecture for motives. The two conjectures in the present unit are the simplest substantive cases of the entire Langlands research programme.

Historical & philosophical context [Master]

Bryan Birch and Peter Swinnerton-Dyer formulated the conjecture bearing their names in a pair of 1963-65 papers on the EDSAC-2 computer at the University of Cambridge ^{[BirchSwinnertonDyer1965]} in Journal für die reine und angewandte Mathematik 218. The original conjecture was formulated as an asymptotic prediction for the product $\prod_{p \leq X} (N_{p} / p)$ over primes $p$ of good reduction, with the prediction that this product should grow like $C (lo g X)^{r}$ as $X \to \infty$ , with $r$ equal to the rank of $E (Q)$ . The computational evidence was based on tables of Mordell-Weil ranks for elliptic curves of small conductor, computed using descent algorithms and Heegner-point constructions. Tate 1974 Invent. Math. 23 reformulated the conjecture in its modern leading-coefficient form ^[Tate1974], identifying the constant $C$ explicitly with the product of arithmetic invariants — real period, regulator, Tate-Shafarevich group, Tamagawa numbers, torsion — that appears in the refined statement. Tate's reformulation made BSD a precise predictive equation rather than a vague asymptotic, and remains the standard formulation in 2026.

The modularity theorem has its origins in two parallel threads. On the modular-form side, Hecke 1936-37 Math. Ann. 112 + 114 introduced the operators and proved the Euler-product structure of modular $L$ -functions; Eichler 1954 Arch. Math. 5 + Shimura 1958 Tohoku Math. J. 10 realised weight- $2$ Hecke eigenforms on the cohomology of modular curves $X_{0} (N)$ , identifying Hecke eigenvalues with Frobenius traces. Shimura 1971 Introduction to the Arithmetic Theory of Automorphic Functions codified the modular-abelian-variety construction. On the elliptic-curve side, Weil 1967 Math. Ann. 168 proved the converse theorem: an elliptic curve over $Q$ is modular if and only if its $L$ -function has the expected analytic continuation and functional equation. Taniyama 1955 had conjectured something close to modularity at the 1955 Tokyo-Nikko symposium; Shimura developed the precise statement through the 1960s. The combined Taniyama-Shimura-Weil conjecture — every elliptic curve over $Q$ is modular — was the standard formulation by 1970.

Frey 1985 ^[Frey1986] observed in a Saarbrücken preprint (published 1986 Annales Universitatis Saraviensis 1) that a hypothetical Fermat solution $a^{ℓ} + b^{ℓ} = c^{ℓ}$ would construct an elliptic curve $E_{a, b, c} : y^{2} = x (x - a^{ℓ}) (x + b^{ℓ})$ with a mod- $ℓ$ Galois representation of impossibly small level — incompatible with modularity at any positive level. Serre 1987 Duke Math. J. 54 ^[Serre1987] formulated the level-lowering $ε$ -conjecture making Frey's observation precise: the mod- $ℓ$ representation of the Frey curve should come from a newform of level $2$ , but $S_{2} (Γ_{0} (2)) = 0$ . Ribet 1990 Invent. Math. 100 ^[Ribet1990] proved the $ε$ -conjecture, closing the implication chain Taniyama-Shimura-Weil ⇒ Fermat's Last Theorem.

Andrew Wiles announced his proof of modularity for semistable elliptic curves over $Q$ in three lectures at the Isaac Newton Institute in Cambridge in June 1993. A gap was identified in the Euler-system argument later that year; Wiles and Richard Taylor closed the gap in 1994 via the Taylor-Wiles patching argument, and the two papers Wiles 1995 Ann. Math. 141 ^[Wiles1995] and Taylor-Wiles 1995 Ann. Math. 141 ^{[TaylorWiles1995]} were published together in Annals of Mathematics 141 in 1995. The Wiles paper develops the deformation-theoretic framework and the $R = T$ identification; the Taylor-Wiles companion paper supplies the patching argument and the numerical criterion. The full modularity theorem for every elliptic curve over $Q$ , including additive-reduction cases, was completed by Breuil-Conrad-Diamond-Taylor 2001 J. AMS 14 ^[BCDT2001], removing the semistability hypothesis via wild $3$ -adic deformation theory.

The partial results towards BSD form a separate fifty-year lineage. Coates-Wiles 1977 Invent. Math. 39 ^{[CoatesWiles1977]} proved the rank- $0$ implication for CM elliptic curves via the elliptic-unit Euler system. Gross-Zagier 1986 Invent. Math. 84 ^{[GrossZagier1986]} supplied the analytic-rank- $1$ lower bound via the Néron-Tate height of Heegner points. Kolyvagin 1988-89 Math. USSR-Izv. 32 ^{[Kolyvagin1989]} supplied the analytic-rank $\leq 1$ upper bound via the Kolyvagin Euler system. Kato 2004 Astérisque 295 ^[Kato2004] and Skinner-Urban 2014 Invent. Math. 195 ^{[SkinnerUrban2014]} proved the Iwasawa main conjecture for modular elliptic curves in many cases, implying the $p$ -part of the refined BSD formula in analytic rank $\leq 1$ . The Bhargava-Skinner-Zhang programme of the 2010s proved BSD holds in a positive proportion of elliptic curves over $Q$ when ordered by naive height, the first density result for the conjecture. BSD remains open for individual elliptic curves of analytic rank $\geq 2$ and is one of the seven Clay Millennium Open Problems announced in 2000, with the original $1,000,000 prize unclaimed as of 2026.

Manin-Panchishkin Introduction to Modern Number Theory (Springer EMS 49, 2nd ed. 2005) Ch. 6 ^{[ManinPanchishkin2005]} codifies the modularity / BSD synthesis for the modern student. Silverman The Arithmetic of Elliptic Curves (GTM 106, 2nd ed. 2009) ^{[Silverman2009]} provides the canonical textbook treatment of elliptic curves, with Ch. C.16 and Appendix C §16 surveying modularity and BSD at survey level. Diamond-Shurman A First Course in Modular Forms (GTM 228, 2005) ^{[DiamondShurman2005]} gives the modular-forms-side exposition of the Eichler-Shimura construction and its role in modularity.

Bibliography [Master]

@article{Wiles1995,
  author  = {Wiles, Andrew},
  title   = {Modular elliptic curves and {F}ermat's last theorem},
  journal = {Annals of Mathematics},
  volume  = {141},
  number  = {3},
  year    = {1995},
  pages   = {443--551}
}

@article{TaylorWiles1995,
  author  = {Taylor, Richard and Wiles, Andrew},
  title   = {Ring-theoretic properties of certain {H}ecke algebras},
  journal = {Annals of Mathematics},
  volume  = {141},
  number  = {3},
  year    = {1995},
  pages   = {553--572}
}

@article{BCDT2001,
  author  = {Breuil, Christophe and Conrad, Brian and Diamond, Fred and Taylor, Richard},
  title   = {On the modularity of elliptic curves over {$\mathbb{Q}$}: wild {$3$}-adic exercises},
  journal = {Journal of the American Mathematical Society},
  volume  = {14},
  number  = {4},
  year    = {2001},
  pages   = {843--939}
}

@article{BirchSwinnertonDyer1965,
  author  = {Birch, Bryan J. and Swinnerton-Dyer, Peter},
  title   = {Notes on elliptic curves. {II}},
  journal = {Journal f{\"u}r die reine und angewandte Mathematik},
  volume  = {218},
  year    = {1965},
  pages   = {79--108}
}

@article{Frey1986,
  author  = {Frey, Gerhard},
  title   = {Links between stable elliptic curves and certain diophantine equations},
  journal = {Annales Universitatis Saraviensis, Mathematische Schriften},
  volume  = {1},
  year    = {1986},
  pages   = {1--40}
}

@article{Ribet1990,
  author  = {Ribet, Kenneth A.},
  title   = {On modular representations of {$\mathrm{Gal}(\overline{\mathbb{Q}}/\mathbb{Q})$} arising from modular forms},
  journal = {Inventiones Mathematicae},
  volume  = {100},
  number  = {2},
  year    = {1990},
  pages   = {431--476}
}

@article{Serre1987,
  author  = {Serre, Jean-Pierre},
  title   = {Sur les repr{\'e}sentations modulaires de degr{\'e} {$2$} de {$\mathrm{Gal}(\overline{\mathbb{Q}}/\mathbb{Q})$}},
  journal = {Duke Mathematical Journal},
  volume  = {54},
  number  = {1},
  year    = {1987},
  pages   = {179--230}
}

@article{CoatesWiles1977,
  author  = {Coates, John and Wiles, Andrew},
  title   = {On the conjecture of {B}irch and {S}winnerton-{D}yer},
  journal = {Inventiones Mathematicae},
  volume  = {39},
  number  = {3},
  year    = {1977},
  pages   = {223--251}
}

@article{GrossZagier1986,
  author  = {Gross, Benedict H. and Zagier, Don B.},
  title   = {{H}eegner points and derivatives of {$L$}-series},
  journal = {Inventiones Mathematicae},
  volume  = {84},
  number  = {2},
  year    = {1986},
  pages   = {225--320}
}

@article{Kolyvagin1989,
  author  = {Kolyvagin, Victor A.},
  title   = {Finiteness of {$E(\mathbb{Q})$} and {$\mathrm{Sha}(E, \mathbb{Q})$} for a subclass of {W}eil curves},
  journal = {Mathematics of the USSR-Izvestiya},
  volume  = {32},
  number  = {3},
  year    = {1989},
  pages   = {523--541}
}

@article{Kato2004,
  author  = {Kato, Kazuya},
  title   = {{$p$}-adic {H}odge theory and values of zeta functions of modular forms},
  journal = {Ast{\'e}risque},
  volume  = {295},
  year    = {2004},
  pages   = {117--290}
}

@article{SkinnerUrban2014,
  author  = {Skinner, Christopher and Urban, Eric},
  title   = {The {I}wasawa main conjectures for {$\mathrm{GL}_2$}},
  journal = {Inventiones Mathematicae},
  volume  = {195},
  number  = {1},
  year    = {2014},
  pages   = {1--277}
}

@article{Tate1974,
  author  = {Tate, John},
  title   = {The arithmetic of elliptic curves},
  journal = {Inventiones Mathematicae},
  volume  = {23},
  number  = {3-4},
  year    = {1974},
  pages   = {179--206}
}

@book{Silverman2009,
  author    = {Silverman, Joseph H.},
  title     = {The Arithmetic of Elliptic Curves},
  series    = {Graduate Texts in Mathematics},
  volume    = {106},
  edition   = {2nd},
  publisher = {Springer},
  year      = {2009}
}

@book{ManinPanchishkin2005,
  author    = {Manin, Yuri I. and Panchishkin, Alexei A.},
  title     = {Introduction to Modern Number Theory},
  series    = {Encyclopaedia of Mathematical Sciences},
  volume    = {49},
  edition   = {2nd},
  publisher = {Springer},
  year      = {2005}
}

@book{DiamondShurman2005,
  author    = {Diamond, Fred and Shurman, Jerry},
  title     = {A First Course in Modular Forms},
  series    = {Graduate Texts in Mathematics},
  volume    = {228},
  publisher = {Springer},
  year      = {2005}
}

@article{Mazur1977,
  author  = {Mazur, Barry},
  title   = {Modular curves and the {E}isenstein ideal},
  journal = {Publications Math{\'e}matiques de l'IH{\'E}S},
  volume  = {47},
  year    = {1977},
  pages   = {33--186}
}

@article{Faltings1983,
  author  = {Faltings, Gerd},
  title   = {Endlichkeitss{\"a}tze f{\"u}r abelsche {V}ariet{\"a}ten {\"u}ber {Z}ahlk{\"o}rpern},
  journal = {Inventiones Mathematicae},
  volume  = {73},
  number  = {3},
  year    = {1983},
  pages   = {349--366}
}

@article{LanglandsTunnell1980,
  author  = {Tunnell, Jerrold},
  title   = {Artin's conjecture for representations of octahedral type},
  journal = {Bulletin of the American Mathematical Society},
  volume  = {5},
  number  = {2},
  year    = {1981},
  pages   = {173--175}
}

Prerequisites

04.04.03
21.03.01
21.03.02
21.04.01
21.04.02

Tier anchors

beginner: Manin-Panchishkin *Introduction to Modern Number Theory* (Springer EMS 49, 2nd ed. 2005) Ch. 6 §5 informal opening — every elliptic curve over the rationals is modular, and the rank of the rational points is conjecturally read off from the $L$-function
intermediate: Manin-Panchishkin *Introduction to Modern Number Theory* (Springer EMS 49, 2nd ed. 2005) Ch. 6 §5-§6; Silverman *The Arithmetic of Elliptic Curves* (GTM 106, 2nd ed. 2009) Ch. C.16 (modularity statement) and App. C §16 (BSD); Diamond-Shurman *A First Course in Modular Forms* (GTM 228, 2005) Ch. 9
master: Wiles 1995 *Annals of Mathematics* 141 (2), 443-551 (originator — modularity of semistable elliptic curves over $\mathbb{Q}$); Taylor-Wiles 1995 *Annals of Mathematics* 141 (2), 553-572 (companion paper — Hecke algebra patching, the numerical criterion); Breuil-Conrad-Diamond-Taylor 2001 *Journal of the American Mathematical Society* 14 (4), 843-939 (full modularity for every elliptic curve over $\mathbb{Q}$); Birch-Swinnerton-Dyer 1965 *Journal für die reine und angewandte Mathematik* 218, 79-108 (BSD originator, the rank-$L$ vanishing-order conjecture with the leading-coefficient refinement); Frey 1986 *Annales Universitatis Saraviensis* 1, 1-40 (Frey curve idea); Ribet 1990 *Inventiones Mathematicae* 100, 431-476 (level-lowering, Serre's $\varepsilon$-conjecture $\Rightarrow$ Fermat); Serre 1987 *Duke Mathematical Journal* 54, 179-230 (Serre's conjecture on mod-$\ell$ Galois representations); Coates-Wiles 1977 *Inventiones Mathematicae* 39, 223-251 (CM rank-$0$ implication); Gross-Zagier 1986 *Inventiones Mathematicae* 84, 225-320 (heights of Heegner points and $L'(E, 1)$); Kolyvagin 1989 *Mathematics of the USSR-Izvestiya* 32, 523-541 (Euler systems and BSD rank $0$ and $1$); Kato 2004 *Astérisque* 295 (Euler system of Beilinson elements, Iwasawa main conjecture for elliptic curves); Skinner-Urban 2014 *Inventiones Mathematicae* 195, 1-277 (full main conjecture for many cases); Tate 1974 *Inventiones Mathematicae* 23, 179-206 (conjectural BSD framework and refined leading-coefficient formula); Silverman *The Arithmetic of Elliptic Curves* (GTM 106, 2nd ed. 2009) Ch. C (modern anchor)

References

TODO_REF
Wiles, A. — Modular elliptic curves and Fermat's last theorem · *Annals of Mathematics* (2) 141 (3), 443-551 (1995). The originator paper on the modularity theorem for semistable elliptic curves over $\mathbb{Q}$; introduces the Galois-deformation framework, the Hecke algebra side of the $R = \mathbb{T}$ identification, and the strategy of bootstrapping from residual modularity at $\ell = 3$ via a $3$-$5$ switch.
TODO_REF
Taylor, R. and Wiles, A. — Ring-theoretic properties of certain Hecke algebras · *Annals of Mathematics* (2) 141 (3), 553-572 (1995). The companion paper completing the proof; the Taylor-Wiles patching argument and the numerical criterion for $R = \mathbb{T}$ via auxiliary primes.
TODO_REF
Breuil, C., Conrad, B., Diamond, F. and Taylor, R. — On the modularity of elliptic curves over $\mathbb{Q}$: wild $3$-adic exercises · *Journal of the American Mathematical Society* 14 (4), 843-939 (2001). The full modularity theorem: every elliptic curve over $\mathbb{Q}$ is modular, removing the semistability hypothesis from Wiles-Taylor-Wiles.
TODO_REF
Birch, B. J. and Swinnerton-Dyer, H. P. F. — Notes on elliptic curves. II · *Journal für die reine und angewandte Mathematik* 218, 79-108 (1965). The originator paper introducing the conjecture, based on Cambridge EDSAC-2 computer calculations of $\prod_{p \leq X}(N_p / p)$ where $N_p$ is the number of $\mathbb{F}_p$-points on $E$; conjectures the asymptotic $\prod_{p \leq X}(N_p / p) \sim C (\log X)^r$ as $X \to \infty$, with $r = \mathrm{rank}\, E(\mathbb{Q})$.
TODO_REF
Frey, G. — Links between stable elliptic curves and certain diophantine equations · *Annales Universitatis Saraviensis, Mathematische Schriften* 1, 1-40 (1986). The Frey-curve construction: from a hypothetical Fermat solution $a^\ell + b^\ell = c^\ell$ build the elliptic curve $E_{a, b, c}: y^2 = x(x - a^\ell)(x + b^\ell)$ and show its mod-$\ell$ Galois representation has level $1$ — incompatible with modularity at any positive level.
TODO_REF
Ribet, K. A. — On modular representations of $\mathrm{Gal}(\overline{\mathbb{Q}}/\mathbb{Q})$ arising from modular forms · *Inventiones Mathematicae* 100 (2), 431-476 (1990). Ribet's level-lowering theorem: Serre's $\varepsilon$-conjecture for mod-$\ell$ Galois representations is proved, closing the implication modularity-of-Frey-curve $\Rightarrow$ Fermat's Last Theorem.
TODO_REF
Serre, J.-P. — Sur les représentations modulaires de degré $2$ de $\mathrm{Gal}(\overline{\mathbb{Q}}/\mathbb{Q})$ · *Duke Mathematical Journal* 54 (1), 179-230 (1987). Serre's conjecture on mod-$\ell$ Galois representations: every continuous odd irreducible $2$-dimensional mod-$\ell$ Galois representation arises from a modular form of specified weight, level, and character; the $\varepsilon$-conjecture is the level-lowering part.
TODO_REF
Coates, J. and Wiles, A. — On the conjecture of Birch and Swinnerton-Dyer · *Inventiones Mathematicae* 39 (3), 223-251 (1977). First substantial result towards BSD: for elliptic curves with complex multiplication, $L(E, 1) \neq 0$ implies $E(\mathbb{Q})$ is finite (rank-$0$ implication for CM curves), via the elliptic-unit Euler system of Kubota-Leopoldt.
TODO_REF
Gross, B. H. and Zagier, D. B. — Heegner points and derivatives of $L$-series · *Inventiones Mathematicae* 84 (2), 225-320 (1986). The Gross-Zagier formula: for a weight-$2$ newform $f$ on $\Gamma_0(N)$ and a Heegner point $y_K$ on $X_0(N)$ associated to an imaginary quadratic field $K$, $\hat h(y_K) = c \cdot L'(E, 1) \cdot L(E^K, 1)$ where $\hat h$ is the Néron-Tate canonical height; non-torsion Heegner points exist whenever $L'(E, 1) \neq 0$.
TODO_REF
Kolyvagin, V. A. — Finiteness of $E(\mathbb{Q})$ and $\mathrm{Sha}(E, \mathbb{Q})$ for a subclass of Weil curves · *Mathematics of the USSR-Izvestiya* 32 (3), 523-541 (1989) (Russian original 1988). The Kolyvagin Euler system of Heegner points: if $E/\mathbb{Q}$ is modular and $\mathrm{ord}_{s=1} L(E, s) \leq 1$, then BSD's rank equality holds and $\mathrm{Sha}(E/\mathbb{Q})$ is finite.
TODO_REF
Kato, K. — $p$-adic Hodge theory and values of zeta functions of modular forms · *Astérisque* 295, 117-290 (2004). The Kato Euler system of Beilinson elements in $K$-theory of modular curves; bounds the Selmer group above by $L$-values and proves one inclusion of the Iwasawa main conjecture for modular elliptic curves over the cyclotomic $\mathbb{Z}_p$-extension of $\mathbb{Q}$.
TODO_REF
Skinner, C. and Urban, E. — The Iwasawa main conjectures for $\mathrm{GL}_2$ · *Inventiones Mathematicae* 195 (1), 1-277 (2014). The reverse inclusion of the Iwasawa main conjecture in the $\mathrm{GL}_2$ setting, completing the main conjecture for a large class of elliptic curves; combined with Kato 2004 gives the full main conjecture, which implies the $p$-part of BSD in analytic rank $\leq 1$ for these curves.
TODO_REF
Tate, J. — The arithmetic of elliptic curves · *Inventiones Mathematicae* 23 (3-4), 179-206 (1974). Tate's exposition of BSD with the refined leading-coefficient formula $L^{(r)}(E, 1)/r! = \Omega_E \cdot R_E \cdot \#\mathrm{Sha}(E/\mathbb{Q}) \cdot \prod_p c_p / (\#E(\mathbb{Q})_\mathrm{tors})^2$; the conjectural framework with all invariants explicit.
TODO_REF
Silverman, J. H. — The Arithmetic of Elliptic Curves · Graduate Texts in Mathematics 106, 2nd ed., Springer (2009). The canonical modern textbook on elliptic curves; Ch. C.16 surveys the modularity theorem and BSD with citations to the primary literature; Appendix C §16 gives the refined BSD formula and the current state of knowledge.
TODO_REF
Manin, Yu. I. and Panchishkin, A. A. — Introduction to Modern Number Theory · Encyclopaedia of Mathematical Sciences 49, 2nd ed., Springer (2005). Ch. 6 §5 (modularity theorem statement, Wiles 1995 + BCDT 2001), §6 (BSD conjecture and refined leading-coefficient formula, current state).
TODO_REF
Diamond, F. and Shurman, J. — A First Course in Modular Forms · Graduate Texts in Mathematics 228, Springer (2005). Ch. 9 (Galois representations and modularity); the canonical introductory exposition of the modularity bridge from a modular-forms viewpoint.
TODO_REF
Mazur, B. — Modular curves and the Eisenstein ideal · *Publications Mathématiques de l'IHÉS* 47, 33-186 (1977). The classification of rational torsion on elliptic curves over $\mathbb{Q}$ (Mazur's torsion theorem: $E(\mathbb{Q})_\mathrm{tors}$ is one of fifteen explicit groups); foundational background for the BSD invariant $\#E(\mathbb{Q})_\mathrm{tors}$.

Lean module

Codex.NumberTheory.Modularity.ModularityBSD

Mathlib gap

Mathlib at present supplies a developing infrastructure for
elliptic curves over a field via
`Mathlib.AlgebraicGeometry.EllipticCurve.Weierstrass` (the type
`WeierstrassCurve R` with discriminant and $j$-invariant) and
partial support for the group law on `EllipticCurve.Point`, but
the full arithmetic apparatus required to state the modularity
theorem and the Birch-Swinnerton-Dyer conjecture is absent.
Specifically, Mathlib lacks (i) the **conductor** $N_E \in \mathbb{N}$
of an elliptic curve over $\mathbb{Q}$ as a definition tying the
reduction type at each prime to a local exponent via the Ogg-Saito
formula; (ii) the **$L$-function** $L(E, s) = \prod_p L_p(E, s)$
with $L_p(E, s) = (1 - a_p p^{-s} + p^{1 - 2 s})^{-1}$ for primes
of good reduction and the appropriate local factor at bad primes;
(iii) the **modularity theorem statement**
`theorem modularity_theorem (E : EllipticCurve ℚ) : ∃ f :
CuspNewform 2 (Γ₀ (conductor E)), ∀ p, L_p_factor E p =
L_p_factor_of_form f p` and the associated bridge
$L(E, s) = L(f_E, s)$ — provable in Mathlib only once weight-$2$
newforms with $\mathbb{Q}$-rational Hecke eigenvalues, the local
$L$-factor identification at primes of good reduction (Eichler-Shimura
for $E$ and $A_f$), and the Galois-representation comparison at
bad primes are all formalised, none of which exist today;
(iv) the **Mordell-Weil theorem** $E(\mathbb{Q}) = \mathbb{Z}^r \oplus
T$ as a structural statement on the type `EllipticCurve.Point E`
with `r = rank E ℚ` and `T = torsion subgroup` — Mazur's torsion
classification at the torsion side, descent at the rank side;
(v) the **Tate-Shafarevich group**
`def shaftaGroup (E : EllipticCurve ℚ) : Type` realised as
$\mathrm{Sha}(E/\mathbb{Q}) = \ker(H^1(\mathrm{Gal}(\overline{\mathbb{Q}}/\mathbb{Q}),
E(\overline{\mathbb{Q}})) \to \prod_v H^1(G_v, E(\overline{\mathbb{Q}_v})))$
via Galois cohomology and local-global duality; (vi) the
**Néron-Tate height pairing** $\hat h : E(\mathbb{Q}) \times
E(\mathbb{Q}) \to \mathbb{R}$ and the **regulator** $R_E := \det(\hat
h(P_i, P_j))$ on a basis of $E(\mathbb{Q})/E(\mathbb{Q})_\mathrm{tors}$;
(vii) the **real period** $\Omega_E := \int_{E(\mathbb{R})} |\omega|$
for the Néron differential $\omega$; (viii) the **Tamagawa numbers**
$c_p := [E(\mathbb{Q}_p) : E^0(\mathbb{Q}_p)]$ via Kodaira's local
classification; (ix) the **BSD conjecture as a structure**
`structure BirchSwinnertonDyerConjecture (E : EllipticCurve ℚ)`
packaging the rank statement $\mathrm{ord}_{s=1} L(E, s) =
\mathrm{rank}\, E(\mathbb{Q})$ together with the refined
leading-coefficient formula
$\lim_{s \to 1}(s - 1)^{-r} L(E, s) = \Omega_E R_E \#\mathrm{Sha}(E/\mathbb{Q})
\prod_p c_p / (\#E(\mathbb{Q})_\mathrm{tors})^2$. The companion file
records these as `sorry`-stubbed declarations on the developing
Mathlib elliptic-curve type, with the proofs awaiting most of
modern Diophantine geometry. The aggregated gap is the single
most ambitious formalisation target in the arithmetic-geometry
corner of the Mathlib roadmap.

Reviewer

TBD

Estimated time

beginner: 25m
intermediate: 60m
master: 110m