---
title: Jacquet-Langlands Dual Overview
url: https://www.emergentmind.com/topics/jacquet-langlands-dual
type: topic
---

# Jacquet-Langlands Dual Overview

Searching arXiv for the cited works to ground the article in current records.
arxiv_search(query="Jacquet-Langlands correspondence dual local inner form GL(n) division algebra parity Mieda", max_results=5)
arxiv_search(query="1603.04691 Parity of the Langlands parameters of conjugate self-dual representations of GL(n) and the local Jacquet-Langlands correspondence", max_results=5)
arxiv_search(query="1112.6116 Geometric approach to the local Jacquet-Langlands correspondence", max_results=5)
arxiv_search(query="2311.10225 ell-adic topological Jacquet-Langlands duality", max_results=5)
arxiv_search(query="1904.09517 Trace Formula and the Proof of the Global Jacquet-Langlands Correspondence", max_results=5)
Jacquet–Langlands dual denotes, in its classical representation-theoretic sense, the object on an inner form that corresponds to a representation of a general linear group under a Jacquet–Langlands correspondence. In the local non-archimedean case, for \(G=\mathrm{GL}_n(F)\) and the unique central division algebra \(D/F\) of invariant \(1/n\), the Jacquet–Langlands dual of a discrete series representation of \(G\) is its corresponding irreducible smooth representation of \(D^\times\). In the global case, it is the automorphic representation on an inner form \(G'=\mathrm{GL}_1(D)\) matching the same placewise \(L\)-parameters. The term has since acquired broader uses in cohomology, arithmetic geometry, chromatic homotopy theory, and geometric Langlands, but its classical core remains the transfer between \(\mathrm{GL}_n\) and its inner forms, characterized by matching characters on corresponding regular elliptic classes and by common Langlands parameters [1603.04691] [1112.6116] [1904.09517].

## 1. Terminology, scope, and basic meaning

The phrase has a stable classical meaning and several derived meanings. Classically, “dual” does not refer to the contragredient representation. In the global inner-form setting, \(\mathrm{GL}_n/F\) and \(\mathrm{GL}_1(D)\) have the same \(L\)-group, namely \(GL_n(\mathbb{C}) \times W_F\) with trivial \(W_F\)-action on \(GL_n(\mathbb{C})\); in that sense, the duality is governed by shared Langlands parameters rather than by linear duality of vector spaces [1904.09517].

| Setting | Objects | Meaning of “Jacquet–Langlands dual” |
|---|---|---|
| Local non-archimedean | \(\mathrm{GL}_n(F)\) and \(D^\times\) | The corresponding representation on the inner form \(D^\times\) |
| Global automorphic | \(\mathrm{GL}_n(\mathbb{A}_F)\) and \(\mathrm{GL}_1(D)(\mathbb{A}_F)\) | The automorphic transfer to the inner form with matching local parameters |
| Function-field geometry | \(J_0(\mathfrak n)^{new}\) and \(J^{\mathfrak n}\) | The quaternionic Jacobian or the JL isogeny partner |
| Shimura-variety cohomology | \(X_{B_1}\), \(X_{B_2}\) | A Hodge-class correspondence inducing the JL isomorphism on \(\pi\)-isotypic cohomology |
| Topological/chromatic | Morava \(E\)-theory, Lubin–Tate/Drinfeld towers | A spectrum-level analogue built from homotopy fixed points or vanishing cycles |
| Geometric Langlands | Whittaker/Jacquet functors, BD Grassmannian | A categorical Jacquet-type duality rather than the classical inner-form correspondence |

A common source of confusion is that some recent papers use the expression more analogically. In the geometric Langlands setting, “Jacquet–Langlands duality” can mean a categorical duality between Poincaré series, Whittaker coefficients, and Jacquet functors, explicitly distinguished from the classical correspondence between inner forms [2211.05282]. Likewise, on the Beilinson–Drinfeld Grassmannian, the phrase can describe the identification of Jacquet-type constructions with factorization modules attached to the Langlands dual group \(\check G\), again in a categorical rather than inner-form sense [2310.19734]. This suggests that the phrase now names a family of structurally similar correspondences whose prototype remains the inner-form transfer.

## 2. Local non-archimedean theory

Let \(F\) be a non-archimedean local field and let \(D\) be the unique central division algebra over \(F\) with invariant \(1/n\). The local Jacquet–Langlands correspondence gives a bijection
\[
JL:\operatorname{Irrdisc}(\mathrm{GL}_n(F)) \xrightarrow{\sim} \operatorname{Irr}(D^\times),
\]
with the Jacquet–Langlands dual of a discrete series representation \(\pi\) defined to be \(JL(\pi)\). On regular elements, the duality is characterized by character identities on matching conjugacy classes: if \(h \in (D^\times)^{reg}\) and \(g_h \in \mathrm{GL}_n(F)^{ell}\) has the same minimal polynomial, then
\[
\theta_\rho(h)=(-1)^{n-1}\theta_{JL(\rho)}(g_h),
\]
and dually
\[
\theta_\pi(g_h)=(-1)^{n-1}\theta_{LJ(\pi)}(h).
\]
Mieda’s geometric construction realizes these identities through the \(\ell\)-adic cohomology of the Drinfeld tower, producing inverse homomorphisms \(JL\) and \(LJ\) on Grothendieck groups, compatible with twists and preserving central characters; when \(n\) is prime, this construction preserves irreducibility and yields a purely local proof of the bijection \(\operatorname{Irr}(D^\times)\to \operatorname{Disc}(\mathrm{GL}_n(F))\) [1112.6116].

The geometric mechanism is rigid-analytic and cohomological. The Drinfeld tower carries commuting actions of \(G=\mathrm{GL}_n(F)\), \(G'=D^\times\), and \(W_F\), and its alternating compactly supported cohomology defines virtual modules whose characters encode the transfer. The key input is a Lefschetz trace formula relating orbital integrals to fixed-point counts, together with transfer of test functions and orthogonality of discrete-series characters. In Mieda’s formulation, the tower produces
\[
JL(\rho)=\frac{(-1)^{n-1}}{n}H_{Dr}[\rho], \qquad
LJ(\pi)=\frac{(-1)^{n-1}}{n}H_{Dr}[\pi],
\]
and these are characterized by the regular elliptic character relations above [1112.6116].

The local theory also admits integral and mod-\(\ell\) refinements. Dat’s non-abelian \(\ell\)-integral Lubin–Tate theory shows that the supercuspidal part of the middle-degree compactly supported cohomology realizes the tensor product of three companions of a supercuspidal representation: the \(\mathrm{GL}_d(K)\)-representation \(\pi\), its Jacquet–Langlands mate \(\rho(\pi)\) on \(D^\times\), and its Langlands–Vignéras parameter \(\sigma(\pi)\) on \(W_K\). In the Grothendieck group, for finite-length mod-\(\ell\) representations one has
\[
[R_\pi]=LJ_{F_\ell}(\pi)\otimes \sigma_{F_\ell}(\pi)^{ss},
\]
while for supercuspidals the middle degree gives an actual tensor product rather than merely a virtual identity. The same framework produces exact functorial correspondences in flat families and identifies universal deformation rings simultaneously on the \(\mathrm{GL}_d\)-side, the \(D^\times\)-side, and the Weil-side [1011.1887].

## 3. Global correspondence and the inner-form viewpoint

For a global field \(F\) of characteristic \(0\), let \(D\) be a central simple \(F\)-algebra of dimension \(n^2\), \(G=\mathrm{GL}_n/F\), and \(G'=\mathrm{GL}_1(D)\). The global Jacquet–Langlands correspondence identifies discrete automorphic representations of \(G'(\mathbb{A}_F)\) with the \(D\)-compatible discrete automorphic representations of \(\mathrm{GL}_n(\mathbb{A}_F)\), namely those whose local components at the non-split places are essentially square-integrable and match under local JL. The correspondence preserves central characters, matches local components place by place, and preserves standard local and global \(L\)- and \(\varepsilon\)-factors because the two sides have the same local and global \(L\)-parameters [1904.09517].

At the local non-split places \(v\), the transfer is characterized by matching of regular semisimple orbital integrals and character identities for corresponding test functions; at split places, it is tautological under \(D_v \simeq M_n(F_v)\). Globally, the proof proceeds through Arthur’s stabilized trace formula and transfer of test functions. The spectral identity compares
\[
\sum_{\Pi} m(\Pi)\operatorname{tr}\Pi(f)
\quad\text{and}\quad
\sum_{\Pi'} m(\Pi')\operatorname{tr}\Pi'(f'),
\]
for matching functions \(f\) and \(f'\), while the geometric side is organized by matched regular semisimple conjugacy classes with transfer factor \(1\) for these inner forms [1904.09517].

For \(PGL_2\) and quaternionic inner forms, the classical correspondence can also be recovered by twisted automorphic descent. In that construction, one begins with a residual Eisenstein representation on \(SO_{4n+2}\), applies Bessel-Fourier coefficients along carefully chosen unipotent radicals, and at \(n=1\) obtains an explicit descent to \(PD^\times\) whose Jacquet–Langlands transfer is the given automorphic representation of \(PGL_2\). The ramified places are selected by the Tunnell–Saito epsilon criterion, and the global descent integral unfolds to an Euler product involving \(L(s,T,\Lambda^2)\) and \(L(s,T\otimes \omega)\), thereby recovering the classical correspondence by a method distinct from the trace formula and from Shimizu’s theta-series construction [1501.00506].

The global inner-form viewpoint has recently been recast in operator-algebraic terms for \(\mathrm{GL}(2)\). For a totally real field \(F\), a quaternion algebra \(D/F\) ramified at all infinite places, and a global pair \(\pi \leftrightarrow \pi'\) under JL, the \(S\)-local constituents satisfy a von Neumann dimension identity
\[
\frac{\dim_{\mathcal L(\mathrm{SL}(2,\mathcal O_S))}(\pi_S)}{\dim_{\mathbb C}(\pi'_S)}
=
\left|\frac{\zeta_D(0)}{\zeta_F(0)}\right|.
\]
Here the spectral mass of the two sides is measured through Murray–von Neumann dimension, and the proportionality constant is expressed by special zeta-values. This does not alter the classical correspondence, but it reframes it as a quantitative identity between \(S\)-arithmetic group-algebra modules [2402.18488].

## 4. Parity, conjugate self-duality, and the JL-dual as a detector

In the local quadratic situation \(F/F^+\), Mieda uses the Jacquet–Langlands dual to extract parity information for Langlands parameters of conjugate self-dual supercuspidal representations of \(\mathrm{GL}_n(F)\). Let \(\pi\) be an irreducible conjugate self-dual supercuspidal representation of \(\mathrm{GL}_n(F)\), and let \(\operatorname{rec}_F(\pi)\) be its \(n\)-dimensional Weil parameter. On both the representation side and the parameter side one defines a parity sign \(C\in\{\pm1\}\) from a \(\tau\)-invariant bilinear pairing, with \(+1\) corresponding to conjugate orthogonal and \(-1\) to conjugate symplectic. Assuming \(F/F^+\) is at worst tamely ramified and \(D/F\) has invariant \(1/n\), Mieda constructs an automorphism \(\tau\) of \(D^\times\) and an element \(t \in D^\times\) with \(\tau^2=\operatorname{Int}(t)\), proves that \(JL(\pi)\) is conjugate self-dual, and establishes the parity formula
\[
C_{\operatorname{rec}_F(\pi)} = (-1)^{n-1} C_{JL(\pi)}.
\]
This extends the self-dual parity theorem of Prasad–Ramakrishnan to the conjugate self-dual supercuspidal case [1603.04691].

The proof is cohomological. It uses the Lubin–Tate tower and the identification
\[
H^{n-1}_{LT/\Gamma,c,\pi}\bigl((n-1)/2\bigr)
\simeq
\pi \boxtimes JL(\pi)^\vee \boxtimes \operatorname{rec}_F(\pi)^\vee,
\]
together with a cup-product pairing and a twisting operator \(\theta\) that converts the pairing into a Hermitian form adapted to conjugate self-duality. The relation \(\theta^2=t\circ c^2\) on cohomology forces the sign discrepancy \((-1)^{n-1}\) between the \(D^\times\)-side and the Weil-side. In this sense, the JL-dual becomes a parity detector: once \(C_{JL(\pi)}\) is computed on the inner form, the sign of the Langlands parameter follows immediately [1603.04691].

This has arithmetic consequences. In characteristic \(0\), when \(F\neq F^+\), the parity \(C_{\operatorname{rec}_F(\pi)}\) determines whether \(\pi\) lies in the image of the standard or twisted base change from the quasi-split unitary group \(U_{F/F^+}(n)\). Mieda also computes the signs explicitly for simple supercuspidals. For example, in the unramified quadratic case, if \(\pi_{\zeta,\chi,c}\) is conjugate self-dual, then
\[
C_{\operatorname{rec}_F(\pi_{\zeta,\chi,c})}=\chi(\varepsilon)c,
\]
while in the ramified quadratic case with \(p\neq 2\), the sign is \(1\) when \(\chi\) is trivial and \(-1\) otherwise; in the self-dual case \(F=F^+\), simple supercuspidals of \(\mathrm{GL}_2(F)\) with trivial central character have symplectic parameter [1603.04691].

## 5. Geometric and arithmetic realizations

In the function-field setting, the Jacquet–Langlands dual can appear as a Jacobian or as the target of an explicit isogeny. For \(K=\mathbb F_q(T)\), square-free level \(\mathfrak n\), and the quaternion algebra \(D/K\) ramified exactly at the primes dividing \(\mathfrak n\) and split at \(\infty\), the global JL correspondence yields a Hecke-equivariant isogeny
\[
J_0(\mathfrak n)^{new} \longrightarrow J^{\mathfrak n},
\]
where \(J^{\mathfrak n}\) is the Jacobian of the quaternionic modular curve attached to \(D\). In the special case \(\mathfrak n=xy\) with \(\deg x=1\) and \(\deg y=2\), the paper computes
\[
J_0(xy)(K)_{tor}=C(xy)\cong \mathbb Z/(q+1)\mathbb Z \oplus \mathbb Z/(q^2+1)\mathbb Z,
\]
shows \(\mathbb T(xy)/\mathfrak E(xy)\cong \mathbb Z/(q^2+1)(q+1)\mathbb Z\), and, under an integral Hecke-module identification together with Gorenstein hypotheses at primes dividing \(q^2+1\), constructs a JL isogeny with kernel
\[
\ker(\phi)=(C_y)\cong \mathbb Z/(q^2+1)\mathbb Z.
\]
Here the JL dual is not a representation but the quaternionic Jacobian \(J^{xy}\), and the transfer is encoded by an explicit cuspidal, Eisenstein-annihilated kernel [1306.3632].

In the setting of quaternionic Shimura varieties, the correspondence can be realized by a Hodge class. Let \(B_1\) and \(B_2\) be quaternion algebras over a totally real field \(F\), split at the same archimedean places \(\Sigma\), and let \(X_{B_i}\) be the associated Shimura varieties of dimension \(d=|\Sigma|\). For a cohomological automorphic representation \(\pi\) of \(\mathrm{GL}_2(\mathbb A_F)\) with JL transfers \(\pi_i\) to \(B_i^\times(\mathbb A_F)\), the paper constructs
\[
\xi \in H^{2d}(X_1\times X_2,\mathbb Q)_{\pi_1\boxtimes \pi_2}
\]
of Hodge type \((d,d)\), and the associated correspondence
\[
\xi(d)^*: H^d(X_1,\mathbb Q)_{\pi_1}\longrightarrow H^d(X_2,\mathbb Q)_{\pi_2}
\]
is an isomorphism of \(\mathbb Q\)-Hodge structures. Conditional on Kottwitz’s conjecture for unitary similitude Shimura varieties, the \(\ell\)-adic realization of \(\xi\) is Galois invariant for every \(\ell\), so the same class induces an isomorphism of \(\operatorname{Gal}(\overline{\mathbb Q}/F_\Sigma)\)-modules on the \(\pi_i\)-isotypic middle-degree étale cohomology. In this formulation, the Jacquet–Langlands dual is a cohomological object produced by a pull-push operator attached to a rational Hodge class rather than by a direct statement about representations alone [1806.10563].

These realizations share a common pattern: the classical transfer between inner forms is encoded by geometry, either through Jacobians and isogenies or through cycles and cohomological correspondences. A plausible implication is that the phrase “JL-dual” remains stable even when the ambient category changes, provided the resulting object still implements the same Hecke- and parameter-theoretic matching.

## 6. Topological, categorical, and factorization analogues

Recent work extends the terminology into chromatic and categorical contexts. In \(\ell\)-adic topological Jacquet–Langlands duality, the Lubin–Tate tower is enlarged to a “degenerating Lubin–Tate tower,” endowed with presheaves of \(E_\infty\)-ring spectra. Applying algebraic \(K\)-theory and spectral vanishing cycles yields a cohomology spectrum
\[
JL_\ell(\mathbb G):=
\operatorname*{hocolim}_{m\to\infty}\operatorname*{holim}_{j\to\infty}
\operatorname{St}_x\Phi\,\operatorname{Fib}\bigl(S/\ell^j \wedge (K\circ O^{topsym}_{lvl\,p^m})\bigr),
\]
with a natural \(GL_n(\mathbb Z_p)\times \operatorname{Aut}(\mathbb G)\)-action. For each supercuspidal irreducible \(\overline{\mathbb Q}_\ell\)-representation \(\pi\) of \(GL_n(\mathbb Q_p)\), the rational homotopy group \(\overline{\mathbb Q}_\ell\otimes \pi_{1-n}(JL_\ell(\mathbb G))\) contains \((\pi|_{GL_n(\mathbb Z_p)}\otimes JL(\pi))\) as a summand. The associated topological Jacquet–Langlands dual of a spectrum \(X\),
\[
TJL_\ell(X;\mathbb G),
\]
converts \(\operatorname{Aut}(\mathbb G)\)-types in \((K(E(\mathbb G))^\wedge_\ell)_*(X)\) into \(GL_n(\mathbb Z_p)\)-types in its homotopy. At height \(1\), the construction preserves \(L\)-factors: the automorphic \(L\)-factor of the \(GL_1\)-representation attached to \(TJL_\ell(X;\mathbb G)\) equals the \(p\)-local Euler factor \(L_p(s;X)\) [2311.10225].

A related integral construction refines the Lubin–Tate tower itself in spectral algebraic geometry. Finite levels are represented by \(E_\infty\)-rings \(\mathrm{jl}_r\), their inverse limit
\[
\mathrm{jl}=\varprojlim_r \mathrm{jl}_r
\]
carries commuting \(GL_n(\mathbb Z_p)\) and Morava stabilizer \(S_n\) actions, and the Jacquet–Langlands dual of Morava \(E\)-theory is defined by
\[
E_n^{JL}:=\mathrm{jl}^{hS_n}.
\]
This is an \(E_\infty\)-ring, and there is a dual homotopy fixed point spectral sequence
\[
E_2^{s,t}=H_c^s\bigl(GL_n(\mathbb Z_p);\pi_t(E_n^{JL})\bigr)\Rightarrow \pi_{t-s}(\mathbb S_{K(n)}),
\]
positioned as an automorphic partner to the Devinatz–Hopkins spectral sequence for \(E_n^{hS_n}\). Here “JL dual” means descent from the Lubin–Tate side to the Drinfeld side via infinite-level spectral moduli and homotopy fixed points [2509.00690].

On the categorical side, the Beilinson–Drinfeld Grassmannian supports a factorization version of derived geometric Satake in which renormalized spherical Hecke categories are identified with spectral factorization categories attached to the Langlands dual group. In the spherical case,
\[
\operatorname{Sph}_G \xrightarrow{\sim} \operatorname{Sph}^{spec}_{\check G},
\]
while in the parabolic setting one has
\[
\operatorname{Sph}_{G,P} \xrightarrow{\sim} \operatorname{Sph}^{spec}_{\check G,\check P}.
\]
The relevant Jacquet-type categories are described by integrable factorization modules for \(\Upsilon(\check{\mathfrak n}_{\check P})\) and by monads such as \(\operatorname{Sym}((\mathfrak g/\mathfrak n_P)[-2])\). This is not the classical inner-form correspondence, but it is presented as a precise “Jacquet–Langlands dual” paradigm in which automorphic Jacquet constructions correspond to spectral chiral induction for \(\check G\) [2310.19734].

The same caveat is explicit in global geometric Langlands for Poincaré series and miraculous duality. There the central statement is a functorial identity
\[
\operatorname{PsId}_! \circ (\operatorname{coeff}_*)^\vee
\simeq
\operatorname{Poinc}_! \circ \operatorname{PsId}_{Whit}^{-1}[2\delta_G],
\]
together with parabolic Jacquet identities controlling constant terms of Poincaré series and Whittaker coefficients of Eisenstein series. The paper emphasizes that this is a geometric, categorical duality between automorphic functors and their Verdier or Serre duals, not the classical JL correspondence between automorphic representations on inner forms [2211.05282].

Across these extensions, the classical notion persists as the reference case. What changes is the ambient category: smooth representations, cohomology groups, Jacobians, \(E_\infty\)-ring spectra, or factorization categories. The continuing role of the phrase “Jacquet–Langlands dual” indicates that the transfer between \(\mathrm{GL}_n\)-type and inner-form or dual-side data has become a template for a broader family of correspondences.

Source: https://www.emergentmind.com/topics/jacquet-langlands-dual