---
title: Cubic Kazhdan–Patterson Cover
url: https://www.emergentmind.com/topics/cubic-kazhdan-patterson-cover
type: topic
---

# Cubic Kazhdan–Patterson Cover

Searching arXiv for the cited works to ground the article in current literature.
The cubic Kazhdan–Patterson cover is the threefold metaplectic central extension of \(\GL_2\), and more generally of \(\GL_r\), by \(\mu_3\) that arises in the Kazhdan–Patterson and Kubota framework. In rank \(2\), it is realized concretely as a central extension \(1\to \mu_3\to \widetilde G^{(c)}\to \GL_2(F)\to 1\) depending on a parameter \(c\in \mathbf Z/3\mathbf Z\), and it provides the metaplectic setting for exceptional theta representations, local and global co-periods, and a conjectural period formula for \(L(1/2,\pi,\Sym^3)\) [2507.15279]. In higher rank, recent work treats the cubic case as a tame Kazhdan–Patterson cover of \(\GL_r\) with explicit Hecke-algebraic, Whittaker-theoretic, derivative, and wavefront-set structure [2502.07262; 2604.01585].

## 1. Local definition and structural features

Let \(F\) be a local field containing all cube roots of unity \(\mu_3\), and fix an embedding
\[
\epsilon:\mu_3\hookrightarrow \mathbf C^\times .
\]
For \(G=\GL_2(F)\) and \(c\in \mathbf Z/3\mathbf Z\), the cubic Kazhdan–Patterson cover is the central extension
\[
1\to \mu_3 \xrightarrow{i} \widetilde G^{(c)} \xrightarrow{p} G \to 1,
\]
realized as \(\widetilde G=G\times \mu_3\) with multiplication
\[
(g_1,z_1)\cdot (g_2,z_2)=\bigl(g_1g_2,\; z_1z_2\,\sigma_K^{(c)}(g_1,g_2)\bigr),
\]
where
\[
\sigma_K^{(c)}(g_1,g_2)=\left(\frac{K(g_1g_2)}{K(g_1)},\frac{K(g_1g_2)}{K(g_2)\det(g_1)}\right)_3 \cdot (\det g_1,\det g_2)_3^c
\]
and
\[
K\!\left(\begin{pmatrix}a&b\\ c&d\end{pmatrix}\right)=
\begin{cases}
c,& c\neq 0,\\
d,& c=0.
\end{cases}
\]
The preferred section is \(s(g)=(g,1)\). In this rank-\(2\) treatment the cover is formulated concretely in Kubota/Kazhdan–Patterson terms rather than primarily in Brylinski–Deligne language, and the dependence on \(c\) is explicit [2507.15279].

The subgroup structure is controlled by
\[
d=\frac{3}{\gcd(1+4c,3)}, \qquad Z^{(c)}=F^{\times,d},
\]
so that the center of \(\widetilde G\) is the preimage \(\widetilde{Z^{(c)}}\), while the center of \(\widetilde T\) is \(\widetilde{Z^{(c)}}\,\widetilde{T_3}\), where
\[
T_3=\{\diag(a,b)\in T:\ a,b\in F^{\times,3}\}.
\]
A maximal abelian subgroup \(\widetilde T_*\subset \widetilde T\) is chosen, and its intersection with \(\widetilde Z\) is denoted \(\widetilde Z_*\).

Several splitting phenomena are basic. The cover splits canonically over the upper unipotent subgroup \(N\) via the preferred section \(s\). If \(F\) is nonarchimedean and \(|3|=1\), then it also admits the Kubota splitting over \(K=\GL_2(\mathcal O_F)\),
\[
\kappa(g)=(g,k(g)),
\qquad
k(g)=
\begin{cases}
(c,d/\det g)_3,& 0<|c|<1,\\
1,& |c|=0,1,
\end{cases}
\]
for \(g=\begin{pmatrix}a&b\\ c&d\end{pmatrix}\). If \(F\) is archimedean and contains \(\mu_3\), then \(F\cong \mathbf C\) and the cover splits as
\[
\widetilde G\cong \mu_3\times G,
\]
so the local metaplectic structure becomes a direct product [2507.15279].

## 2. Adelic form and global splittings

For a number field \(F\) containing \(\mu_3\), with adeles \(\mathbf A\), the adelic cubic cover is defined from the local covers by the restricted product modulo the standard metaplectic compatibility relation
\[
\widetilde G(\mathbf A)=M\backslash \prod_v' \widetilde G(F_v),
\]
where
\[
M=\{(\xi,\xi^{-1})\in F_v^\times\times F_{v'}^\times:\ \xi\in \mu_3,\ v\neq v'\}.
\]
This yields
\[
1\to \mu_3 \to \widetilde G(\mathbf A)\xrightarrow{p} G(\mathbf A)\to 1.
\]

The local preferred sections induce a global splitting
\[
s:G(F)\hookrightarrow \widetilde G(\mathbf A),
\]
which is essential for the automorphic theory: automorphic forms on \(\widetilde G(\mathbf A)\) may be restricted along \(G(F)\), and period integrals can therefore be taken over quotients of the linear group. The adelic analogues \(T_3(\mathbf A)\), \(Z^{(c)}(\mathbf A)\), \(T_*(\mathbf A)\), and \(Z_*(\mathbf A)\) are arranged so that
\[
\widetilde T_*(\mathbf A)=T(F)\,\widetilde{T_3Z^{(c)}(\mathbf A)}
\]
is a maximal abelian subgroup of \(\widetilde T(\mathbf A)\), and
\[
\widetilde Z_*(\mathbf A)=Z(F)\,\widetilde{Z^{(c)}(\mathbf A)}
\]
is a maximal abelian subgroup of \(\widetilde Z(\mathbf A)\) [2507.15279].

## 3. Genuine and exceptional representations

A representation of a metaplectic subgroup \(\widetilde H\) is \(\epsilon\)-genuine if \(\mu_3\) acts through the fixed embedding \(\epsilon\). For the cubic cover of \(\GL_2\), the central genuine objects are the exceptional representations. If \(\chi\) is a genuine character of \(\widetilde{Z^{(c)}}\,\widetilde{T_3}\), extended to \(\widetilde T_*\), the local principal series is
\[
V(\chi)=\Ind_{\widetilde B_*}^{\widetilde G}\chi.
\]
It satisfies \(V(\chi)^\vee\simeq V(\chi^{-1})\), and it is irreducible unless \(\chi\) is exceptional. The exceptional condition is
\[
\chi\!\left(s(\diag(x^3,x^{-3}))\right)=|x|
\qquad \forall x\in F^\times .
\]
For such \(\chi\), there is an exact sequence
\[
0\to V_0(\chi^w)\to V(\chi)\to V_0(\chi)\to 0,
\]
and \(V_0(\chi)\) is the local exceptional representation [2507.15279].

Two distinguished exceptional characters \(\chi_\pm\) satisfy
\[
\chi_\pm|_{\mu_3}=\epsilon^\pm,
\qquad
\chi_\pm\!\left(s(\diag(x^3,x^{-3}))\right)=|x|,
\]
and their local exceptional representations are denoted
\[
\sigma_{\pm,v}=V_0(\chi_{\pm,v}).
\]
Their Jacquet theory is unusually rigid: for exceptional \(\chi\), the twisted Jacquet module \(V_0(\chi)_\psi\) is an irreducible genuine \(\widetilde Z\)-representation. This irreducibility underlies invariant pairings, Whittaker-model constructions, and local multiplicity-one results. A notable subtlety is that, depending on the cocycle parameter \(c\), the exceptional theta series can have multiple Whittaker functionals; the metaplectic theory therefore uses the full \(\widetilde Z\)-module structure of twisted Jacquet modules rather than only scalar Whittaker uniqueness [2507.15279].

Globally, if \(\chi\) is an exceptional genuine character on \(\widetilde{T_3Z^{(c)}(\mathbf A)}\), trivial on \(T(F)\cap \widetilde{T_3Z^{(c)}(\mathbf A)}\), one forms the adelic principal series and the Eisenstein series
\[
E(g,\chi\delta^s,f_s)=\sum_{\gamma\in B(F)\backslash G(F)} f_s(\gamma g).
\]
At \(s=0\), an exceptional \(\chi\) produces a simple pole, and the residue
\[
\theta(g,\chi,f)=\Res_{s=0}E(g,\chi\delta^s,f_s)
\]
spans an irreducible automorphic representation \(\Theta(\chi)\). The two global exceptional theta representations are
\[
\sigma_\pm=\Theta(\chi_\pm)=\otimes_{[\mu_3]}' \sigma_{\pm,v}.
\]
These are the metaplectic theta inputs for the co-period theory [2507.15279].

## 4. Higher-rank cubic covers of \(\GL_r\)

For nonarchimedean \(F\) with residue characteristic \(p\), the higher-rank theory is developed under the tameness condition \(\gcd(n,p)=1\), equivalently \(n\mid q-1\). An \(n\)-fold cover of \(G=\GL_r(F)\) is a central extension
\[
1\to \mu_n\to \widetilde G\xrightarrow{p} G\to 1,
\]
with cocycle \(\sigma\) defined by a section \(s\) through
\[
s(g_1)s(g_2)\sigma(g_1,g_2)=s(g_1g_2).
\]
If \(H\subset G\) is pro-\(p\), then \(H^2(H,\mu_n)=0\), so \(H\) splits uniquely in the cover; in particular, unipotent radicals split canonically [2502.07262].

For \(\GL_r(F)\), one basic cocycle is the determinant cover
\[
\sigma_{\det}(g_1,g_2)=(\det(g_1),\det(g_2))_n,
\]
and another is the standard Kazhdan–Patterson cocycle \(\sigma_{\mathrm{KP}}\), characterized by the block-compatibility formula
\[
\sigma_{\mathrm{KP}}(\diag(g_1,\dots,g_k), \diag(g_1',\dots,g_k'))
=
\prod_{i=1}^k \sigma_{\mathrm{KP}}(g_i,g_i')
\prod_{1\le i<j\le k} (\det(g_i),\det(g_j'))_n .
\]
All Brylinski–Deligne \(n\)-fold covers of \(\GL_r(F)\) arise as Baer sums of copies of the determinant cover and the standard Kazhdan–Patterson cover, represented by cocycles
\[
\sigma_{\det}^c\cdot \sigma_{\mathrm{KP}}^d,
\]
and the Kazhdan–Patterson covers are precisely those with \(d=1\) [2502.07262].

In the cubic case \(n=3\), the relevant Hilbert symbol is \((\cdot,\cdot)_3\), and the structural invariants simplify sharply. A complementary description treats the Kazhdan–Patterson cover as a Brylinski–Deligne cover with
\[
Q(\alpha^\vee)=-1,
\]
equivalently \(2\mathbf a-\mathbf b=-1\), and defines
\[
d_r=\gcd(3,\,2r\mathbf a-r+1).
\]
For cubic covers, \(d_r\in\{1,3\}\). The subgroup
\[
G_r^{(3)}=\{g\in \GL_r(F):\det(g)\in F^{\times 3}\}
\]
plays a central role in restriction and induction theory, and Levi blocks are generally not block-compatible, so one uses the metaplectic tensor product rather than the naive tensor product [2604.01585].

## 5. Hecke modules, Gelfand–Graev representations, and Whittaker dimension

For tame Kazhdan–Patterson covers, the Gelfand–Graev representation
\[
V=\ind_U^{\widetilde G}(\psi)
\]
admits a simple-type description in the Hecke category. If \((J,\lambda)\) is a simple type, then for Kazhdan–Patterson covers the Hecke algebra satisfies
\[
\mathcal H(\widetilde G,\lambda)\cong \mathcal A\rtimes \mathcal H_0,
\]
with \(\mathcal A=\mathbf C[T(\mathfrak b,\lambda)]\) and \(\mathcal H_0\) a finite Hecke algebra of type \(A\). The corresponding Hecke module of the Gelfand–Graev representation decomposes as
\[
V^\lambda \cong \bigoplus_{\mathcal O\in X(\lambda)/\mathfrak S_k} V_{\mathcal O},
\]
where each summand is induced from the sign character of the stabilizer Hecke algebra [2502.07262].

This yields an explicit Whittaker-dimension formula for discrete series. If \(\pi\) is a genuine discrete series representation of inertial class corresponding to \((J,\lambda)\), then
\[
\dim_{\mathbf C}\Hom_{\widetilde G}(\ind_U^{\widetilde G}(\psi),\pi)=|X(\lambda)/\mathfrak S_k|.
\]
For Kazhdan–Patterson covers,
\[
|X(\lambda)/\mathfrak S_k|
=
\binom{k+n_0-1}{k}\frac{d_0}{n_0},
\]
while for Savin’s cover the factor \(d_0/n_0\) is absent. In the cubic Kazhdan–Patterson case,
\[
n_0=\frac{3}{\gcd\!\bigl(3,(2c+1)r_0l_0,l_0\bigr)},
\qquad
d_0=\frac{3}{\gcd\!\bigl(3,l_0(2cr+r-1)\bigr)},
\]
so \(n_0,d_0\in\{1,3\}\). This makes the orbit count and the Whittaker dimension especially explicit [2502.07262].

These higher-rank formulas are consistent with a broader phenomenon: Whittaker multiplicity on metaplectic covers need not be \(1\). In the cubic setting this is not an exceptional pathology but part of the basic representation theory.

## 6. Co-periods and the symmetric cube \(L\)-function

The rank-\(2\) cubic cover enters most directly through the co-period integral attached to one linear \(\GL_2\)-representation and two exceptional theta representations on the cover. If \(f\in \sigma\), where \(\sigma\) is an automorphic representation of \(G(\mathbf A)=\GL_2(\mathbf A)\), and \(f_\pm\in \sigma_\pm\), the global co-period is
\[
\CP(f,f_+,f_-)
=
\int_{Z_*(\mathbf A)G(F)\backslash G(\mathbf A)}
f(g)\,f_+(g)\,f_-(g)\,d^\tau g.
\]
The two theta factors are chosen with opposite genuineness,
\[
\chi_+|_{\mu_3}=\epsilon,
\qquad
\chi_-|_{\mu_3}=\epsilon^{-1},
\]
so the product \(f_+(g)f_-(g)\) descends to a function on the linear group. This is the basic reason the cubic cover appears in the symmetric-cube story [2507.15279].

The local branching problem is the trilinear space
\[
\Hom_{G(F_v)}\!\bigl(\sigma_v\otimes \sigma_{+,v}\otimes \sigma_{-,v},\mathbf C\bigr).
\]
If \(v\nmid 3\) or \(\sigma_v\) is non-supercuspidal, then
\[
\dim \Hom_{G(F_v)}\!\bigl(\sigma_v\otimes \sigma_{+,v}\otimes \sigma_{-,v},\mathbf C\bigr)=1.
\]
For nonarchimedean \(v\), the canonical local period is defined by the matrix-coefficient integral
\[
I(\cdots)
=
\int_{Z^{(c)}(F_v)\backslash G(F_v)}
(\sigma_v(g)f_v,f_v^\vee)_v\,
(\sigma_{+,v}(g)f_{+,v},f_{+,v}^\vee)_{+,v}\,
(\sigma_{-,v}(g)f_{-,v},f_{-,v}^\vee)_{-,v}\,dg_v,
\]
which converges absolutely and spans the product of local trilinear Hom-spaces. In the unramified case, with \(\sigma_v=I_{B(F_v)}^{G(F_v)}\theta_v\) and spherical vectors normalized to pair to \(1\), the principal calculation gives
\[
I=
\frac{L(1/2,\sigma_v,\Sym^3)}{L(1,\sigma_v,\Ad)}
\]
in the case emphasized for the global theory [2507.15279].

For Eisenstein data, the global co-period requires regularization via Yamana’s mixed truncation. The regularized period unfolds to a product of local zeta integrals,
\[
P(\varphi_s,\varphi_+,\varphi_-)
=
\prod_v
\int_{Z_*(F_v)N(F_v)\backslash G(F_v)}
f_{s,v}(g_v)\,W_{+,v}(g_v)\,W_{-,v}(g_v)\,dg_v,
\]
and at \(s=0\) this yields an explicit Euler factorization with symmetric-cube shape. In the irreducible Eisenstein range,
\[
\CP(\varphi,\varphi_+,\varphi_-)\,
\CP^\vee(\varphi^\vee,\varphi_+^\vee,\varphi_-^\vee)
=
\frac{L(1/2,\sigma,\Sym^3)}
{L(1,\theta_1\theta_2^{-1})L(1,\theta_2\theta_1^{-1})}
\prod_v I_v^\sharp(\cdots),
\]
where
\[
I^\sharp(\cdots)=\frac{L(1,\sigma_v,\Ad)}{L(1/2,\sigma_v,\Sym^3)}\,I(\cdots)
\]
is normalized to equal \(1\) at almost all unramified places [2507.15279].

For cuspidal \(\sigma\), the central conjecture is an Ichino–Ikeda type statement:
\[
\CP(f,f_+,f_-)\neq 0
\quad\Longleftrightarrow\quad
L(1/2,\sigma,\Sym^3)\neq 0,
\]
and moreover
\[
\CP(f,f_+,f_-)\CP^\vee(f^\vee,f_+^\vee,f_-^\vee)
=
C\,\frac{L(1/2,\sigma,\Sym^3)}{L(1,\sigma,\Ad)}
\prod_v I^\sharp(\cdots),
\]
for a constant \(C\in \mathbf C^\times\) independent of \(\sigma\). The implication
\[
\CP\neq 0 \implies L(1/2,\sigma,\Sym^3)\neq 0
\]
is recorded as already proved by Ginzburg–Jiang–Rallis, and the converse together with the refined product formula is conjectural. The same paper deduces from local multiplicity one that there exist cuspidal automorphic representations with prescribed local supercuspidal components and nonvanishing central symmetric cube \(L\)-values [2507.15279].

## 7. Derivatives, wavefront sets, and analytic context

For higher-rank cubic Kazhdan–Patterson covers, Bernstein–Zelevinsky theory is now available in a form parallel to the linear case. Genuine irreducible representations are classified by multisegments together with a compatible genuine central character. For a cuspidal line \(\rho\), the reducibility invariant \(l(\rho)\) divides \(3\), so
\[
l(\rho)\in\{1,3\},
\qquad
n(\rho)=\frac{3}{l(\rho)}\in\{3,1\}.
\]
The wavefront set of an irreducible genuine representation \(Z(\mathfrak m)\) is a singleton
\[
\WF(Z(\mathfrak m))=\{\lambda_{\mathfrak m}\},
\]
where \(\lambda_{\mathfrak m}\) is determined by the iterated degrees of the highest derivatives. For a single segment \(\Delta=[0,l-1]_\rho\), the cubic rule becomes especially simple: if \(n(\rho)=1\), then
\[
\lambda_\Delta=(r_0^l),
\]
while if \(n(\rho)=3\) and \(l=3c+d\) with \(0\le d<3\), then
\[
\lambda_\Delta=((3r_0)^c,\ r_0d).
\]
Genericity is governed by segment length: if every segment length \(l_i\le n(\rho_i)\), then \(Z(\mathfrak m)\) is generic, whereas if some \(l_i>n(\rho_i)\), it is not. In particular, for cubic covers, if \(3\nmid r\), then every irreducible genuine representation is generic [2604.01585].

The arithmetic and analytic side of cubic metaplectic theory is represented by work over \(\mathbf Q(\omega)\) on Patterson’s conjecture. That literature does not formulate itself in adelic Kazhdan–Patterson language, but it uses Kubota’s cubic multiplier, Patterson’s cubic theta function, and Fourier coefficients identified with cubic Gauss sums. Conditional on the Generalized Riemann Hypothesis for Hecke \(L\)-functions over \(\mathbf Q(\omega)\), one obtains the asymptotic
\[
\sum_{p\equiv 1 \!\!\!\pmod 3}
\frac{S_p}{2\sqrt p}\,
W\!\left(\frac pX\right)
\sim
\frac{(2\pi)^{2/3}}{3\Gamma(\tfrac23)}
\left(\int_0^\infty W(x)x^{-1/6}\,dx\right)\frac{X^{5/6}}{\log X},
\]
together with an explicit level-aspect Voronoi formula for cubic Gauss sums and a proof that Heath-Brown’s cubic large sieve is sharp up to factors of \(X^{o(1)}\) under GRH. That paper explicitly states that “the Fourier coefficients of the cubic theta function essentially sample cubic Gauss sums,” and it is therefore directly relevant to the analytic behavior of Fourier coefficients in the cubic metaplectic setting, even though it does not present the global cubic Kazhdan–Patterson cover in modern adelic terms [2109.07463].

As a broader comparison, the Kazhdan–Patterson paradigm extends beyond \(\GL_r\). For double covers of odd general spin groups, one again has genuine torus characters, exceptional principal series, unique irreducible exceptional quotients, global Eisenstein residues, vanishing of large classes of twisted Jacquet modules and Fourier coefficients, and co-period applications. This comparison does not supply cubic formulas, and several ingredients there are explicitly degree-\(2\)-specific, including the use of quadratic Weil factors; nevertheless, it shows that exceptional-representation and co-period constructions belong to a wider metaplectic architecture rather than to \(\GL_r\) alone [1411.4082].

Source: https://www.emergentmind.com/topics/cubic-kazhdan-patterson-cover