riemann

The Lean code claims a proof of the Riemann Hypothesis. This site has not checked the claim yet. See what is checked.

Elliptic curves and complex multiplication

Lattices in ℂ\mathbb{C} and the Weierstrass function, elliptic curves with their reduction, twists and periods, and complex multiplication, with the Chowla-Selberg formula and the CM curves of Gross. The last sections are on abelian varieties and abelian schemes.

Prerequisites: Automorphic forms, Class field theory.

An elliptic curve over ℂ\mathbb{C} is a quotient ℂ/L\mathbb{C}/L by a lattice, and also a plane cubic. When the lattice is an ideal of an imaginary quadratic field, the curve has complex multiplication and its invariants generate class fields. The CM divisibility chapter uses a CM curve with good reduction outside one prime, the Chowla-Selberg formula and Eisenstein sums of lattices; abelian varieties, the higher-dimensional case, appear in the window bound and the Weil bound.

Lattices and the Weierstrass function

Definition. A lattice in ℂ\mathbb{C} is a subgroup L=ℤω1+ℤω2L = \mathbb{Z}\omega_1 + \mathbb{Z}\omega_2 with ω1,ω2\omega_1, \omega_2 linearly independent over ℝ\mathbb{R}. Lattices L,L′L, L' are homothetic if L′=cLL' = cL for some c∈ℂ×c \in \mathbb{C}^\times. Each lattice is homothetic to Lτ=ℤ+ℤτL_\tau = \mathbb{Z} + \mathbb{Z}\tau with τ∈ℍ\tau \in \mathbb{H}, and Lτ,Lτ′L_\tau, L_{\tau'} are homothetic exactly when τ′=γτ\tau' = \gamma\tau for some γ∈SL2(ℤ)\gamma \in \mathrm{SL}_2(\mathbb{Z}).

The quotient ℂ/L\mathbb{C}/L, a compact Riemann surface of genus 11 and an abelian group, is a complex torus of dimension 11. An elliptic function for LL is a meromorphic function ff on ℂ\mathbb{C} with f(z+ω)=f(z)f(z + \omega) = f(z) for each ω∈L\omega \in L.

Definition. For an integer k≥3k \geq 3 the Eisenstein sum of LL is Gk(L)=∑ω∈L∖{0}ω−kG_k(L) = \sum_{\omega \in L \setminus \{0\}} \omega^{-k}. It converges absolutely and is 00 for odd kk. Put g2(L)=60 G4(L)g_2(L) = 60\, G_4(L) and g3(L)=140 G6(L)g_3(L) = 140\, G_6(L).

Definition. The Weierstrass function of LL is

℘(z)=1z2+∑ω∈L∖{0}(1(z−ω)2−1ω2).\wp(z) = \frac{1}{z^2} + \sum_{\omega \in L \setminus \{0\}} \left( \frac{1}{(z - \omega)^2} - \frac{1}{\omega^2} \right).

It is an even elliptic function with a double pole at each point of LL and no other pole. The field of elliptic functions for LL is ℂ(℘,℘′)\mathbb{C}(\wp, \wp').

Proposition. ℘′(z)2=4℘(z)3−g2(L)℘(z)−g3(L)\wp'(z)^2 = 4\wp(z)^3 - g_2(L)\wp(z) - g_3(L), and the discriminant Δ(L)=g2(L)3−27g3(L)2\Delta(L) = g_2(L)^3 - 27 g_3(L)^2 is not 00.

Scaling gives g2(cL)=c−4g2(L)g_2(cL) = c^{-4} g_2(L), g3(cL)=c−6g3(L)g_3(cL) = c^{-6} g_3(L) and Δ(cL)=c−12Δ(L)\Delta(cL) = c^{-12}\Delta(L). The function τ↦Δ(Lτ)\tau \mapsto \Delta(L_\tau) is the cusp form (2π)12η(τ)24(2\pi)^{12}\eta(\tau)^{24} of weight 1212, with η\eta the Dedekind eta function of the automorphic forms chapter.

Elliptic curves and the j-invariant

Definition. An elliptic curve over a field KK is a smooth projective curve EE over KK of genus 11 with a point O∈E(K)O \in E(K). If the characteristic of KK is not 22 or 33, EE has a Weierstrass equation y2=x3+ax+by^2 = x^3 + ax + b with a,b∈Ka, b \in K and discriminant Δ=−16(4a3+27b2)≠0\Delta = -16(4a^3 + 27b^2) \neq 0, and OO is the point at infinity.

The points E(K)E(K) form an abelian group with neutral element OO, in which three points add to OO when they lie on a line, counted with multiplicity.

Definition. The j-invariant of y2=x3+ax+by^2 = x^3 + ax + b is j=1728⋅4a3/(4a3+27b2)j = 1728 \cdot 4a^3 / (4a^3 + 27b^2). For a lattice, j(L)=1728 g2(L)3/Δ(L)j(L) = 1728\, g_2(L)^3/\Delta(L).

Over an algebraically closed field, two elliptic curves are isomorphic exactly when they have the same jj-invariant.

Theorem (uniformization). For each lattice LL, the map z↦(℘(z),℘′(z)/2)z \mapsto (\wp(z), \wp'(z)/2) is an isomorphism of complex Lie groups from ℂ/L\mathbb{C}/L onto EL:y2=x3−g2(L)4x−g3(L)4E_L : y^2 = x^3 - \tfrac{g_2(L)}{4} x - \tfrac{g_3(L)}{4}, whose discriminant is Δ(L)\Delta(L). Each elliptic curve over ℂ\mathbb{C} is ELE_L for a lattice LL unique up to homothety, and j(EL)=j(L)j(E_L) = j(L).

The function j(τ)=j(Lτ)j(\tau) = j(L_\tau) is a modular function for SL2(ℤ)\mathrm{SL}_2(\mathbb{Z}) with qq-expansion q−1+744+196884q+⋯q^{-1} + 744 + 196884q + \cdots. The proof starts from the model ELE_L in the statement CMCurveGoodOutsidePrimeSt of the CM divisibility chapter.

Reduction and minimal models

Let KK be a number field, vv a finite place of KK with valuation ring 𝒪v\mathcal{O}_v, and EE an elliptic curve over KK.

Definition. A Weierstrass model of EE at vv is a general Weierstrass equation y2+a1xy+a3y=x3+a2x2+a4x+a6y^2 + a_1 xy + a_3 y = x^3 + a_2 x^2 + a_4 x + a_6 for EE with all ai∈𝒪va_i \in \mathcal{O}_v. It is minimal if v(Δ)v(\Delta) is least among such models. The minimal discriminant of EE is the ideal ∏v𝔭vv(Δv)\prod_v \mathfrak{p}_v^{v(\Delta_v)}, with Δv\Delta_v the discriminant of a minimal model at vv.

The change of variables x=u2x′+rx = u^2 x' + r, y=u3y′+su2x′+ty = u^3 y' + su^2 x' + t changes Δ\Delta to u−12Δu^{-12}\Delta. A minimal model exists at each vv, unique up to such a change with u∈𝒪v×u \in \mathcal{O}_v^\times and r,s,t∈𝒪vr, s, t \in \mathcal{O}_v.

Definition. EE has good reduction at vv if a minimal model reduces modulo vv to a smooth curve, that is, if v(Δv)=0v(\Delta_v) = 0, and bad reduction otherwise. Bad reduction is multiplicative if the reduced curve has a node and additive if it has a cusp.

A curve has bad reduction at finitely many places. The Néron model of EE over 𝒪K\mathcal{O}_K is the smooth group scheme that extends EE and has the universal property for maps from smooth schemes; at a place of good reduction it is the minimal model.

Theorem (criterion of Néron, Ogg and Shafarevich). Let ℓ\ell be a prime not below vv. Then EE has good reduction at vv exactly when the inertia group at vv acts trivially on the Tate module Tℓ(E)T_\ell(E).

Quadratic twists

Definition. Let E:y2=x3+ax+bE : y^2 = x^3 + ax + b over KK of characteristic not 22 or 33, and let d∈K×d \in K^\times. The quadratic twist of EE by dd is E(d):dy2=x3+ax+bE^{(d)} : dy^2 = x^3 + ax + b, which is isomorphic to y2=x3+ad2x+bd3y^2 = x^3 + ad^2 x + bd^3.

The twist has the same jj-invariant, becomes isomorphic to EE over K(d)K(\sqrt d), and has discriminant d6Δd^6\Delta. It can change the reduction type at the primes dividing 2d2d, for example from additive to good.

The proof of CMCurveGoodOutsidePrimeSt twists ELE_L by dF=1d_F = 1 if p≡3(mod8)p \equiv 3 \pmod 8 and by dF=−pd_F = -p if p≡7(mod8)p \equiv 7 \pmod 8, then finds minimal models prime by prime.

Periods

Let EE be an elliptic curve over a subfield KK of ℂ\mathbb{C} and ω\omega a nonzero invariant differential on EE over KK, for example dx/2ydx/2y on y2=x3+ax+by^2 = x^3 + ax + b.

Definition. The period lattice of (E,ω)(E, \omega) is {∫γω:γ∈H1(E(ℂ),ℤ)}\{\int_\gamma \omega : \gamma \in H_1(E(\mathbb{C}), \mathbb{Z})\}. It is a lattice LL with E≅ELE \cong E_L, and a period of EE is an element of it.

Replacing ω\omega by cωc\omega with c∈K×c \in K^\times replaces LL by cLcL, so a period is well defined up to a factor in K×K^\times. For ω=dx/2y\omega = dx/2y, g2(L)=−4ag_2(L) = -4a and g3(L)=−4bg_3(L) = -4b lie in KK.

If Ω12=Δ(L)\Omega^{12} = \Delta(L), then Ω\Omega scales like c−1c^{-1}, so Gk(L)/ΩkG_k(L)/\Omega^k is invariant under scaling. It is a polynomial with rational coefficients in g2/Ω4g_2/\Omega^4 and g3/Ω6g_3/\Omega^6, hence algebraic when j(L)j(L) is algebraic. The periods aia_i of JetInt (below) are chosen so that ai/Ωa_i/\Omega, with Ω12=Δ(Λ)\Omega^{12} = \Delta(\Lambda), has a known ll-adic size.

Complex multiplication

Definition. The endomorphism ring End(E)\mathrm{End}(E) of an elliptic curve over ℂ\mathbb{C} is the ring of holomorphic group maps E→EE \to E. For E=ℂ/LE = \mathbb{C}/L it is {α∈ℂ:αL⊂L}\{\alpha \in \mathbb{C} : \alpha L \subset L\}, which is ℤ\mathbb{Z} or an order in an imaginary quadratic field KK. In the second case EE has complex multiplication (CM) by that order.

Example. For a fractional ideal 𝔞\mathfrak{a} of an imaginary quadratic field KK, the curve ℂ/𝔞\mathbb{C}/\mathfrak{a} has End=𝒪K\mathrm{End} = \mathcal{O}_K. For K=ℚ(i)K = \mathbb{Q}(i) the curve y2=x3−xy^2 = x^3 - x has j=1728j = 1728, and ii acts by (x,y)↦(−x,iy)(x, y) \mapsto (-x, iy).

Up to homothety, 𝔞\mathfrak{a} depends only on its ideal class, so there are exactly hKh_K curves with CM by 𝒪K\mathcal{O}_K up to isomorphism, and Cl(K)\mathrm{Cl}(K) acts simply transitively on them by 𝔟∗ℂ/𝔞=ℂ/𝔟−1𝔞\mathfrak{b} \ast \mathbb{C}/\mathfrak{a} = \mathbb{C}/\mathfrak{b}^{-1}\mathfrak{a}.

Theorem (j and the Hilbert class field). Let KK be imaginary quadratic and 𝔞\mathfrak{a} a fractional ideal of 𝒪K\mathcal{O}_K. Then j(𝔞)j(\mathfrak{a}) is an algebraic integer, [ℚ(j(𝔞)):ℚ]=hK[\mathbb{Q}(j(\mathfrak{a})) : \mathbb{Q}] = h_K, and K(j(𝔞))K(j(\mathfrak{a})) is the Hilbert class field HH of KK (class field theory). For each fractional ideal 𝔟\mathfrak{b} with Artin symbol σ𝔟∈Gal(H/K)\sigma_{\mathfrak{b}} \in \mathrm{Gal}(H/K), σ𝔟(j(𝔞))=j(𝔟−1𝔞)\sigma_{\mathfrak{b}}(j(\mathfrak{a})) = j(\mathfrak{b}^{-1}\mathfrak{a}).

So the field F=ℚ(j(𝒪K))F = \mathbb{Q}(j(\mathcal{O}_K)) of CMCurveGoodOutsidePrimeSt has degree hKh_K, and the curve has a model over FF.

Theorem (main theorem of complex multiplication). Let EE be an elliptic curve over ℂ\mathbb{C} with End(E)=𝒪K\mathrm{End}(E) = \mathcal{O}_K, and f:ℂ/𝔞→E(ℂ)f : \mathbb{C}/\mathfrak{a} \to E(\mathbb{C}) an analytic isomorphism. Let σ∈Aut(ℂ/K)\sigma \in \mathrm{Aut}(\mathbb{C}/K) and let s∈𝔸K×s \in \mathbb{A}_K^\times be an idele whose Artin symbol is σ\sigma on KabK^{\mathrm{ab}}. Then there is a unique analytic isomorphism f′:ℂ/s−1𝔞→Eσ(ℂ)f' : \mathbb{C}/s^{-1}\mathfrak{a} \to E^\sigma(\mathbb{C}) with f(x)σ=f′(s−1x)f(x)^\sigma = f'(s^{-1}x) for each x∈K/𝔞x \in K/\mathfrak{a} (Shimura, 1971).

Theorem (Shimura reciprocity). Let τ∈ℍ∩K\tau \in \mathbb{H} \cap K and let qτ:𝔸K,f×→GL2(𝔸ℚ,f)q_\tau : \mathbb{A}_{K,f}^\times \to \mathrm{GL}_2(\mathbb{A}_{\mathbb{Q},f}) be given by the action of 𝔸K,f×\mathbb{A}_{K,f}^\times on 𝔸K,f=𝔸ℚ,fτ+𝔸ℚ,f\mathbb{A}_{K,f} = \mathbb{A}_{\mathbb{Q},f}\tau + \mathbb{A}_{\mathbb{Q},f}. For each modular function ff of some level with Fourier coefficients in ℚab\mathbb{Q}^{\mathrm{ab}} that is finite at τ\tau, and each s∈𝔸K,f×s \in \mathbb{A}_{K,f}^\times, the value f(τ)f(\tau) lies in KabK^{\mathrm{ab}} and f(τ)[s−1,K]=fqτ(s)(τ)f(\tau)^{[s^{-1}, K]} = f^{q_\tau(s)}(\tau), where f↦fgf \mapsto f^g is the action of GL2(𝔸ℚ,f)\mathrm{GL}_2(\mathbb{A}_{\mathbb{Q},f}) on modular functions (Shimura, 1971, Theorem 6.31).

Both theorems write a Galois conjugate of a CM value as a value of the same shape at another lattice. In Transport, each conjugate σ(β)\sigma(\beta) of β=(C0⋅T)12/Δ(Λ)wt\beta = (C_0 \cdot T)^{12}/\Delta(\Lambda)^{\mathrm{wt}} is such a value for an ideal 𝔞σ\mathfrak{a}_\sigma and a lattice Λσ\Lambda_\sigma. The record CM and abelian inputs has the fields EllipticMainTheoremCM and HilbertClassFieldJ.

The Chowla-Selberg formula

The classical form of the formula writes the periods of CM curves through values of the Gamma function at rationals with denominator dd (Chowla and Selberg, 1967). The proof uses the form below, which is equivalent by the functional equation of L(s,χ)L(s, \chi) and Lerch's formula.

Theorem (Chowla-Selberg). Let KK be an imaginary quadratic field of discriminant −d-d with d>4d > 4, class number hh and quadratic character χ\chi. For each ideal class cc choose an ideal 𝔞c\mathfrak{a}_c in cc with lattice Lc⊂ℂL_c \subset \mathbb{C}. Then

1h∑clog⁡(N(𝔞c)6 ∣g2(Lc)3−27g3(Lc)2∣)=12log⁡2π−6log⁡d+6γ−6 L′(1,χ)L(1,χ),\frac{1}{h} \sum_{c} \log\Big( N(\mathfrak{a}_c)^6 \,\big| g_2(L_c)^3 - 27 g_3(L_c)^2 \big| \Big) = 12 \log 2\pi - 6 \log d + 6\gamma - 6\, \frac{L'(1, \chi)}{L(1, \chi)},

where γ\gamma is Euler's constant.

The left side does not depend on the choice of 𝔞c\mathfrak{a}_c, since N(𝔞)6∣Δ(𝔞)∣N(\mathfrak{a})^6|\Delta(\mathfrak{a})| is invariant under scaling. It follows from the first Kronecker limit formula for the real analytic Eisenstein series and the class number formula L(1,χ)=πh/dL(1, \chi) = \pi h/\sqrt d. In terms of η\eta, the mean of log⁡(Im⁡τ ∣η(τ)∣4)\log(\operatorname{Im}\tau\, |\eta(\tau)|^4) over the CM points τ\tau of the classes is γ−log⁡(2d)−L′/L(1,χ)\gamma - \log(2\sqrt d) - L'/L(1, \chi).

The code proves both forms: BetaArch averages the bounds on the conjugates of β\beta with the first, and the volume theorem of the Weil bound gets its term L′/L(1,χd)L'/L(1, \chi_d) from the second.

Gross's curves

Theorem (Gross's curves; Gross, 1980). Let p>3p > 3 be a prime with p≡3(mod4)p \equiv 3 \pmod 4, let K=ℚ(−p)K = \mathbb{Q}(\sqrt{-p}), L=𝒪K⊂ℂL = \mathcal{O}_K \subset \mathbb{C} and F=ℚ(j(L))F = \mathbb{Q}(j(L)). There is an elliptic curve AA over FF with j(A)=j(L)j(A) = j(L), good reduction at each prime of FF not dividing pp, and minimal discriminant (−p)3(-p)^3, that is, at each prime of FF a minimal model with discriminant a unit times (−p)3(-p)^3.

This is the statement CMCurveGoodOutsidePrimeSt, a field of the bundle UnitaryGroupThirdPlaceHypotheses that the code proves. Inside baseCond_KPx it is applied with p=dp = d, where −d-d is the discriminant of the host field and d≡7(mod16)d \equiv 7 \pmod{16}.

Eisenstein sums of lattices

A CM type φ=(φi)i∈I\varphi = (\varphi_i)_{i \in I} of EE is a set of embeddings E→ℂE \to \mathbb{C} that contains exactly one from each pair of complex conjugate embeddings. It embeds a lattice L⊂EL \subset E as a lattice in ℂI\mathbb{C}^I, and the Eisenstein lattice sums of weight 2s2s, analogues of G2sG_{2s}, are functions on ℂI\mathbb{C}^I built from LL. For E=KE = K imaginary quadratic they are the G2s(L)G_{2s}(L), and G2s(L)/Ω2sG_{2s}(L)/\Omega^{2s} is algebraic.

The statement JetInt of the CM divisibility chapter bounds the ll-adic size of the partial derivatives at 00 of these sums, divided by powers of periods aia_i. Its proof takes EllipticMainTheoremCM and HilbertClassFieldJ as inputs but uses only the functional equation of partial zeta functions.

Abelian varieties and abelian schemes

Definition. An abelian variety over a field kk is a connected smooth proper group scheme over kk. It is commutative and projective. An elliptic curve is an abelian variety of dimension 11, and over ℂ\mathbb{C} an abelian variety of dimension gg is a complex torus ℂg/Λ\mathbb{C}^g/\Lambda with a positive definite Riemann form.

Definition. The dual abelian variety of AA is A∨=PicA/k0A^\vee = \mathrm{Pic}^0_{A/k}, the abelian variety of line bundles on AA algebraically equivalent to 00. A line bundle ℒ\mathcal{L} on AA gives a homomorphism φℒ:A→A∨\varphi_{\mathcal{L}} : A \to A^\vee, x↦tx∗ℒ⊗ℒ−1x \mapsto t_x^*\mathcal{L} \otimes \mathcal{L}^{-1}.

Definition. A polarization of AA is a homomorphism λ:A→A∨\lambda : A \to A^\vee that over kˉ\bar k equals φℒ\varphi_{\mathcal{L}} for an ample line bundle ℒ\mathcal{L}. It is an isogeny, and it is principal if it is an isomorphism. An elliptic curve has a unique principal polarization, given by ℒ=𝒪(O)\mathcal{L} = \mathcal{O}(O), so E≅E∨E \cong E^\vee.

Definition. Let SS be a scheme. An abelian scheme over SS is a smooth proper group scheme over SS with geometrically connected fibres. A semi-abelian scheme is a smooth separated commutative group scheme over SS whose fibres are extensions of an abelian variety by a torus.

The dual, the polarizations and the Tate modules Tℓ(A)=lim←⁡A[ℓn]T_\ell(A) = \varprojlim A[\ell^n] extend to abelian schemes. An abelian variety AA over a number field KK has good reduction at vv if it extends to an abelian scheme over 𝒪v\mathcal{O}_v.

Theorem (Serre and Tate). Let KK be a field with a discrete valuation vv of residue characteristic pp, let AA be an abelian variety over KK and let ℓ≠p\ell \neq p be a prime. Then AA has good reduction at vv exactly when inertia at vv acts trivially on Tℓ(A)T_\ell(A). An abelian variety with complex multiplication has potentially good reduction at every place (Serre and Tate, 1968).

Theorem (semistable reduction). An abelian variety over a number field KK acquires semistable reduction over a finite extension K′K' of KK: the identity components of the fibres of its Néron model over 𝒪K′\mathcal{O}_{K'} are semi-abelian (Grothendieck, SGA 7).

The largest field of CM and abelian inputs, used by the window bound, is SerreTateGoodReductionSt. Semi-abelian schemes and their Lie algebras enter the slope bound of the Weil bound.

Further reading

  • J. H. Silverman, The Arithmetic of Elliptic Curves, Chapters III, VI and VII.
  • J. H. Silverman, Advanced Topics in the Arithmetic of Elliptic Curves, Chapter II.
  • G. Shimura, Introduction to the Arithmetic Theory of Automorphic Functions, Chapters 4 to 6.
  • D. A. Cox, Primes of the Form x2+ny2x^2 + ny^2, Sections 10 and 11.
  • D. Mumford, Abelian Varieties, Chapters I and II.

The source of this page. Markdown. Something is wrong? Report it.