The Hiraga-Ichino-Ikeda conjecture on formal degrees for classical groups

Raphaël Beuzart-Plessis · 2025

Read The Hiraga-Ichino-Ikeda conjecture on formal degrees for classical groups by Raphaël Beuzart-Plessis (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-08470.txt.

Introduction

Raphaël Beuzart-Plessis wrote a 2025 mathematics paper proving a conjecture about formal degrees, a measure of how large certain representations of classical groups are. The conjecture, by Hiraga, Ichino, and Ikeda, ties that measure to special values called adjoint gamma factors. It is a keystone prediction of the Langlands programme over local fields.

The proof covers symplectic and even orthogonal groups over non-Archimedean fields of characteristic zero. It leans on the twisted endoscopic description of the local correspondence. One operator's residue is computed two ways: once spectrally, once through orbital integrals. Matching the two computations against a known formula yields the theorem, and the method adapts to other classical groups.

Catalog record

CREATORRaphaël Beuzart-Plessis
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE129 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-08470

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

The Hiraga-Ichino-Ikeda conjecture on formal degrees for classical groups Raphaël Beuzart-Plessis Abstract We prove a conjecture of Hiraga-Ichino-Ikeda relating formal degrees of square-integrable representations to adjoint gamma factors for symplectic and even orthogonal groups over characteristic zero non-Archimedean local fields. The proof is based on the twisted endoscopic characterization of the local Langlands correspondence for such groups and extends an approach already appearing in the original paper of Hiraga-Ichino-Ikeda [ HII08 ] itself inspired from earlier work of Shahidi [ Sha92 ] . Namely, the argument consists in computing in two different ways a residue of certain standard intertwining operators suitably extended to a large representation realized on an explicit space of functions. This gives rise to the spectral decomposition of certain singular twisted orbital integrals. The main theorem follows from a comparison of this identity with the Plancherel formula of a symplectic or even orthogonal group. The proof can be adapted to deal with other classical groups, namely odd orthogonal and unitary groups, but the formal degree conjecture was already established in these cases. 1 Introduction 1.1 The formal degree conjecture Let \underline{G} be a connected reductive group defined over a local field F and G=\underline{G}(F) . Recall that an admissible irreducible representation \pi of G is said to be square-integrable if its central character is unitary and its matrix coefficients are square-integrable over G/A_{G} , where A_{G} denotes the maximal split torus in the center of G . To a square-integrable representation \pi , we can attach its formal degree d(\pi) which is the unique positive real number such that \displaystyle\int_{G/A_{G}}\langle\pi(g)v,v^{\vee}\rangle\langle w,\pi^{\vee}(g)w^{\vee}\rangle\;\frac{dg}{da}=\frac{\langle v,w^{\vee}\rangle\langle w,v^{\vee}\rangle}{d(\pi)},\;\mbox{ for every }v,w\in V_{\pi},\;v^{\vee},w^{\vee}\in V_{\pi}^{\vee}, where V_{\pi} denotes the space of (smooth vectors of) \pi and V_{\pi}^{\vee} the space of the contragredient representation. This invariant however depends a priori on the choice of a Haar measure on G/A_{G} . Following [ HII08 ] , we can choose such a Haar measure by fixing a non trivial additive character \psi:F\to S^{1} . More precisely, denoting by \underline{G}_{\bZ} the split form of \underline{G} over \bZ , up to \pm 1 there is a unique invariant volume form \omega_{\bZ} on \underline{G}_{\bZ} that is nowhere vanishing. Pulling back \omega_{\bZ} by any isomorphism \underline{G}_{F^{sep}}=\underline{G}\times_{F}F^{sep}\simeq\underline{G}_{\bZ}\times F^{sep} over a separable closure F^{sep} of F gives rise to a nonzero invariant volume form \omega_{F^{sep}} on \underline{G}_{F^{sep}} , canonical up to \pm 1 . We can then pass from \omega_{F^{sep}} to a Haar measure \lvert\omega_{F^{sep}}\rvert_{\psi} on G by Weil’s construction [ Wei82 ] : if in local coordinates x_{1},\ldots,x_{d} we have \displaystyle\omega_{F^{sep}}=f(x_{1},\ldots,x_{d})dx_{1}\wedge\ldots\wedge dx_{d}, then \displaystyle\lvert\omega_{F^{sep}}\rvert_{\psi}=\lvert f(x_{1},\ldots,x_{d})\rvert d_{\psi}x_{1}\ldots d_{\psi}x_{d} where \lvert.\rvert denotes the normalised absolute value on F , extended to F^{sep} , and d_{\psi}x stands for the \psi -self dual Haar measure on F . Of course, the \pm 1 ambiguity in the definition of \omega_{F^{sep}} disappears when we pass to \lvert\omega_{F^{sep}}\rvert_{\psi} and we shall denote the resulting Haar measure, which depends only on \psi , by d_{\psi}g . We construct similarly a Haar measure d_{\psi}a on A_{G} . We can then take dg=d_{\psi}g , da=d_{\psi}a in the definition of the formal degree, that we will denote by d_{\psi}(\pi) to emphasize the dependence on \psi . 1.2 The conjecture of Hirage-Ichino-Ikeda In [ HII08 ] , Hiraga, Ichino and Ikeda have proposed a conjectural formula for d_{\psi}(\pi) in terms of the local Langlands parameterization for G . To be more specific, let us assume that \underline{G} admits a pure inner form \underline{G}_{qd} that is quasi-split. We recall that the pure inner forms of \underline{G} are naturally parameterized by the Galois cohomology set H^{1}(\Gamma_{F},\underline{G}(F^{sep})) (where F^{sep} denotes a separable closure of F and \Gamma_{F}=\Gal(F^{sep}/F) the absolute Galois group) and that the assumption is satisfied by all classical groups (that is, symplectic, special orthogonal and unitary groups). The local Langlands correspondence predicts the existence of a natural decomposition \displaystyle\Irr(G)=\bigsqcup_{\phi}\Pi^{G}(\phi) of the admissible dual \Irr(G) of G into L -packets \Pi^{G}(\phi) . The latter are finite sets indexed by equivalence classes of L -parameters that is group morphisms \displaystyle\phi:W^{\prime}_{F}\to{}^{L}G, where W^{\prime}_{F} denotes the Weil-Deligne group of F (just its Weil group W_{F} if F is Archimedean and W_{F}\times\SL_{2}(\mathbb{C}) if F is non-Archimedean) and {}^{L}G=\widehat{G}\rtimes\Gamma_{F} the L -group of G , that are continuous, semisimple, algebraic on \SL_{2}(\mathbb{C}) and commuting with the natural projections to \Gamma_{F} ; two L -parameters \phi , \phi^{\prime} being equivalent if they are conjugate by an element of the dual group \widehat{G} . We also expect that each L -packet \Pi^{G}(\phi) can be parametrized by a certain subset of irreducible representations of the (finite) group of components \displaystyle S_{\phi}:=\pi_{0}(Z_{\widehat{G}}(\phi)) of the centralizer Z_{\widehat{G}}(\phi) of \phi in \widehat{G} . That is to say there should exists an injective map \displaystyle\Pi^{G}(\phi)\hookrightarrow\Irr(S_{\phi}), \displaystyle\pi\mapsto\rho_{\pi,\varpi}, which, as the notation suggests, depends on an auxiliary choice \varpi which is that of a Whittaker datum for one of the quasi-split pure inner form G_{qd} of G 1 1 1 Recall that a Whittaker datum is a pair (N,\xi) consisting of the unipotent radical N of a Borel subgroup (defined over F ) and of a non-degenerate character \xi of N , taken up to G_{qd} -conjugacy. . Let A_{\widehat{G}}=Z(\widehat{G})^{\Gamma_{F},0} be the neutral component of the subgroup of \Gamma_{F} -fixed points in the center of \widehat{G} . We write \Ad_{G} for the adjoint representation of {}^{L}G on \Lie(\widehat{G})/\Lie(A_{\widehat{G}}) . For \pi\in\Irr(G) , let us denote by \phi_{\pi} its L -parameter (i. e. \pi\in\Pi^{G}(\phi_{\pi}) ) and let \displaystyle\gamma(s,\pi,\Ad_{G},\psi):=\gamma(s,\Ad_{G}\circ\phi_{\pi},\psi)=\epsilon(s,\Ad_{G}\circ\phi_{\pi},\psi)\frac{L(1-s,\Ad\circ\phi_{\pi})}{L(s,\Ad\circ\phi_{\pi})} be its adjoint gamma factor. Here, L(s,\Ad_{G}\circ\phi_{\pi}) denotes the local L -factor of the Weil-Deligne representation \Ad_{G}\circ\phi_{\pi} whereas \epsilon(s,\Ad_{G}\circ\phi_{\pi},\psi) stands for the local \epsilon -factor. The main conjecture of [ HII08 ] on formal degrees can now be stated as follows. Conjecture 1.1 (Hiraga-Ichino-Ikeda) . Let \pi be a square-integrable irreducible representation of G . Then, we have \displaystyle d_{\psi}(\pi)=\frac{\deg(\rho_{\pi,\varpi})}{[X^{*}(A_{G}):X^{*}(G)]\lvert S_{\phi_{\pi}}\rvert}\lvert\gamma(0,\pi,\Ad_{G},\psi)\rvert, where \deg(\rho_{\pi,\varpi}) denotes the degree of the representation \rho_{\pi,\varpi} of S_{\phi_{\pi}} . Remark 1.1 . The conjecture includes implicitely the fact that \gamma(s,\pi,\Ad_{G},\psi) is regular at s=0 . However, it is expected that, for \pi square-integrable, the L -parameter \phi_{\pi} is discrete, which means that its centralizer Z_{\widehat{G}}(\phi_{\pi}) is finite modulo A_{\widehat{G}} , and this last condition is readily seen to imply that \gamma(s,\pi,\Ad_{G},\psi) is regular for \Re(s)\geq 0 . Since the formal degree d_{\psi}(\pi) is obviously independent of the Whittaker datum \varpi , the conjecture also implies that the degree \deg(\rho_{\pi,\varpi}) doesn’t depend on \varpi either. There is however a precise expectation on the dependence of \rho_{\pi,\varpi} on \varpi , see [ GGP12 , §9, (3)] , namely changing the Whittaker datum should result in twisting \rho_{\pi,\varpi} by a certain one-dimensional character S_{\phi_{\pi}}\to\mathbb{C}^{\times} ; an operation that clearly doesn’t affect the degree of the representation. Finally, the denominator in the above formula is slightly different from the one in [ HII08 ] . Namely, in op. cit. the factor in the denominator is the cardinality of the centralizer S_{\phi_{\pi}}^{\natural} in the dual group of G^{\natural}:=G/A_{G} . However it is easy to see that for discrete L -parameters we have \displaystyle\lvert S_{\phi_{\pi}}^{\natural}\rvert=[X^{*}(A_{G}):X^{*}(G)]\lvert S_{\phi_{\pi}}\rvert. We also emphasize that this conjecture translates into a more natural one, that is without the factor [X^{*}(A_{G}):X^{*}(G)] , for the Plancherel density (suitably normalised) of square-integrable representations; see the next section. 1.3 Plancherel density When combined with a

Read the complete text