riemann

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

Galois representations

Galois representations, Weil-Deligne representations and the local Langlands correspondence, with the p-adic Hodge theory needed at the places above ℓ\ell. The row statement of the proof asks for such a representation attached to an automorphic representation.

Prerequisites: Automorphic forms, Class field theory.

A Galois representation is a continuous action of the absolute Galois group of a field on a finite-dimensional vector space. Over a number field it is described place by place: by Frobenius eigenvalues where it is unramified, by a Weil-Deligne representation at the other places away from ℓ\ell, and by p-adic Hodge theory above ℓ\ell. The row statement of the proof asks for such a representation attached to a cuspidal automorphic representation.

Profinite groups

Definition. A profinite group is a topological group that is an inverse limit of finite discrete groups; equivalently, a compact, Hausdorff, totally disconnected topological group. Its open subgroups have finite index and form a basis of neighbourhoods of 11. For example, ℤℓ=lim←⁡nℤ/ℓnℤ\mathbb{Z}_\ell = \varprojlim_n \mathbb{Z}/\ell^n\mathbb{Z} and ℤ^=lim←⁡nℤ/nℤ≅∏ℓℤℓ\widehat{\mathbb{Z}} = \varprojlim_n \mathbb{Z}/n\mathbb{Z} \cong \prod_\ell \mathbb{Z}_\ell.

Definition. Let KK be a field with separable closure K‾\overline{K}. The absolute Galois group of KK is GK=Gal(K‾/K)=lim←⁡LGal(L/K)G_K = \mathrm{Gal}(\overline{K}/K) = \varprojlim_L \mathrm{Gal}(L/K), over the finite Galois extensions LL of KK in K‾\overline{K}, with the inverse limit (Krull) topology.

Theorem (infinite Galois theory). The map H↦K‾HH \mapsto \overline{K}^H is a bijection from the closed subgroups of GKG_K to the extensions of KK in K‾\overline{K}. Open subgroups correspond to finite extensions, and closed normal subgroups to Galois extensions.

Definition (the group GSG_S). Let KK be a number field and SS a set of places of KK containing the infinite places. Let KSK_S be the largest extension of KK in K‾\overline{K} unramified at each finite place outside SS, and GS=Gal(KS/K)G_S = \mathrm{Gal}(K_S/K). A representation of GKG_K unramified outside SS factors through the quotient GSG_S.

Row0b writes Gal(Ω/E)\mathrm{Gal}(\Omega/E) as G E Ω, for an algebraic closure Ω\Omega of a CM field EE, and VWin is about the cohomology of a group GS of this kind.

Decomposition groups, inertia and Frobenius

Let KK be a number field and vv a finite place, with completion KvK_v and residue field kvk_v of size qvq_v (number fields). Fix an embedding K‾→K‾v\overline{K} \to \overline{K}_v over KK; restriction gives an injective continuous map GKv→GKG_{K_v} \to G_K.

Definition. The image of GKvG_{K_v} in GKG_K is the decomposition group at vv; another embedding replaces it by a conjugate. For finite Galois L/KL/K, its image in Gal(L/K)\mathrm{Gal}(L/K) is {σ:σ(w)=w}\{\sigma : \sigma(w) = w\}, for the place ww of LL given by the embedding.

Definition. The action of GKvG_{K_v} on the integers of K‾v\overline{K}_v gives a surjection GKv→Gkv=Gal(k‾v/kv)G_{K_v} \to G_{k_v} = \mathrm{Gal}(\overline{k}_v/k_v), whose kernel is the inertia group IvI_v. An algebraic extension of KvK_v in K‾v\overline{K}_v is unramified if and only if IvI_v acts trivially on it. The group Gkv≅ℤ^G_{k_v} \cong \widehat{\mathbb{Z}} has topological generator x↦xqvx \mapsto x^{q_v}.

Definition. A Frobenius element (or arithmetic Frobenius) at vv is an element Frobv∈GKv\mathrm{Frob}_v \in G_{K_v} whose image in GkvG_{k_v} is x↦xqvx \mapsto x^{q_v}; its inverse is a geometric Frobenius.

Two Frobenius elements at vv differ by conjugation and by an element of IvI_v. If L/KL/K is finite Galois and unramified at vv, their images in Gal(L/K)\mathrm{Gal}(L/K) form one conjugacy class. By the Chebotarev density theorem (L-functions), for finite SS the conjugates of the Frobv\mathrm{Frob}_v with v∉Sv \notin S are dense in GSG_S.

The condition ESAt in GeoOne says that ρ\rho is trivial on inertia at ww and that a given polynomial kills ρ(σ−1)\rho(\sigma^{-1}) for each Frobenius element σ\sigma at ww.

Galois representations

Definition. Let AA be a topological commutative ring and n≥1n \geq 1. A Galois representation of KK of dimension nn over AA is a continuous homomorphism ρ:GK→GLn(A)\rho : G_K \to \mathrm{GL}_n(A); two are isomorphic if they are conjugate by an element of GLn(A)\mathrm{GL}_n(A). It is ℓ\ell-adic when AA is a finite extension of ℚℓ\mathbb{Q}_\ell or ℚ‾ℓ\overline{\mathbb{Q}}_\ell, with the topology of the ℓ\ell-adic absolute value. It is semisimple if each GKG_K-stable subspace has a GKG_K-stable complement.

Proposition. If ρ:GK→GLn(ℚ‾ℓ)\rho : G_K \to \mathrm{GL}_n(\overline{\mathbb{Q}}_\ell) is continuous, then ρ(GK)⊂GLn(L)\rho(G_K) \subset \mathrm{GL}_n(L) for a finite extension LL of ℚℓ\mathbb{Q}_\ell, and ρ\rho is conjugate to a representation with values in GLn(𝒪L)\mathrm{GL}_n(\mathcal{O}_L).

Its reduction modulo the maximal ideal of 𝒪L\mathcal{O}_L is a residual representation ρˉ:GK→GLn(𝔽‾ℓ)\bar\rho : G_K \to \mathrm{GL}_n(\overline{\mathbb{F}}_\ell), whose semisimplification does not depend on the lattice (Brauer-Nesbitt).

Definition. A Galois representation ρ\rho is an unramified representation at a finite place vv if ρ(Iv)=1\rho(I_v) = 1. Then ρ(Frobv)\rho(\mathrm{Frob}_v) is well defined up to conjugacy, and the roots of its characteristic polynomial are the Frobenius eigenvalues of ρ\rho at vv.

Theorem. Let KK be a number field, SS a finite set of places, and ρ,ρ′\rho, \rho' semisimple ℓ\ell-adic representations of GKG_K unramified outside SS. If ρ(Frobv)\rho(\mathrm{Frob}_v) and ρ′(Frobv)\rho'(\mathrm{Frob}_v) have the same characteristic polynomial for each v∉Sv \notin S, then ρ≅ρ′\rho \cong \rho'.

(The traces agree on a dense set by Chebotarev, so everywhere, and they determine a semisimple representation in characteristic 00.)

Example. A continuous character χ:GK→ℚ‾ℓ×\chi : G_K \to \overline{\mathbb{Q}}_\ell^\times factors through GKabG_K^{\mathrm{ab}}. If it is unramified almost everywhere and de Rham (defined below) at the places above ℓ\ell, then by class field theory it is the ℓ\ell-adic form of an algebraic Hecke character.

The fields cont and ss of Row0b ask that rr is continuous and semisimple, and CoreAll in the Galois row asks that at almost all ww, rr is unramified with Frobenius eigenvalues given by the Satake parameters of πw\pi_w (the formula is below).

The cyclotomic character and Tate modules

Definition. Let ℓ\ell be a prime different from the characteristic of KK, and μℓn⊂K‾\mu_{\ell^n} \subset \overline{K} the roots of unity of order dividing ℓn\ell^n. The cyclotomic character is the continuous character εℓ:GK→ℤℓ×\varepsilon_\ell : G_K \to \mathbb{Z}_\ell^\times with σ(ζ)=ζεℓ(σ) mod ℓn\sigma(\zeta) = \zeta^{\varepsilon_\ell(\sigma) \bmod \ell^n} for all nn and ζ∈μℓn\zeta \in \mu_{\ell^n}. For a ℤℓ[GK]\mathbb{Z}_\ell[G_K]-module MM and k∈ℤk \in \mathbb{Z}, the Tate twist M(k)M(k) is MM with σ\sigma acting as εℓ(σ)kσ\varepsilon_\ell(\sigma)^k \sigma.

Example. For a number field KK and a finite place v∤ℓv \nmid \ell, εℓ\varepsilon_\ell is unramified at vv with εℓ(Frobv)=qv\varepsilon_\ell(\mathrm{Frob}_v) = q_v. For K=ℚK = \mathbb{Q} it is the isomorphism Gal(ℚ(μℓ∞)/ℚ)≅ℤℓ×\mathrm{Gal}(\mathbb{Q}(\mu_{\ell^\infty})/\mathbb{Q}) \cong \mathbb{Z}_\ell^\times.

The discrete module ℚℓ/ℤℓ(k)=lim→⁡nℤ/ℓnℤ(k)\mathbb{Q}_\ell/\mathbb{Z}_\ell(k) = \varinjlim_n \mathbb{Z}/\ell^n\mathbb{Z}(k) is μℓ∞\mu_{\ell^\infty} for k=1k = 1.

Definition. Let AA be an abelian variety of dimension gg over KK (Chapter 12), and ℓ≠char K\ell \neq \mathrm{char}\, K. The Tate module of AA is TℓA=lim←⁡nA[ℓn](K‾)T_\ell A = \varprojlim_n A[\ell^n](\overline{K}), along multiplication by ℓ\ell. It is a free ℤℓ\mathbb{Z}_\ell-module of rank 2g2g with a continuous action of GKG_K. Put VℓA=TℓA⊗ℤℓℚℓV_\ell A = T_\ell A \otimes_{\mathbb{Z}_\ell} \mathbb{Q}_\ell. The Tate module of the multiplicative group is lim←⁡nμℓn=ℤℓ(1)\varprojlim_n \mu_{\ell^n} = \mathbb{Z}_\ell(1).

Theorem (Néron-Ogg-Shafarevich criterion; Serre and Tate, 1968). Let KK be a number field and v∤ℓv \nmid \ell a finite place. Then AA has good reduction at vv if and only if VℓAV_\ell A is unramified at vv. In that case the characteristic polynomial of Frobv\mathrm{Frob}_v on VℓAV_\ell A has integer coefficients independent of ℓ\ell.

The group of VWin lies in an H1H^1 with coefficients ℚl/ℤl(−e)\mathbb{Q}_l/\mathbb{Z}_l(-e), and the input FarguesExact of the Weil bound is about the Tate modules of a p-divisible group (Chapter 13) and its dual.

The Weil group

Let FF be a nonarchimedean local field with residue field kk of size qq, a power of the prime pp, and let IF=ker⁡(GF→Gk)I_F = \ker(G_F \to G_k).

Definition. The Weil group WFW_F is the group of σ∈GF\sigma \in G_F whose image in GkG_k is x↦xqnx \mapsto x^{q^n} for some n∈ℤn \in \mathbb{Z}; put deg⁡σ=n\deg \sigma = n and ∣σ∣=qdeg⁡σ|\sigma| = q^{\deg \sigma}. Its topology makes IFI_F, with the profinite topology, an open subgroup.

So WFW_F is an extension of ℤ\mathbb{Z} by IFI_F, dense in GFG_F, and a geometric Frobenius Φ\Phi has deg⁡Φ=−1\deg \Phi = -1, ∣Φ∣=q−1|\Phi| = q^{-1}. We normalize the local Artin map ArtF:F×→WFab\mathrm{Art}_F : F^\times \to W_F^{\mathrm{ab}} of class field theory to send uniformizers to geometric Frobenius elements; then ∣ArtF(x)∣=∣x∣F|\mathrm{Art}_F(x)| = |x|_F.

Definition. The wild inertia group PFP_F is the unique pro-pp Sylow subgroup of IFI_F. A representation of WFW_F is tame if it is trivial on PFP_F.

The quotient IF/PFI_F/P_F is isomorphic to ∏ℓ≠pℤℓ(1)\prod_{\ell \neq p} \mathbb{Z}_\ell(1). For ℓ≠p\ell \neq p let tℓ:IF→ℤℓ(1)t_\ell : I_F \to \mathbb{Z}_\ell(1) be the projection; then tℓ(στσ−1)=εℓ(σ) tℓ(τ)t_\ell(\sigma \tau \sigma^{-1}) = \varepsilon_\ell(\sigma)\, t_\ell(\tau) for σ∈WF\sigma \in W_F, τ∈IF\tau \in I_F.

In IsRecT the map a:Wh→Ew×a : W_h \to E_w^\times with IsLocalArtin inverts the local Artin map on abelianizations; its field ord_eq says that a(σ)a(\sigma) has valuation −deg⁡σ-\deg \sigma.

Weil-Deligne representations

Fix a field Ω\Omega of characteristic 00, such as ℂ\mathbb{C} or ℚ‾ℓ\overline{\mathbb{Q}}_\ell.

Definition. A Weil-Deligne representation of WFW_F over Ω\Omega is a pair (r,N)(r, N) of a homomorphism r:WF→GL(V)r : W_F \to \mathrm{GL}(V), on a finite-dimensional Ω\Omega-vector space VV, trivial on an open subgroup of IFI_F, and a nilpotent endomorphism NN of VV, the monodromy operator, with r(σ)Nr(σ)−1=∣σ∣Nr(\sigma) N r(\sigma)^{-1} = |\sigma| N for σ∈WF\sigma \in W_F. Morphisms commute with rr and NN. The pair is Frobenius-semisimple if r(Φ)r(\Phi) is semisimple.

The tensor product of (r,N)(r, N) and (r′,N′)(r', N') is (r⊗r′,N⊗1+1⊗N′)(r \otimes r', N \otimes 1 + 1 \otimes N'). The Frobenius-semisimplification (r,N)F-ss(r, N)^{\mathrm{F\text{-}ss}} replaces r(Φ)r(\Phi) by its semisimple part.

Example (Sp(m)\mathrm{Sp}(m)). For m≥1m \geq 1, Sp(m)\mathrm{Sp}(m) is Ωm\Omega^m with basis e0,…,em−1e_0, \dots, e_{m-1}, r(σ)ei=∣σ∣ieir(\sigma) e_i = |\sigma|^i e_i, Nei=ei+1N e_i = e_{i+1} for i<m−1i < m - 1 and Nem−1=0N e_{m-1} = 0. It is indecomposable, and each indecomposable Frobenius-semisimple Weil-Deligne representation is (r,0)⊗Sp(m)(r, 0) \otimes \mathrm{Sp}(m) for an irreducible rr and a unique mm.

Theorem (Grothendieck monodromy theorem). Let ℓ≠p\ell \neq p, fix a generator of ℤℓ(1)\mathbb{Z}_\ell(1) so that tℓt_\ell takes values in ℤℓ\mathbb{Z}_\ell, and fix a geometric Frobenius Φ\Phi. Let ρ:WF→GLn(ℚ‾ℓ)\rho : W_F \to \mathrm{GL}_n(\overline{\mathbb{Q}}_\ell) be continuous. There is a unique nilpotent NN with ρ(τ)=exp⁡(tℓ(τ)N)\rho(\tau) = \exp(t_\ell(\tau) N) for all τ\tau in an open subgroup of IFI_F. Then r(Φkτ)=ρ(Φkτ)exp⁡(−tℓ(τ)N)r(\Phi^k \tau) = \rho(\Phi^k \tau) \exp(-t_\ell(\tau) N), for k∈ℤk \in \mathbb{Z} and τ∈IF\tau \in I_F, defines a Weil-Deligne representation WD(ρ)=(r,N)\mathrm{WD}(\rho) = (r, N), whose isomorphism class does not depend on the choices and determines ρ\rho up to isomorphism.

So ρ\rho is trivial on an open subgroup of inertia exactly when N=0N = 0.

In IsRecT, rcrc takes values in the set WD(h)\mathrm{WD}(h) of classes of Weil-Deligne representations, and the law langlands uses Sp(m)\mathrm{Sp}(m). In Row0b, wdOf h hl ι r is the class of WD(r∣GEw)\mathrm{WD}(r|_{G_{E_w}}), and the clause tor of AwayB concerns classes with N=0N = 0 that factor through WFabW_F^{\mathrm{ab}}.

The local Langlands correspondence

Let FF be a finite extension of ℚp\mathbb{Q}_p. For n≥1n \geq 1 let 𝒜n(F)\mathcal{A}_n(F) be the set of isomorphism classes of irreducible smooth representations of GLn(F)\mathrm{GL}_n(F) (representations), and 𝒢n(F)\mathcal{G}_n(F) the set of isomorphism classes of nn-dimensional Frobenius-semisimple Weil-Deligne representations of WFW_F over ℂ\mathbb{C}.

Theorem (the local Langlands correspondence; Harris and Taylor, 2001; Henniart, 2000). There is a unique family of bijections recF:𝒜n(F)→𝒢n(F)\mathrm{rec}_F : \mathcal{A}_n(F) \to \mathcal{G}_n(F), n≥1n \geq 1, such that:

  1. for n=1n = 1, recF(χ)=χ∘ArtF−1\mathrm{rec}_F(\chi) = \chi \circ \mathrm{Art}_F^{-1};
  2. for πi∈𝒜ni(F)\pi_i \in \mathcal{A}_{n_i}(F) and each nontrivial additive character ψ\psi of FF, L(s,π1×π2)=L(s,recF(π1)⊗recF(π2))L(s, \pi_1 \times \pi_2) = L(s, \mathrm{rec}_F(\pi_1) \otimes \mathrm{rec}_F(\pi_2)) and ε(s,π1×π2,ψ)=ε(s,recF(π1)⊗recF(π2),ψ)\varepsilon(s, \pi_1 \times \pi_2, \psi) = \varepsilon(s, \mathrm{rec}_F(\pi_1) \otimes \mathrm{rec}_F(\pi_2), \psi);
  3. det⁡recF(π)=recF(ωπ)\det \mathrm{rec}_F(\pi) = \mathrm{rec}_F(\omega_\pi), for the central character ωπ\omega_\pi of π\pi;
  4. recF(π∨)=recF(π)∨\mathrm{rec}_F(\pi^\vee) = \mathrm{rec}_F(\pi)^\vee;
  5. recF(π⊗(χ∘det⁡))=recF(π)⊗recF(χ)\mathrm{rec}_F(\pi \otimes (\chi \circ \det)) = \mathrm{rec}_F(\pi) \otimes \mathrm{rec}_F(\chi) for each smooth character χ\chi of F×F^\times.

Uniqueness follows from 1, 2 and the local converse theorem. Scholze (2013) gave another proof of existence.

Proposition. Moreover:

  1. π\pi is supercuspidal if and only if recF(π)\mathrm{rec}_F(\pi) is irreducible, and then N=0N = 0.
  2. For supercuspidal σ\sigma, a segment Δ=[σ,σ∣det⁡∣m−1]\Delta = [\sigma, \sigma |{\det}|^{m-1}] and its Langlands quotient Q(Δ)Q(\Delta), recF(Q(Δ))=recF(σ)⊗Sp(m)\mathrm{rec}_F(Q(\Delta)) = \mathrm{rec}_F(\sigma) \otimes \mathrm{Sp}(m).
  3. If π\pi is the Langlands quotient of a multisegment {Δ1,…,Δk}\{\Delta_1, \dots, \Delta_k\}, then recF(π)=⨁irecF(Q(Δi))\mathrm{rec}_F(\pi) = \bigoplus_i \mathrm{rec}_F(Q(\Delta_i)).
  4. If π\pi is unramified with Satake parameters b1,…,bnb_1, \dots, b_n, then recF(π)\mathrm{rec}_F(\pi) is unramified with N=0N = 0, and r(Φ)r(\Phi) has eigenvalues b1,…,bnb_1, \dots, b_n.
  5. If π\pi has a nonzero vector fixed by 1+ϖMn(𝒪F)1 + \varpi \mathrm{M}_n(\mathcal{O}_F), then recF(π)\mathrm{rec}_F(\pi) is tame.

The predicate IsRecT lists twelve laws of recF\mathrm{rec}_F, some only on a class Good, among them the case n=1n = 1, segments, twists, central characters, duals and depth. With no law on L-factors or epsilon factors, it does not single out recF\mathrm{rec}_F. Its law langlands is property 2 with the Lean normalization of Sp(m)\mathrm{Sp}(m), and RecTExists gives such a map at each place.

Period rings

Let KK be a finite extension of ℚp\mathbb{Q}_p with residue field kk, K0=W(k)[1/p]K_0 = W(k)[1/p], CC the completion of an algebraic closure K‾\overline{K}, and 𝒪C\mathcal{O}_C its ring of integers; GKG_K acts continuously on CC. p-adic Hodge theory studies the p-adic representations of GKG_K, the continuous representations on finite-dimensional ℚp\mathbb{Q}_p-vector spaces, where the Grothendieck monodromy theorem does not apply.

Theorem (Tate, 1967). CGK=KC^{G_K} = K, and C(k)GK=0C(k)^{G_K} = 0 for each k≠0k \neq 0.

Definition. The tilt of CC is C♭=lim←⁡x↦xpCC^\flat = \varprojlim_{x \mapsto x^p} C, the set of sequences x=(x(0),x(1),… )x = (x^{(0)}, x^{(1)}, \dots) in CC with (x(n+1))p=x(n)(x^{(n+1)})^p = x^{(n)}, with termwise product and the sum (x+y)(n)=lim⁡m→∞(x(n+m)+y(n+m))pm(x + y)^{(n)} = \lim_{m \to \infty} (x^{(n+m)} + y^{(n+m)})^{p^m}. With ∣x∣=∣x(0)∣|x| = |x^{(0)}| it is an algebraically closed complete field of characteristic pp, and 𝒪C♭≅lim←⁡x↦xp𝒪C/p\mathcal{O}_{C^\flat} \cong \varprojlim_{x \mapsto x^p} \mathcal{O}_C/p.

Definition. Fontaine's ring is Ainf=W(𝒪C♭)A_{\mathrm{inf}} = W(\mathcal{O}_{C^\flat}), the Witt vectors of 𝒪C♭\mathcal{O}_{C^\flat}, with the Frobenius φ\varphi induced by x↦xpx \mapsto x^p. There is a unique ring map θ:Ainf→𝒪C\theta : A_{\mathrm{inf}} \to \mathcal{O}_C with θ([x])=x(0)\theta([x]) = x^{(0)} for Teichmüller lifts [x][x]. It is surjective, with kernel generated by ξ=p−[p♭]\xi = p - [p^\flat], where (p♭)(0)=p(p^\flat)^{(0)} = p.

Definition (Fontaine period rings). Let ϵ=(1,ζp,ζp2,… )∈𝒪C♭\epsilon = (1, \zeta_p, \zeta_{p^2}, \dots) \in \mathcal{O}_{C^\flat} for a compatible system of primitive pnp^n-th roots of unity.

  1. BdR+B_{\mathrm{dR}}^+ is the ker⁡θ\ker\theta-adic completion of Ainf[1/p]A_{\mathrm{inf}}[1/p]. The series t=log⁡[ϵ]t = \log [\epsilon] converges in it and generates its maximal ideal, and σ(t)=εp(σ)t\sigma(t) = \varepsilon_p(\sigma) t. The field BdR=BdR+[1/t]B_{\mathrm{dR}} = B_{\mathrm{dR}}^+[1/t] has the filtration Fili=tiBdR+\mathrm{Fil}^i = t^i B_{\mathrm{dR}}^+, with griBdR≅C(i)\mathrm{gr}^i B_{\mathrm{dR}} \cong C(i).
  2. AcrisA_{\mathrm{cris}} is the pp-adic completion of the divided power envelope of AinfA_{\mathrm{inf}} with respect to ker⁡θ\ker\theta, and Bcris=Acris[1/t]B_{\mathrm{cris}} = A_{\mathrm{cris}}[1/t]. It has a Frobenius φ\varphi and an injection into BdRB_{\mathrm{dR}}.
  3. Bst=Bcris[u]B_{\mathrm{st}} = B_{\mathrm{cris}}[u] for an indeterminate uu, with φ(u)=pu\varphi(u) = pu and monodromy operator N=−d/duN = -d/du. An embedding Bst→BdRB_{\mathrm{st}} \to B_{\mathrm{dR}} sends uu to log⁡[p♭]\log [p^\flat].

These are Fontaine's period rings, each with a continuous action of GKG_K. For B=Bcris,Bst,BdRB = B_{\mathrm{cris}}, B_{\mathrm{st}}, B_{\mathrm{dR}}, BGKB^{G_K} is K0K_0, K0K_0, KK. For a p-adic representation VV put DB(V)=(B⊗ℚpV)GKD_B(V) = (B \otimes_{\mathbb{Q}_p} V)^{G_K}. Then dim⁡BGKDB(V)≤dim⁡ℚpV\dim_{B^{G_K}} D_B(V) \leq \dim_{\mathbb{Q}_p} V, and VV is BB-admissible when equality holds.

The bundle nLink of UnitaryGroupThirdPlaceHypotheses has a statement named after BcrisB_{\mathrm{cris}}, and the tilt and AinfA_{\mathrm{inf}} enter the prismatic Dieudonné statement of CM divisibility.

Crystalline, de Rham and potentially semistable representations

Definition (Hodge-Tate weights). Let VV be a p-adic representation of GKG_K. An integer ii is a Hodge-Tate weight of VV, with multiplicity dim⁡K(V⊗ℚpC(i))GK\dim_K (V \otimes_{\mathbb{Q}_p} C(i))^{G_K}, when this number is positive. VV is Hodge-Tate if the multiplicities add up to dim⁡V\dim V. With this convention ℚp(1)\mathbb{Q}_p(1) has Hodge-Tate weight −1-1.

If VV has coefficients in a finite extension LL of ℚp\mathbb{Q}_p containing the images of all embeddings τ:K→ℚ‾p\tau : K \to \overline{\mathbb{Q}}_p, then K⊗ℚpL=∏τLK \otimes_{\mathbb{Q}_p} L = \prod_\tau L, and the τ\tau-components of the (V⊗C(i))GK(V \otimes C(i))^{G_K} give a multiset HTτ(V)\mathrm{HT}_\tau(V) of dim⁡LV\dim_L V integers; likewise for ℚ‾p\overline{\mathbb{Q}}_p coefficients. Proof pages call its elements Hodge-Tate jumps.

Definition. VV is crystalline, semistable or de Rham if it is admissible for BcrisB_{\mathrm{cris}}, BstB_{\mathrm{st}} or BdRB_{\mathrm{dR}}, and potentially semistable if its restriction to GK′G_{K'} is semistable for some finite extension K′/KK'/K.

Since Bcris⊂Bst⊂BdRB_{\mathrm{cris}} \subset B_{\mathrm{st}} \subset B_{\mathrm{dR}}, crystalline implies semistable, semistable implies de Rham, and de Rham implies Hodge-Tate.

Theorem (p-adic monodromy theorem; Berger, 2002). A p-adic representation is de Rham if and only if it is potentially semistable.

Example. Unramified characters and ℚp(1)\mathbb{Q}_p(1) are crystalline. If an abelian variety AA over KK has good reduction, then VpAV_p A is crystalline with Hodge-Tate weights 00 and −1-1, each of multiplicity gg (Tate, 1967; Fontaine, 1982).

Definition. Let VV be semistable over a finite Galois extension K′/KK'/K. The space Dst,K′(V)=(Bst⊗V)GK′D_{\mathrm{st}, K'}(V) = (B_{\mathrm{st}} \otimes V)^{G_{K'}} carries φ\varphi, NN and an action of Gal(K′/K)\mathrm{Gal}(K'/K), from which Fontaine (1994) builds a Weil-Deligne representation WD(V)\mathrm{WD}(V) of WKW_K. When VV is crystalline, WD(V)\mathrm{WD}(V) is unramified with N=0N = 0, and r(Φ)r(\Phi) acts through φf\varphi^f, where pf=∣k∣p^f = |k|.

The field above of Row0b asks that rr at w∣ℓw \mid \ell is crystalline with Hodge-Tate jumps HTHT when πw\pi_w is spherical, and, only when rc(πw)rc(\pi_w) has N=0N = 0 and some further conditions hold, potentially semistable with a class wdAt matching rc(πw)rc(\pi_w). PhiKilled in GeoOne concerns the characteristic polynomial of wdAt at an element of degree −1-1.

The Galois representation of an automorphic representation

Let EE be a CM field with maximal totally real subfield E+E^+ and complex conjugation cc. Fix a prime ℓ\ell and an isomorphism ι:ℚ‾ℓ≅ℂ\iota : \overline{\mathbb{Q}}_\ell \cong \mathbb{C}. Let π\pi be a regular algebraic cuspidal automorphic representation of GLn(𝔸E)\mathrm{GL}_n(\mathbb{A}_E) and χ\chi a Hecke character of E+E^+ with (π,χ)(\pi, \chi) polarized (automorphic forms).

Theorem (Chenevier and Harris, 2013; Shin, 2011; Caraiani, 2012 and 2014; Barnet-Lamb, Gee, Geraghty and Taylor, 2014). There is a continuous semisimple representation r=rℓ,ι(π):GE→GLn(ℚ‾ℓ)r = r_{\ell,\iota}(\pi) : G_E \to \mathrm{GL}_n(\overline{\mathbb{Q}}_\ell) such that:

  1. rc≅r∨⊗εℓ1−n rℓ,ι(χ)∣GEr^c \cong r^\vee \otimes \varepsilon_\ell^{1-n}\, r_{\ell,\iota}(\chi)|_{G_E}, where rℓ,ι(χ)r_{\ell,\iota}(\chi) is the ℓ\ell-adic character of χ\chi;
  2. for each finite place v∤ℓv \nmid \ell, ι WD(r∣GEv)F-ss≅recEv(πv⊗∣det⁡∣v(1−n)/2)\iota\, \mathrm{WD}(r|_{G_{E_v}})^{\mathrm{F\text{-}ss}} \cong \mathrm{rec}_{E_v}(\pi_v \otimes |{\det}|_v^{(1-n)/2});
  3. for each v∣ℓv \mid \ell, r∣GEvr|_{G_{E_v}} is potentially semistable, crystalline when πv\pi_v is unramified, and 2 holds with Fontaine's WD\mathrm{WD};
  4. for each τ:E→ℚ‾ℓ\tau : E \to \overline{\mathbb{Q}}_\ell, HTτ(r)\mathrm{HT}_\tau(r) consists of nn distinct integers given by the infinitesimal character of π∞\pi_\infty at ι∘τ\iota \circ \tau.

Properties 2 and 3 are local-global compatibility. At an unramified v∤ℓv \nmid \ell with Satake parameters BB, property 2 says that rr is unramified at vv and r(Frobv−1)r(\mathrm{Frob}_v^{-1}) has characteristic polynomial ∏b∈B(X−ι−1(qv(n−1)/2b))\prod_{b \in B}(X - \iota^{-1}(q_v^{(n-1)/2} b)); the twist by ∣det⁡∣(1−n)/2|{\det}|^{(1-n)/2} makes these eigenvalues algebraic.

Row0b is this theorem, proved in Lean in the Galois row holds: away is property 2 for each rcrc with IsRecT and the twist algNorm σ↦σ⊗∣det⁡∣(1−a)/2\sigma \mapsto \sigma \otimes |{\det}|^{(1-a)/2}, weights and weights_conj are property 4, and above is a weak form of property 3. The field caraianiLocalGlobalDual of the literature structure is named after Caraiani's theorem, but on the trunk it has the same Lean proof.

Modularity lifting

A representation ρ:GF→GLn(ℚ‾ℓ)\rho : G_F \to \mathrm{GL}_n(\overline{\mathbb{Q}}_\ell) of a number field FF is automorphic if ρ≅rℓ,ι(π)\rho \cong r_{\ell,\iota}(\pi) for some automorphic representation π\pi of GLn(𝔸F)\mathrm{GL}_n(\mathbb{A}_F).

Definition. Let 𝒪\mathcal{O} be the ring of integers of a finite extension of ℚℓ\mathbb{Q}_\ell, with residue field kk, let SS be finite, and let ρˉ:GS→GLn(k)\bar\rho : G_S \to \mathrm{GL}_n(k) be absolutely irreducible. A deformation of ρˉ\bar\rho to a complete local noetherian 𝒪\mathcal{O}-algebra AA with residue field kk is a lift GS→GLn(A)G_S \to \mathrm{GL}_n(A) of ρˉ\bar\rho, up to conjugation by ker⁡(GLn(A)→GLn(k))\ker(\mathrm{GL}_n(A) \to \mathrm{GL}_n(k)). The deformations with given local conditions are the maps from a universal deformation ring RR (Mazur, 1989).

A modularity lifting theorem says that if ρˉ\bar\rho is automorphic, ρ\rho has local properties at ℓ\ell like those of an automorphic representation, and ρˉ\bar\rho has large image, then ρ\rho is automorphic. The Taylor-Wiles patching method proves such a theorem by showing R=𝕋R = \mathbb{T} for a localized Hecke algebra 𝕋\mathbb{T} acting on automorphic forms.

Theorem (Barnet-Lamb, Gee, Geraghty and Taylor, 2014, Theorem 4.2.1). Let FF be an imaginary CM field, ℓ≥2(n+1)\ell \geq 2(n+1) a prime, and r:GF→GLn(ℚ‾ℓ)r : G_F \to \mathrm{GL}_n(\overline{\mathbb{Q}}_\ell) continuous. Suppose:

  1. rr is unramified at all but finitely many places;
  2. rc≅r∨⊗εℓ1−nμ∣GFr^c \cong r^\vee \otimes \varepsilon_\ell^{1-n} \mu|_{G_F} for a character μ\mu of GF+G_{F^+} with μ(cv)\mu(c_v) the same at each real place vv;
  3. at each v∣ℓv \mid \ell, r∣GFvr|_{G_{F_v}} is potentially diagonalizable and has nn distinct Hodge-Tate weights for each τ\tau;
  4. rˉ∣GF(ζℓ)\bar r|_{G_{F(\zeta_\ell)}} is irreducible, and the field cut out by ad rˉ\mathrm{ad}\, \bar r does not contain F(ζℓ)F(\zeta_\ell);
  5. rˉ≅rˉℓ,ι(π)\bar r \cong \bar r_{\ell,\iota}(\pi) for a regular algebraic polarized cuspidal π\pi with rℓ,ι(π)r_{\ell,\iota}(\pi) potentially diagonalizable at each v∣ℓv \mid \ell.

Then rr is automorphic.

Potential diagonalizability is a condition at ℓ\ell; it holds, for example, for crystalline representations with Hodge-Tate weights in a range [a,a+ℓ−2][a, a + \ell - 2] when Fv/ℚℓF_v/\mathbb{Q}_\ell is unramified.

Example (known cases of automorphy). Elliptic curves over ℚ\mathbb{Q} are modular (Wiles, 1995, and Taylor and Wiles, 1995, in the semistable case; Breuil, Conrad, Diamond and Taylor, 2001). Serre's modularity conjecture holds (Khare and Wintenberger, 2009), all symmetric powers of holomorphic newforms without CM are automorphic (Newton and Thorne, 2021), and potential automorphy holds in many cases over CM fields (Allen, Calegari, Caraiani and seven others, 2023).

In the window bound, HypothesesH' comes from published_hypotheses_from_patching_alt, and the counts of the cases II, III and V use automorphy of representations of dimensions four and three.

Further reading

  • J.-P. Serre, Abelian ℓ\ell-adic Representations and Elliptic Curves, chapter I.
  • J. Tate, Number theoretic background, in Automorphic Forms, Representations and L-functions, Proc. Sympos. Pure Math. 33 (1979), part 2, sections 1 to 4.
  • T. Wedhorn, The local Langlands correspondence for GL(n)\mathrm{GL}(n) over p-adic fields, ICTP Lecture Notes 21 (2008).
  • O. Brinon and B. Conrad, CMI Summer School Notes on p-adic Hodge Theory (2009).
  • T. Barnet-Lamb, T. Gee, D. Geraghty and R. Taylor, Potential automorphy and change of weight, Ann. of Math. 179 (2014), sections 1 and 2.

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