riemann

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

Periods

The Bessel and Fourier-Jacobi periods of unitary groups, local multiplicity one, the Gan-Gross-Prasad conjectures, descent, interlacing and orbit integrals. The Bessel seesaw chapter of the proof is built on them.

Prerequisites: The Weil representation and theta lifts.

A period of an automorphic form is its integral over a subgroup, often against a character or a theta function. The Gan-Gross-Prasad conjectures relate Bessel and Fourier-Jacobi periods of unitary groups to central L-values. The Bessel seesaw chapter uses a Bessel period, the method of Ginzburg, Rallis and Soudry run upwards, and real local multiplicity one with an interlacing condition.

Generic characters

Let FF be a number field, ψ\psi a nontrivial character of 𝔸F/F\mathbb{A}_F / F, and GG quasi-split over FF with a Borel subgroup B=TNB = TN.

Definition. A character ψN\psi_N of N(𝔸F)N(\mathbb{A}_F), trivial on N(F)N(F), is a generic character if it is nontrivial on each simple root subgroup. The Whittaker period of φ\varphi is ∫[N]φ(n) ψN(n)‾ dn\int_{[N]} \varphi(n)\, \overline{\psi_N(n)}\, dn, and π\pi is globally generic if this period is nonzero on π\pi.

For GLn\mathrm{GL}_n the standard generic character is ψN(u)=ψ(u12+u23+⋯+un−1,n)\psi_N(u) = \psi(u_{12} + u_{23} + \dots + u_{n-1,n}), as in the chapter on automorphic forms.

Let E/FE/F be a CM extension and W⊂VW \subset V nondegenerate spaces over EE, both Hermitian or both skew-Hermitian, in one of two cases:

  1. Bessel case: W⊥=X⊕L⊕X∗W^\perp = X \oplus L \oplus X^* has odd dimension 2r+12r + 1, with XX, X∗X^* isotropic of dimension rr and LL an anisotropic line;
  2. Fourier-Jacobi case: W⊥=X⊕X∗W^\perp = X \oplus X^* has even dimension 2r2r.

Let NN be the unipotent radical of the parabolic of U(V)\mathrm{U}(V) that stabilizes a full isotropic flag in XX, and H=U(W)⋉NH = \mathrm{U}(W) \ltimes N. In the Bessel case ψ\psi, a basis of XX and a vector of LL give a character λ\lambda of NN, trivial on N(F)N(F), generic on the root subgroups of XX and fixed by U(W)\mathrm{U}(W) (Gan, Gross and Prasad, 2012, Section 12). The Bessel period of the Bessel seesaw chapter uses such a character, besChar.

Bessel and Fourier-Jacobi periods

Let π\pi and σ\sigma be cuspidal automorphic representations of U(V)\mathrm{U}(V) and U(W)\mathrm{U}(W).

Definition. In the Bessel case, the Bessel period of φ∈π\varphi \in \pi and f∈σf \in \sigma is ℬ(φ,f)=∫[U(W)]∫[N]φ(nh) f(h)‾ λ(n)‾ dn dh, \mathcal{B}(\varphi, f) = \int_{[\mathrm{U}(W)]} \int_{[N]} \varphi(nh)\, \overline{f(h)}\, \overline{\lambda(n)}\, dn\, dh, which converges since φ\varphi decays rapidly. For r=0r = 0, that is dim⁡V=dim⁡W+1\dim V = \dim W + 1, it is the Gross-Prasad period ∫[U(W)]φ(h) f(h)‾ dh\int_{[\mathrm{U}(W)]} \varphi(h)\, \overline{f(h)}\, dh.

Definition. In the Fourier-Jacobi case, NN maps onto a Heisenberg group and HH has a Weil representation ν\nu with theta functions θ(⋅,ϕ)\theta(\cdot, \phi) (the chapter on theta lifts). The Fourier-Jacobi period is ℱ𝒥(φ,f,ϕ)=∫[U(W)]∫[N]φ(nh) f(h)‾ θ(nh,ϕ)‾ dn dh. \mathcal{FJ}(\varphi, f, \phi) = \int_{[\mathrm{U}(W)]} \int_{[N]} \varphi(nh)\, \overline{f(h)}\, \overline{\theta(nh, \phi)}\, dn\, dh.

In the Bessel seesaw chapter the functional Bplus is a global Bessel period, and its bottom period P0(F,u)P_0(F, u), an integral over [U(V0,0)][\mathrm{U}(V_{0,0})], is a Gross-Prasad period.

Local multiplicity one

Now let FF be local, ν=λ\nu = \lambda in the Bessel case and ν\nu the Weil representation of HH in the Fourier-Jacobi case, and π\pi, σ\sigma irreducible smooth representations of U(V)\mathrm{U}(V), U(W)\mathrm{U}(W).

Definition. A Bessel functional (or Fourier-Jacobi functional) is an element of HomH(π⊗σˉ⊗νˉ,ℂ)\mathrm{Hom}_H(\pi \otimes \bar\sigma \otimes \bar\nu, \mathbb{C}).

Theorem (local multiplicity one; Aizenbud, Gourevitch, Rallis and Schiffmann, 2010; Sun and Zhu, 2012; Gan, Gross and Prasad, 2012). For all π\pi and σ\sigma, dim⁡HomH(π⊗σˉ⊗νˉ,ℂ)≤1\dim \mathrm{Hom}_H(\pi \otimes \bar\sigma \otimes \bar\nu, \mathbb{C}) \le 1.

The first two papers treat r=0r = 0 (pp-adic and archimedean); Gan, Gross and Prasad reduce the general pp-adic case to it, and Jiang, Sun and Zhu (2010) treat archimedean Bessel models. So on factorizable vectors a global Bessel period is a constant times a product of local Bessel functionals. The condition BurGapPos of the Bessel seesaw chapter is the bound for Bessel functionals at the real place, as a rank at most 11.

The Gan-Gross-Prasad conjectures

Theorem (local Gan-Gross-Prasad conjecture; Beuzart-Plessis, 2016 and 2020; Gan and Ichino, 2016). Let (π,σ)(\pi, \sigma) run over the representations of the relevant pure inner forms U(Vα)×U(Wα)\mathrm{U}(V_\alpha) \times \mathrm{U}(W_\alpha) in the Vogan L-packet of a tempered L-parameter. Then HomH(π⊗σˉ⊗νˉ,ℂ)≠0\mathrm{Hom}_H(\pi \otimes \bar\sigma \otimes \bar\nu, \mathbb{C}) \ne 0 for exactly one pair, and an explicit character built from local root numbers determines it (local Gan-Gross-Prasad conjecture).

Beuzart-Plessis proved the Bessel case; Gan and Ichino deduced the Fourier-Jacobi case by theta lifts.

Theorem (global Gan-Gross-Prasad conjecture; Beuzart-Plessis, Liu, Zhang and Zhu, 2021; Beuzart-Plessis, Chaudouard and Zydor, 2022). Let dim⁡V=dim⁡W+1\dim V = \dim W + 1, and let π\pi, σ\sigma be cuspidal automorphic representations of U(V)\mathrm{U}(V), U(W)\mathrm{U}(W) whose base changes Π\Pi, Σ\Sigma to GL\mathrm{GL} are isobaric sums of distinct cuspidal representations. Then L(12,Π×Σ)≠0L(\frac12, \Pi \times \Sigma) \ne 0 if and only if the Gross-Prasad period is nonzero on some π′⊗σ′\pi' \otimes \sigma', where W′⊂V′W' \subset V' are Hermitian of codimension one, U(V′)×U(W′)\mathrm{U}(V') \times \mathrm{U}(W') is an inner form of U(V)×U(W)\mathrm{U}(V) \times \mathrm{U}(W), and (π′,σ′)(\pi', \sigma') is cuspidal on it with base changes Π\Pi and Σ\Sigma.

This is the global Gan-Gross-Prasad conjecture for unitary groups in codimension one. Its refined form (Ichino and Ikeda; N. Harris for unitary groups) writes the squared Gross-Prasad period as the central L-value times local integrals of matrix coefficients. The statement BesselPeriodGGP of the Bessel seesaw chapter, local GGP at the real place, lies in the formal layer, outside the spine of the proof.

Interlacing and the branching law for discrete series

By the Weyl branching law of the chapter on representations, the irreducible representation of U(n)\mathrm{U}(n) of highest weight λ\lambda restricts to U(n−1)\mathrm{U}(n-1) as the multiplicity-free sum of those of highest weight μ\mu with λ1≥μ1≥λ2≥μ2≥⋯≥μn−1≥λn. \lambda_1 \ge \mu_1 \ge \lambda_2 \ge \mu_2 \ge \dots \ge \mu_{n-1} \ge \lambda_n.

Definition. Sequences a1>⋯>ana_1 > \dots > a_n and b1>⋯>bn−1b_1 > \dots > b_{n-1} are interlaced if a1>b1>a2>⋯>bn−1>ana_1 > b_1 > a_2 > \dots > b_{n-1} > a_n.

A discrete series representation of U(p,q)\mathrm{U}(p, q) has a Harish-Chandra parameter: p+qp + q distinct numbers in p+q−12+ℤ\frac{p+q-1}{2} + \mathbb{Z}, each signed by whether it belongs to the pp part or the qq part.

Theorem (branching law for discrete series; He, 2017). Let W⊂VW \subset V be Hermitian spaces for ℂ/ℝ\mathbb{C}/\mathbb{R} of codimension one, and π\pi, σ\sigma discrete series of U(V)\mathrm{U}(V), U(W)\mathrm{U}(W). Then HomU(W)(π⊗σˉ,ℂ)≠0\mathrm{Hom}_{\mathrm{U}(W)}(\pi \otimes \bar\sigma, \mathbb{C}) \ne 0 if and only if the signed Harish-Chandra parameters of π\pi and σ\sigma satisfy an explicit combinatorial condition, the one that local GGP predicts.

For compact groups the condition is interlacing of the ρ\rho-shifted highest weights. The predicate Interlaced of the Bessel seesaw chapter is a sign form: merge the two parameters in decreasing order, with signs, and ask that neighbours have equal signs exactly when they come from different lists.

The descent of Ginzburg, Rallis and Soudry

Theorem (Ginzburg, Rallis and Soudry, 2011). Let Π=Π1⊞⋯⊞Πk\Pi = \Pi_1 \boxplus \dots \boxplus \Pi_k be an isobaric sum of distinct conjugate self-dual cuspidal Πi\Pi_i on GLni(𝔸E)\mathrm{GL}_{n_i}(\mathbb{A}_E) whose Asai L-functions L(s,Πi,Asϵ)L(s, \Pi_i, \mathrm{As}^{\epsilon}) have a pole at s=1s = 1, for a sign ϵ\epsilon depending only on n=∑nin = \sum n_i. Then a quasi-split Un\mathrm{U}_n has a globally generic cuspidal σ\sigma with base change Π\Pi, spanned by Bessel or Fourier-Jacobi coefficients of a residue of an Eisenstein series induced from Π\Pi on a larger unitary group.

This is the descent of Ginzburg, Rallis and Soudry. The Bessel seesaw chapter runs it backwards as an ascent, with rungs Rung0, RungB, RungJ between Fourier-Jacobi and Bessel periods; HypothesesH' of the window bound has a weak descent input DescWeak.

Orbit integrals and the relative trace formula

Let HH be a group over FF acting on a variety XX, and x∈X(F)x \in X(F) with stabilizer HxH_x.

Definition. The orbit integral of Φ\Phi on X(𝔸F)X(\mathbb{A}_F) at xx is ∫Hx(𝔸F)\H(𝔸F)Φ(h−1x) dh\int_{H_x(\mathbb{A}_F) \backslash H(\mathbb{A}_F)} \Phi(h^{-1} x)\, dh.

Example. Let U(V)\mathrm{U}(V) act on VnV^n, and let TT be Hermitian of size nn with det⁡T≠0\det T \ne 0. By Witt's theorem the x∈V(F)nx \in V(F)^n with Gram matrix TT form at most one U(V)(F)\mathrm{U}(V)(F)-orbit, so the TT-th Fourier coefficient of the theta integral of the chapter on theta lifts is one orbit integral (or 00).

Definition (Jacquet). Let H1,H2⊆GH_1, H_2 \subseteq G, f∈Cc∞(G(𝔸F))f \in C_c^\infty(G(\mathbb{A}_F)) and Kf(x,y)=∑γ∈G(F)f(x−1γy)K_f(x, y) = \sum_{\gamma \in G(F)} f(x^{-1}\gamma y). The relative trace formula equates two expansions of ∫[H1]∫[H2]Kf(h1,h2) dh1 dh2\int_{[H_1]} \int_{[H_2]} K_f(h_1, h_2)\, dh_1\, dh_2: a spectral sum over π\pi of products of an H1H_1-period and an H2H_2-period, and a geometric sum over H1(F)\G(F)/H2(F)H_1(F) \backslash G(F) / H_2(F) of relative orbit integrals.

The proofs of the global GGP conjecture compare two relative trace formulas, following Jacquet and Rallis (2011). This proof uses none and reaches periods by theta lifts and seesaw identities, but its orb_T in the Siegel-Weil chapter, a sum over the XX with gram(X,X)=−2T\mathrm{gram}(X, X) = -2T averaged over a fundamental domain, is an orbit integral.

Further reading

  • W. T. Gan, B. Gross and D. Prasad, Symplectic local root numbers, central critical L-values, and restriction problems in the representation theory of classical groups, Astérisque 346 (2012), Sections 2 to 4, 12 to 17 and 24.
  • D. Ginzburg, S. Rallis and D. Soudry, The Descent Map from Automorphic Representations of GL(n) to Classical Groups (2011), Chapters 2 and 3.
  • H. Jacquet and S. Rallis, On the Gross-Prasad conjecture for unitary groups, in Clay Math. Proc. 13 (2011).
  • R. Beuzart-Plessis, P.-H. Chaudouard and M. Zydor, The global Gan-Gross-Prasad conjecture for unitary groups: the endoscopic case, Publ. Math. IHES 135 (2022), Section 1.

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