riemann

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

The Weil representation and theta lifts

The Weil representation of a reductive dual pair, the local and global theta correspondence, and the Siegel-Weil and Rallis inner product formulas. The theta matching and Siegel-Weil chapters of the proof use them.

Prerequisites: Automorphic forms.

The Weil representation of the metaplectic group, restricted to two groups that centralize each other, gives the theta correspondence, and its theta kernels carry automorphic forms from one group to the other. The proof builds the representations of the theta matching chapter as theta lifts and computes them with the Siegel-Weil and Rallis inner product formulas.

Reductive dual pairs

Let FF be a field of characteristic 00 and WW a symplectic space over FF (the chapter on classical groups defines Sp(W)\mathrm{Sp}(W)).

Definition (Howe). A reductive dual pair in Sp(W)\mathrm{Sp}(W) is a pair (G,G′)(G, G') of reductive subgroups, each the centralizer of the other. It is irreducible if WW is not an orthogonal sum of two nonzero G⋅G′G \cdot G'-stable subspaces.

Let EE be FF (with trivial conjugation) or a quadratic extension of FF, and ϵ=±1\epsilon = \pm 1. Let VV be an ϵ\epsilon-Hermitian and V′V' a (−ϵ)(-\epsilon)-Hermitian space over EE. Then W=V⊗EV′W = V \otimes_E V' with the form ⟨v⊗v′,u⊗u′⟩=Tr⁡E/F((v,u) (v′,u′)′‾) \langle v \otimes v', u \otimes u' \rangle = \operatorname{Tr}_{E/F}\big((v, u)\, \overline{(v', u')'}\big) is symplectic over FF, and the isometry groups of VV and V′V' form a dual pair, of type I. (Quaternion algebras give the remaining type I pairs.)

Example. E=FE = F with VV quadratic and V′V' symplectic of dimension 2n2n gives (O(V),Sp2n)(\mathrm{O}(V), \mathrm{Sp}_{2n}); E/FE/F quadratic with VV Hermitian and V′V' skew-Hermitian gives (U(V),U(V′))(\mathrm{U}(V), \mathrm{U}(V')).

IsTriple in ThetaIntegralsAgree fixes Hermitian VV, V′V' of signatures (a−p,p)(a - p, p), (q,a′−q)(q, a' - q) and a skew-Hermitian WW of dimension a+a′a + a', hence the dual pairs (U(V),U(W))(\mathrm{U}(V), \mathrm{U}(W)) and (U(V′),U(W))(\mathrm{U}(V'), \mathrm{U}(W)).

The Stone-von Neumann theorem

Let FF be local and ψ\psi a nontrivial character of FF. The Heisenberg group H(W)H(W) (the chapter on classical groups) is W×FW \times F with (w,t)(w′,t′)=(w+w′,t+t′+12⟨w,w′⟩)(w, t)(w', t') = (w + w', t + t' + \frac12 \langle w, w' \rangle); its center is FF.

Theorem (Stone-von Neumann). Up to isomorphism, H(W)H(W) has exactly one irreducible smooth representation ρψ\rho_\psi (unitary when FF is archimedean) with central character ψ\psi.

As Sp(W)\mathrm{Sp}(W) acts on H(W)H(W) by g(w,t)=(gw,t)g(w, t) = (gw, t), fixing the center, ρψ∘g≅ρψ\rho_\psi \circ g \cong \rho_\psi, so some ω(g)\omega(g), unique up to a scalar, satisfies ω(g) ρψ(h) ω(g)−1=ρψ(gh)\omega(g)\, \rho_\psi(h)\, \omega(g)^{-1} = \rho_\psi(g h). This gives a projective representation ω\omega of Sp(W)\mathrm{Sp}(W). The theorem is stoneVonNeumann in ThetaLocalInputs.

The Weil representation

Let Mp(W)\mathrm{Mp}(W) be the metaplectic double cover of Sp(W)\mathrm{Sp}(W) (the chapter on classical groups).

Theorem (Weil, 1964). The projective representation ω\omega lifts to a genuine representation ωψ\omega_\psi of Mp(W)\mathrm{Mp}(W), unique up to isomorphism. When F≠ℂF \ne \mathbb{C} the cover is nontrivial and ωψ\omega_\psi does not come from Sp(W)\mathrm{Sp}(W).

Definition. ωψ\omega_\psi is the Weil representation (or oscillator representation) of Mp(W)\mathrm{Mp}(W). It is the sum of two irreducible representations. ThetaLocalInputs gives it over pp-adic fields and ℝ\mathbb{R} (weilRepresentationExists, metaplecticRepresentation, realMetaplecticRepresentation).

The Schrödinger model

A Lagrangian decomposition W=X⊕YW = X \oplus Y realizes ωψ\omega_\psi on the Schwartz space S(Y)S(Y). Let VV be a quadratic space of dimension mm and Sp2n\mathrm{Sp}_{2n} the isometry group of the standard symplectic F2nF^{2n}, so W=Vn⊕VnW = V^n \oplus V^n. For x∈Vnx \in V^n let (x,x)=((xi,xj))i,j(x, x) = ((x_i, x_j))_{i,j} be the Gram matrix, and let χV\chi_V be the quadratic character of the discriminant of VV. Put m(a)=diag(a,ta−1)m(a) = \mathrm{diag}(a, {}^t a^{-1}) for a∈GLn(F)a \in \mathrm{GL}_n(F), n(b)=(1b01)n(b) = \begin{pmatrix} 1 & b \\ 0 & 1 \end{pmatrix} for symmetric bb, and w=(01−10)w = \begin{pmatrix} 0 & 1 \\ -1 & 0 \end{pmatrix}.

Definition (Schrödinger model). The Schrödinger model of the Weil representation of O(V)×Sp2n\mathrm{O}(V) \times \mathrm{Sp}_{2n} is the action on S(Vn)S(V^n) given by ω(m(a))φ(x)=χV(det⁡a) ∣det⁡a∣m/2φ(xa),ω(n(b))φ(x)=ψ(12tr⁡(b (x,x)))φ(x), \omega(m(a))\varphi(x) = \chi_V(\det a)\, |\det a|^{m/2} \varphi(xa), \qquad \omega(n(b))\varphi(x) = \psi\big(\tfrac12 \operatorname{tr}(b\, (x, x))\big) \varphi(x), ω(w)φ(x)=γ φ^(x),ω(h)φ(x)=φ(h−1x)(h∈O(V)), \omega(w)\varphi(x) = \gamma\, \hat\varphi(x), \qquad \omega(h)\varphi(x) = \varphi(h^{-1} x) \quad (h \in \mathrm{O}(V)), with φ^\hat\varphi the Fourier transform for the self-dual measure and γ\gamma an eighth root of unity (the Weil index). For mm even this is a representation of Sp2n(F)\mathrm{Sp}_{2n}(F); for mm odd, with χV\chi_V corrected by Weil indices, of Mp2n(F)\mathrm{Mp}_{2n}(F).

For Hermitian VV over EE and the pair (U(V),U(n,n))(\mathrm{U}(V), \mathrm{U}(n, n)) the same formulas hold, with a∈GLn(E)a \in \mathrm{GL}_n(E), b∈Hermnb \in \mathrm{Herm}_n, ∣det⁡a∣E|\det a|_E for ∣det⁡a∣|\det a|, and a splitting character χ\chi (defined below) for χV\chi_V.

Definition. For FF pp-adic, ψ\psi of conductor 𝒪F\mathcal{O}_F and L⊂VL \subset V a self-dual lattice, the vacuum vector of S(Vn)S(V^n) is the characteristic function of LnL^n; for F=ℝF = \mathbb{R} and VV positive definite it is the Gaussian φ0(x)=e−πtr⁡(x,x)\varphi^0(x) = e^{-\pi \operatorname{tr}(x, x)}. In both cases the inverse image in Mp\mathrm{Mp} of a maximal compact subgroup acts on it by a character.

These formulas are TPar and TWeyl in the Siegel-Weil chapter, whose MetaplecticRepresentation uses the Gaussian e−πx⋅xe^{-\pi x \cdot x}; LocalMatching matches vacuum vectors outside a finite set.

The Fock model

Let F=ℝF = \mathbb{R}, W=ℝ2nW = \mathbb{R}^{2n}, K≅U(n)K \cong \mathrm{U}(n) the maximal compact subgroup of Sp2n(ℝ)\mathrm{Sp}_{2n}(\mathbb{R}) and K~\tilde K its inverse image in Mp2n(ℝ)\mathrm{Mp}_{2n}(\mathbb{R}).

Definition. The Fock space ℱn\mathcal{F}_n is the Hilbert space of holomorphic ff on ℂn\mathbb{C}^n with ∫ℂn∣f(z)∣2e−π∣z∣2 dz<∞\int_{\mathbb{C}^n} |f(z)|^2 e^{-\pi |z|^2}\, dz < \infty. The Bargmann transform is a unitary isomorphism L2(ℝn)→ℱnL^2(\mathbb{R}^n) \to \mathcal{F}_n that sends the Gaussian e−πx⋅xe^{-\pi x \cdot x} to a constant.

Definition (Fock model). The Fock model is the transport of the Weil representation to ℱn\mathcal{F}_n by the Bargmann transform.

Its K~\tilde K-finite vectors are the polynomials, with K~\tilde K preserving degree and vacuum vector 11, and 𝔰𝔭2n(ℂ)\mathfrak{sp}_{2n}(\mathbb{C}) acts on them by zizjz_i z_j, ∂i∂j\partial_i \partial_j and zi∂j+12δijz_i \partial_j + \frac12 \delta_{ij} (up to constants). LocalMatching puts the real Lie algebra and K∞K_\infty actions on this (𝔰𝔭2n,K~)(\mathfrak{sp}_{2n}, \tilde K)-module, named Pol.

Splitting characters

Let E/FE/F be a quadratic extension of local or global fields, with quadratic character η\eta of F×F^\times or 𝔸F×/F×\mathbb{A}_F^\times / F^\times. Let VV be Hermitian of dimension mm and V′V' skew-Hermitian of dimension nn over EE.

Theorem (Kudla, 1994; Harris, Kudla and Sweet, 1996). A character χ\chi of E×E^\times (or 𝔸E×/E×\mathbb{A}_E^\times / E^\times) with χ∣F×=ηm\chi|_{F^\times} = \eta^m determines a splitting U(V′)→Mp(V⊗V′)\mathrm{U}(V') \to \mathrm{Mp}(V \otimes V'). A pair (χ,χ′)(\chi, \chi') with χ∣F×=ηm\chi|_{F^\times} = \eta^m and χ′∣F×=ηn\chi'|_{F^\times} = \eta^n determines a splitting over U(V)×U(V′)\mathrm{U}(V) \times \mathrm{U}(V').

Definition. Such (χ,χ′)(\chi, \chi') are splitting characters of the Weil representation; they make ωψ\omega_\psi a representation ωψ,χ,χ′\omega_{\psi, \chi, \chi'} of U(V)×U(V′)\mathrm{U}(V) \times \mathrm{U}(V').

The theorem is kudlaSplittingSplitUnitary in ThetaLocalInputs. The Weil representations om and om' of IsTriple have splitting characters (β,β/α)(\beta, \beta/\alpha) and (α,α/β)(\alpha, \alpha/\beta), and TLiftK calls the restriction condition ResEtaPow.

Theta kernels

Let FF be a number field and (G,G′)(G, G') a type I dual pair over FF with splitting characters. The global Weil representation is the restricted tensor product ω=⨂v′ωv\omega = \bigotimes'_v \omega_v on S(V(𝔸F)n)S(V(\mathbb{A}_F)^n), along the vacuum vectors.

Theorem (Weil, 1964). The form Θ(φ)=∑ξ∈V(F)nφ(ξ)\Theta(\varphi) = \sum_{\xi \in V(F)^n} \varphi(\xi) satisfies Θ(ω(γ)φ)=Θ(φ)\Theta(\omega(\gamma)\varphi) = \Theta(\varphi) for γ∈G(F)×G′(F)\gamma \in G(F) \times G'(F). For the Weyl element ww this is Poisson summation.

Definition. The theta kernel of φ∈S(V(𝔸F)n)\varphi \in S(V(\mathbb{A}_F)^n) is θ(g,h;φ)=∑ξ∈V(F)n(ω(g,h)φ)(ξ),g∈G(𝔸F), h∈G′(𝔸F), \theta(g, h; \varphi) = \sum_{\xi \in V(F)^n} (\omega(g, h)\varphi)(\xi), \qquad g \in G(\mathbb{A}_F),\ h \in G'(\mathbb{A}_F), a smooth function of moderate growth on [G]×[G′][G] \times [G'], where [G]=G(F)\G(𝔸F)[G] = G(F) \backslash G(\mathbb{A}_F).

Example. For V=ℚV = \mathbb{Q} with the form x2x^2 (the group is Mp2\mathrm{Mp}_2), the Gaussian at ∞\infty and 𝟏ℤp\mathbf{1}_{\mathbb{Z}_p} at each pp, the theta kernel at n(x) m(y)n(x)\, m(\sqrt{y}) is y1/4y^{1/4} times the theta function ∑neπin2τ\sum_n e^{\pi i n^2 \tau}, τ=x+iy\tau = x + iy, of the chapter on complex analysis.

The kernel is thetaSum in ThetaIntegralsAgree, its invariance thetaRationalInvariance in ThetaLocalInputs, and Poisson summation TTheta in the Siegel-Weil chapter.

The local theta correspondence

Let FF be local, (G,G′)(G, G') a type I dual pair with splitting characters, and π\pi an irreducible smooth representation of GG.

Definition. The maximal π\pi-isotypic quotient of ω\omega is π⊗Θ(π)\pi \otimes \Theta(\pi) for a smooth representation Θ(π)\Theta(\pi) of G′G', possibly zero, the big theta lift of π\pi.

Theorem (Howe duality; Howe, 1989; Waldspurger, 1990; Gan and Takeda, 2016; Gan and Sun, 2017). If Θ(π)≠0\Theta(\pi) \ne 0, it has a unique irreducible quotient θ(π)\theta(\pi). If θ(π1)≅θ(π2)≠0\theta(\pi_1) \cong \theta(\pi_2) \ne 0, then π1≅π2\pi_1 \cong \pi_2.

Definition. The map π↦θ(π)\pi \mapsto \theta(\pi) is the theta correspondence for (G,G′)(G, G'), a bijection between the irreducible representations of GG and of G′G' that occur as quotients of ω\omega.

For pp odd and unramified data, an unramified π\pi with θ(π)≠0\theta(\pi) \ne 0 has unramified θ(π)\theta(\pi), with explicit Satake parameters (Howe; Rallis). Howe duality is gLHoweDuality and howeDualityReal in ThetaLocalInputs; LocalMatching asks that two local theta lifts share an irreducible quotient Πv\Pi_v.

Theta towers

Let V0V_0 be an anisotropic Hermitian space and Vr=V0⊕ℍrV_r = V_0 \oplus \mathbb{H}^r, with ℍ\mathbb{H} the hyperbolic plane; the VrV_r form a Witt tower. Fix a skew-Hermitian WW and an irreducible π\pi of U(W)\mathrm{U}(W).

Definition. The theta tower of π\pi is the sequence of lifts ΘVr(π)\Theta_{V_r}(\pi) to U(Vr)\mathrm{U}(V_r). The first occurrence of π\pi is the least dim⁡Vr\dim V_r with ΘVr(π)≠0\Theta_{V_r}(\pi) \ne 0.

Theorem (persistence; Kudla, 1986). If ΘVr(π)≠0\Theta_{V_r}(\pi) \ne 0, then ΘVr+1(π)≠0\Theta_{V_{r+1}}(\pi) \ne 0.

Theorem (conservation; Sun and Zhu, 2015). Let FF be pp-adic and V±V^\pm the two Witt towers of Hermitian spaces of a fixed dimension parity. The first occurrences of π\pi in them satisfy n+(π)+n−(π)=2dim⁡W+2n^+(\pi) + n^-(\pi) = 2 \dim W + 2.

Each of the steps StepAlpha and StepBeta of the theta matching chapter splits off a hyperbolic plane: one place up a Witt tower.

The global theta lift

Let FF be a number field, (G,G′)=(U(V),U(W))(G, G') = (\mathrm{U}(V), \mathrm{U}(W)) with splitting characters, and π\pi a cuspidal automorphic representation of G′G'.

Definition. For f∈πf \in \pi and φ∈S(V(𝔸F)n)\varphi \in S(V(\mathbb{A}_F)^n), the theta lift of ff is θ(φ,f)(g)=∫[G′]θ(g,h;φ) f(h)‾ dh,g∈G(𝔸F), \theta(\varphi, f)(g) = \int_{[G']} \theta(g, h; \varphi)\, \overline{f(h)}\, dh, \qquad g \in G(\mathbb{A}_F), which converges since ff decays rapidly. The global theta lift ΘV(π)\Theta_V(\pi) is the span of these automorphic forms on GG.

Theorem (Rallis, 1984). In a Witt tower VrV_r, the first nonzero global lift ΘVr(π)\Theta_{V_r}(\pi) is cuspidal.

Theorem (Kudla and Rallis; Moeglin, 1997). If ΘV(π)\Theta_V(\pi) is nonzero and cuspidal, it is irreducible and ΘV(π)≅⨂v′θ(πv)\Theta_V(\pi) \cong \bigotimes'_v \theta(\pi_v).

In ThetaIntegralsAgree, IM(Φ)(h)=∫domθ(g,h;Φ) dμ(g)I_M(\Phi)(h) = \int_{\mathrm{dom}} \theta(g, h; \Phi)\, d\mu(g) lifts the constant function over a fundamental domain; TLiftK lifts a definite automorphic representation of Un\mathrm{U}_n to Ud\mathrm{U}_d.

Seesaw pairs

Definition. A seesaw pair is two dual pairs (G,G′)(G, G') and (H,H′)(H, H') in one Sp(W)\mathrm{Sp}(W) with H⊆GH \subseteq G and G′⊆H′G' \subseteq H'.

Example. For V=V1⊕V2V = V_1 \oplus V_2 Hermitian and WW skew-Hermitian, (U(V),U(W))(\mathrm{U}(V), \mathrm{U}(W)) and (U(V1)×U(V2),U(W)×U(W))(\mathrm{U}(V_1) \times \mathrm{U}(V_2), \mathrm{U}(W) \times \mathrm{U}(W)) form a seesaw pair, with U(W)\mathrm{U}(W) embedded diagonally.

Proposition (seesaw identity). For cusp forms ff on HH and f′f' on G′G', ∫[H]θ(φ,f′)(h) f(h)‾ dh=∫[G′]θ(φ,f)(g) f′(g)‾ dg, \int_{[H]} \theta(\varphi, f')(h)\, \overline{f(h)}\, dh = \int_{[G']} \theta(\varphi, f)(g)\, \overline{f'(g)}\, dg, with f′f' lifted to GG on the left and ff lifted to H′H' on the right. Both sides are ∬θ(h,g;φ) f′(g)‾ f(h)‾ dg dh\iint \theta(h, g; \varphi)\, \overline{f'(g)}\, \overline{f(h)}\, dg\, dh, integrated in the two orders; this exchange is the step upLast of the Bessel seesaw chapter.

The Siegel-Weil formula

Let WW be skew-Hermitian of dimension nn over EE.

Definition. The doubled group of U(W)\mathrm{U}(W) is U(W□)\mathrm{U}(W^\square), W□=W⊕(−W)W^\square = W \oplus (-W). It is isomorphic to U(n,n)\mathrm{U}(n, n) and contains U(W)×U(W)\mathrm{U}(W) \times \mathrm{U}(W). The diagonal WΔW^\Delta is Lagrangian, and its stabilizer is the Siegel parabolic PP.

Let VV be Hermitian of dimension mm and χ\chi a Hecke character with χ∣𝔸F×=ηm\chi|_{\mathbb{A}_F^\times} = \eta^m. For Φ∈S(V(𝔸F)n)\Phi \in S(V(\mathbb{A}_F)^n), fΦ(g)=(ω(g)Φ)(0)f_\Phi(g) = (\omega(g)\Phi)(0) lies in I(s0,χ)I(s_0, \chi) on U(n,n)\mathrm{U}(n, n), s0=(m−n)/2s_0 = (m - n)/2. Let fΦ,sf_{\Phi, s} be the standard section through fΦf_\Phi, and E(g,s,fΦ)E(g, s, f_\Phi) the Siegel Eisenstein series of the chapter on automorphic forms.

Theorem (Siegel-Weil formula; Weil, 1965; Kudla and Rallis, 1988; Ichino, 2004 and 2007). Let VV be anisotropic and Φ\Phi KK-finite. The theta integral I(g,Φ)=∫[U(V)]θ(g,h;Φ) dhI(g, \Phi) = \int_{[\mathrm{U}(V)]} \theta(g, h; \Phi)\, dh, with vol([U(V)])=1\mathrm{vol}([\mathrm{U}(V)]) = 1, converges absolutely, E(g,s,fΦ)E(g, s, f_\Phi) is holomorphic at s0s_0, and E(g,s0,fΦ)=κ I(g,Φ),κ={1m>n,2m≤n. E(g, s_0, f_\Phi) = \kappa\, I(g, \Phi), \qquad \kappa = \begin{cases} 1 & m > n, \\ 2 & m \le n. \end{cases}

On TT-th Fourier coefficients it is Siegel's mass formula: an orbit integral (defined in Chapter 17) over the vectors with Gram matrix TT equals a product of local densities.

Remark (regularized Siegel-Weil formula). When Weil's convergence condition fails, Kudla and Rallis (1994) regularize the theta integral with a Hecke operator. The regularized Siegel-Weil formula gives the leading Laurent coefficients of E(g,s,fΦ)E(g, s, f_\Phi) at s0s_0 by regularized theta integrals of VV and of a complementary space (Ichino, 2004; Yamana, 2013; Gan, Qiu and Takeda, 2014). The proof does not need it.

SiegelWeilStd in the Siegel-Weil chapter is the formula for definite VV on G2=U(Y⊕−Y)G_2 = \mathrm{U}(Y \oplus -Y), with cc for κ\kappa; SiegelWeilUnitaryConvergent adds the Siegel densities g∞(T)∏vgv(T)g_\infty(T) \prod_v g_v(T).

The doubling method

Let G=U(W)G = \mathrm{U}(W), dim⁡W=n\dim W = n, and ι:G×G→U(W□)\iota : G \times G \to \mathrm{U}(W^\square) the inclusion. G×GG \times G has an open orbit on P\U(W□)P \backslash \mathrm{U}(W^\square), with representative δ\delta.

Definition. For π\pi cuspidal automorphic on GG, φ1,φ2∈π\varphi_1, \varphi_2 \in \pi and fsf_s a section of I(s,χ)I(s, \chi), the doubling zeta integral is Z(s,φ1,φ2,fs)=∫[G×G]E(ι(g1,g2),s,fs) φ1(g1) φ2(g2)‾ χ−1(det⁡g2) dg1 dg2. Z(s, \varphi_1, \varphi_2, f_s) = \int_{[G \times G]} E(\iota(g_1, g_2), s, f_s)\, \varphi_1(g_1)\, \overline{\varphi_2(g_2)}\, \chi^{-1}(\det g_2)\, dg_1\, dg_2.

Theorem (doubling method; Piatetski-Shapiro and Rallis, 1987). For Re⁡(s)\operatorname{Re}(s) large, unfolding gives Z(s,φ1,φ2,fs)=∫G(𝔸F)fs(δ ι(g,1)) ⟨π(g)φ1,φ2⟩ dg, Z(s, \varphi_1, \varphi_2, f_s) = \int_{G(\mathbb{A}_F)} f_s(\delta\, \iota(g, 1))\, \langle \pi(g)\varphi_1, \varphi_2 \rangle\, dg, with the Petersson inner product. For factorizable data this is a product of local zeta integrals Zv(s,fv,ξv)=∫G(Fv)fs,v(δ ι(g,1)) ξv(g) dgZ_v(s, f_v, \xi_v) = \int_{G(F_v)} f_{s,v}(\delta\, \iota(g, 1))\, \xi_v(g)\, dg of matrix coefficients ξv\xi_v of πv\pi_v. At a finite place with unramified spherical data, Zv(s)=L(s+12,πv×χv)dv(s), Z_v(s) = \frac{L(s + \frac12, \pi_v \times \chi_v)}{d_v(s)}, with dv(s)d_v(s) an explicit product of abelian L-factors.

These are doublingMethod and unramifiedDoublingGL in ThetaLocalInputs; the tool dbl of the determinants chapter uses the integral at s0=(d−n)/2s_0 = (d - n)/2.

The Rallis inner product formula

The Siegel-Weil formula for VV and W□W^\square at Φ=φ1⊗φˉ2\Phi = \varphi_1 \otimes \bar\varphi_2, where fΦ(δ ι(g,1))=⟨ω(g)φ1,φ2⟩f_\Phi(\delta\, \iota(g, 1)) = \langle \omega(g)\varphi_1, \varphi_2 \rangle, turns the doubling integral into an inner product of theta lifts.

Theorem (Rallis inner product formula; Rallis, 1984; Kudla and Rallis, 1994; Gan, Qiu and Takeda, 2014). Let π\pi be cuspidal automorphic on U(W)\mathrm{U}(W), dim⁡W=n\dim W = n, VV anisotropic Hermitian of dimension mm, f1,f2∈πf_1, f_2 \in \pi and φ1,φ2∈S(V(𝔸F)n)\varphi_1, \varphi_2 \in S(V(\mathbb{A}_F)^n) factorizable, and SS a finite set of places outside of which all data are unramified. Then ⟨θ(φ1,f1),θ(φ2,f2)⟩=κ LS(s0+12,π×χ)dS(s0)∏v∈SZv(s0),s0=m−n2, \langle \theta(\varphi_1, f_1), \theta(\varphi_2, f_2) \rangle = \kappa\, \frac{L^S(s_0 + \frac12, \pi \times \chi)}{d^S(s_0)} \prod_{v \in S} Z_v(s_0), \qquad s_0 = \frac{m - n}{2}, where Zv(s0)=∫U(W)(Fv)⟨ωv(g)φ1,v,φ2,v⟩ ⟨πv(g)f1,v,f2,v⟩‾ dgZ_v(s_0) = \int_{\mathrm{U}(W)(F_v)} \langle \omega_v(g)\varphi_{1,v}, \varphi_{2,v} \rangle\, \overline{\langle \pi_v(g) f_{1,v}, f_{2,v} \rangle}\, dg and κ\kappa is the Siegel-Weil constant for VV and W□W^\square.

It reduces the nonvanishing of ΘV(π)\Theta_V(\pi) to that of an L-value and of local zeta integrals. The lemma rallis_of_dm of the Bessel seesaw chapter is the field rallisInnerProductConvergent, and Rallis_k of the determinants chapter proves one direction.

Further reading

  • S. Kudla, Notes on the local theta correspondence (1996), Chapters I to III.
  • C. Moeglin, M.-F. Vignéras and J.-L. Waldspurger, Correspondances de Howe sur un corps p-adique, LNM 1291 (1987), Chapters 2 and 3.
  • S. Gelbart, I. Piatetski-Shapiro and S. Rallis, Explicit Constructions of Automorphic L-functions, LNM 1254 (1987), Part A.
  • W. T. Gan, Y. Qiu and S. Takeda, The regularized Siegel-Weil formula (second term identity) and the Rallis inner product formula, Invent. Math. 198 (2014).

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