betaFeedback

WorksMathematics

Exceptional zeros for GSp(4)×GL(2)\mathrm{GSp}(4) \times \mathrm{GL}(2) and Yoshida lifts

Abstract

The aim of this article is to study an exceptional zero phenomenon for the pp-adic LL-functions attached to GSp4×GL2\mathrm{GSp}_4\times\mathrm{GL}_2 in the critical regions (F) and (D). We construct improved pp-adic LL-functions by replacing the pp-depleted Eisenstein family with its pp-ordinary counterpart. Under a local compatibility hypothesis at the ramified places and classical control for the families, we show that their poles detect exactly the Yoshida lifts whose smaller constituent is dual to the GL2\mathrm{GL}_2 representation. The proof compares their residues with distinguished automorphic periods, extending the coherent period formula to the case where one modular form has weight zero. An explicit Euler-factor identity relates each improved function to the usual pp-adic LL-function. At these Yoshida points, the vanishing factor cancels the pole, and the resulting value is expressed through the distinguished period and the logarithmic derivative of the ordinary elliptic root. We also discuss the expected factorization along Yoshida families into Hida–Rankin pp-adic LL-functions.

Introduction

The poles of the complex LL-function for GSp4×GL2\mathrm{GSp}_{4} \times\mathrm{GL}_{2} detect a natural relation between its two inputs: the Siegel representation is a Yoshida lift, and the GL2\mathrm{GL}_{2} (elliptic) representation is dual to one of its constituents. This relation is also characterized by the nonvanishing of a distinguished period. We seek a pp-adic counterpart of this criterion in the critical regions (F) and (D). At the relevant weights, an interpolation factor may vanish and cancel a pole. Thus, removing this factor leads to improved pp-adic LL-functions whose residues recover the distinguished period.

The two regions lead to different coherent realizations: in (F), a degree-two class on the Siegel threefold pairs with two degree-zero elliptic classes, while in (D) the Siegel class and the second elliptic class both have degree one. In either construction, the first GL2\mathrm{GL}_{2} factor contributes an Eisenstein series. Before pp-depletion, its pp-adic family has a simple pole with constant residue. Taking this residue should then recover the distinguished period. The difficulty is that the resulting coherent pairing has first elliptic weight zero, beyond the range of the positive-weight period formula used in the interpolation theorem.

We establish the period formula in both coherent degrees. Using the automorphic realization of coherent cohomology, we identify the cocycles and coefficient contractions that enter the pairing, and recover the global integral through the Serre trace. The complex and pp-adic constructions can then be compared as realizations of the same rational pairing. Together with the local test-vector calculations, this gives a common Yoshida criterion for the poles of the two improved functions. Moreover, an explicit Euler-factor identity relates each of them to its ordinary counterpart. We call a Yoshida point matching if the GL2\mathrm{GL}_2 representation is contragredient to the smaller Yoshida constituent, with the corresponding ordinary refinement. At a matching Yoshida point, the ordinary value is governed by the coherent period and the logarithmic derivative of the elliptic ordinary root.

The complex pole and the distinguished period

Let Π\Pi be a unitary globally generic cuspidal automorphic representation of GSp4(A)\mathrm{GSp}_4(\mathbb{A}), of cohomological weight (k1,k2)(k_1,k_2) with k1≥k2≥3k_1 \ge k_2 \ge3, and let Σ\Sigma be a unitary cuspidal representation of GL2(A)\mathrm{GL}_2(\mathbb{A}). Put

H=GL2×GL1GL2,H = \mathrm{GL}_2 \mathop{\times}_{\mathrm{GL}_1} \mathrm{GL}_2,

with the equal-determinant embedding ι:H↪GSp4\iota: H \hookrightarrow\mathrm{GSp}_4. When ωΠωΣ=1\omega_\Pi\omega_\Sigma= 1, the global period is

I(Ψ,F)=∫ZG(A)H(Q)\H(A)Ψ(ι(h))F(h2) dh,\mathcal{I}(\Psi,F) = \int_{Z_G(\mathbb{A})H(\mathbb{Q})\backslash H(\mathbb{A})} \Psi(\iota(h))F(h_2)\,dh,

where ZGZ_G is the common scalar centre. We call the pair distinguished if this integral is nonzero for some Ψ∈Π\Psi\in\Pi and F∈ΣF \in\Sigma.

Write Λ(s,Π×Σ)\Lambda(s,\Pi\times\Sigma) for the completed degree-eight Langlands LL-function in unitary normalization. Its relation to Yoshida lifts follows from the generic spin transfer of Asgari–Shahidi and the Rankin–Selberg pole theorem. Novodvorsky’s integral connects the pole with I\mathcal{I}: the residue of its Eisenstein series at s=1s=1 is constant. For this integral comparison we require equality of the local Novodvorsky and Langlands factors at ramified finite places, as in the assumption 3.1. This holds, in particular, when each corresponding Σv\Sigma_v is non-supercuspidal [22]. We recall the resulting criterion in Section 3.2.1.

Theorem. Let Π\Pi and Σ\Sigma be as above, with ωΠωΣ=1\omega_\Pi\omega_\Sigma= 1. Under the local compatibility expressed in 3.1, the following conditions are equivalent:

(i) Λ(s,Π×Σ)\Lambda(s,\Pi\times\Sigma) has a pole at s=1s=1;

(ii) Π\Pi is a Yoshida lift of Σ∨⊠Σ′\Sigma^\vee\boxtimes\Sigma' for a cuspidal representation Σ′≄Σ∨\Sigma' \not\simeq\Sigma^\vee with ωΣ′=ωΣ−1\omega_{\Sigma'} = \omega_\Sigma^{-1};

(iii) the pair (Π,Σ)(\Pi,\Sigma) is distinguished.

When these conditions hold, the pole is simple.

This criterion supplies the complex input to our argument. Its proof combines [1 GT11, 15] with the residue and local-integral results recalled in Section 3. For the pp-adic application, we evaluate this period on automorphic forms with prescribed compact types at infinity and ordinary vectors at pp. When the pair is distinguished, the test-vector calculations below show that these choices allow a nonzero period for suitable data at the remaining finite places.

The coherent period at weight zero

The boundary relevant to both regions occurs at elliptic weight

c0=k1−k2+2,c_0 = k_1 - k_2 + 2,

the weight of the smaller Yoshida constituent. Taking the Eisenstein residue replaces the form on the first GL2\mathrm{GL}_2 factor by 11, leaving a pairing between the Siegel class and an elliptic cusp class of weight c0c_0.

The positive-weight formula of [23 Theorem 5.6(b)] is stated with first elliptic weight 1+t11+t_1, where t1≥0t_1 \ge0, while our specialization has t1=−1t_1=-1. To reach it, we work with the Dolbeault representatives in the Harris–Su realization [10, 39] and check that restriction to the HH-surface, exterior product and algebraic coefficient contraction commute with the automorphic lift. At weight zero, the contractions select Siegel vectors of compact weights (0,−c0)(0,-c_0) and (0,c0)(0,c_0), paired respectively with holomorphic and antiholomorphic elliptic representatives. Evaluating the corresponding cochains gives nonzero constants. Then, cuspidal decay allows us to identify the integral with the trace on the toroidal compactification, with the constant form in the first elliptic factor.

The resulting formula is stated in Theorem 4.22. Writing A⋄A_{\diamond} for the proportionality scalar in the weight-zero formula of that theorem, we obtain

PF(η,1⊠g)=AFIFem,PD(ξ,1⊠v)=ADIDem,m=k1+k2−6.P_F(\eta,1\boxtimes g)=A_F I_F e_m,\qquad P_D(\xi,1\boxtimes v)=A_D I_D e_m,\qquad m=k_1+k_2-6.

Here PF,PDP_F,P_D are rational coherent pairings, and IF,IDI_F,I_D are the periods of the corresponding automorphic representatives. The pairings take values in the same one-dimensional character space E{m}E\{m\}; its basis eme_m records the twist relating the arithmetic coefficients to unitary automorphic forms. The nonzero constants AF,ADA_F,A_D account for the cocycle, contraction, measure and trace normalizations.

On the pp-adic side, the analytic branching kernels of [LZ21, LR25] remain polynomials at weight zero. Lemma 6.11 therefore identifies their specializations with the algebraic contractions. This allows us to compare the complex period and the pp-adic residue through the same rational pairing.

Improved functions and their poles

Fix ordinary families Π,Σ\Pi,\Sigma satisfying the classical-control hypotheses of § 7. Their classical weights are denoted (k1,k2,c)(k_1,k_2,c), with the underlying eigenfamily points understood. If ss is the unitary variable and s′s' the motivic variable, then

s′=s+k1+k2+c−42.s'=s+\frac{k_1+k_2+c-4}{2}.

Our constructions lie on the three-variable domains

WFimpr:s′=k1−1,WDimpr:s′=c+k2−3.\mathcal{W}^{\mathrm{impr}}_F:s'=k_1-1,\qquad\mathcal{W}^{\mathrm{impr}}_D:s'=c+k_2-3.

Their Eisenstein weights are

tF=k1−k2−c+2,tD=−k1+k2+c−2.t_F=k_1-k_2-c+2,\qquad t_D=-k_1+k_2+c-2.

At the boundary of interest both Eisenstein characters are trivial and s=0s=0 on the two domains. This is the point paired with s=1s=1 by the functional equation, which also replaces the representations by their contragredients. For a distinguished pair, the completed function has a pole at both points. In Novodvorsky’s integral, the Eisenstein residue at s=0s=0 is again constant: with the notation of Proposition 3.4, it is −cEφ(0)-c_E\varphi(0), instead of cEφ^(0)c_E\widehat{\varphi}(0) at s=1s=1 [30 Proposition 2.5]. Our comparison passes through this common distinguished period.

We construct Lp(F),imprL_p^{(F),\mathrm{impr}} and Lp(D),imprL_p^{(D),\mathrm{impr}} by using the ordinary Eisenstein family before depletion in the two coherent pairings. Its pole supplies the possible pole of each improved function. To compute the residue, we first regularize the Eisenstein family in the analytic coefficient module where the pairing is defined. We can then specialize inside the pairing, replacing the Eisenstein series by a multiple of the constant form. The generators and auxiliary test data are fixed on the family neighbourhood, as explained in § 8.

Fix a good classical point PP of the Siegel family and a crystalline point Q0Q_0 of weight c0c_0 of the GL2\mathrm{GL}_2 family, and put Π=Π(P)\Pi=\Pi(P) and Σ=Σ(Q0)\Sigma=\Sigma(Q_0). We assume the unramified ordinary refinements, classical control and étale elliptic weight map specified in § 9.3, with both Eisenstein characters trivial at Q0Q_0. We also retain the central-character condition and local compatibility of Theorem 1.1.

Theorem A. For suitable tame test data, the following conditions are equivalent:

  1. Lp(F),impr(Π×Σ)L_p^{(F),\mathrm{impr}}(\Pi\times\Sigma) has a pole at Q0Q_0;

  2. Lp(D),impr(Π×Σ)L_p^{(D),\mathrm{impr}}(\Pi\times\Sigma) has a pole at Q0Q_0;

  3. Π\Pi is a Yoshida lift with Σ∨\Sigma^\vee as its constituent of weight c0c_0;

(iv) the pair (Π,Σ)(\Pi,\Sigma) is distinguished.

When these conditions hold, both poles are simple. For any choice of tame data, a pole of either improved function implies (iii).

The equivalence comes from comparing the two residues with the distinguished period. Let zz be the logarithmic elliptic weight centred at Q0Q_0, so that ℓ(tF)=−z\ell(t_F)=-z and ℓ(tD)=z\ell(t_D)=z. The Kubota–Leopoldt normalization gives ℓ(κ)E(κ)∣κ=0=−(1−p−1)1\ell(\kappa)\mathcal{E}(\kappa)|_{\kappa=0}=-(1-p^{-1})1, and hence

Res⁡z=0Lp(F),impr=(1−p−1)BF(1),Res⁡z=0Lp(D),impr=−(1−p−1)BD(1),\operatorname{Res}_{z=0} L_p^{(F),\mathrm{impr}}=(1-p^{-1})B_F(1),\qquad\operatorname{Res}_{z=0} L_p^{(D),\mathrm{impr}}=-(1-p^{-1})B_D(1),

where B⋄(1)B_{\diamond}(1) is the constant-input pairing defined in (68). Comparing rational structures as in [29 Theorem 3.10], we obtain

Res⁡z=0Lp(⋄),imprΩ⋄,p=ε⋄(1−p−1)ιp(νKA⋄I⋄Ω⋄,∞),εF=1,εD=−1.\frac{\operatorname{Res}_{z=0}L_p^{(\diamond),\mathrm{impr}}}{\Omega_{\diamond,p}}=\varepsilon_{\diamond}(1-p^{-1})\iota_p\left(\frac{\nu_KA_{\diamond}I_{\diamond}}{\Omega_{\diamond,\infty}}\right),\qquad\varepsilon_F=1,\quad\varepsilon_D=-1.

Here νK\nu_K is the tame volume multiplier in the definition of the global functions; the constants A⋄A_{\diamond} still refer to the geometric pairings P⋄P_{\diamond}. The expression inside ιp\iota_p is algebraic by Proposition 6.24: observe that after dividing by their comparison periods, the complex integral and the pp-adic residue come from the same rational pairing. In particular, the residue is nonzero precisely when the period is nonzero on the chosen vectors.

For the converse, the test-vector calculations show that the distinguished period can be nonzero with our prescribed vectors at pp and infinity. Choosing suitable tame data then gives a nonzero residue in both regions. The residue also varies analytically along the Eisenstein divisor; near a matching Yoshida point it identifies the smaller constituent at nearby good classical points, as explained after Theorem 6.25.

The Euler-factor identities

Let α,β,γ,δ\alpha,\beta,\gamma,\delta be the motivic Siegel Hecke parameters, ordered by the ordinary refinement, and let aa be the ordinary elliptic root. The normalized parameters

References

  1. [1]Mahdi Asgari and Freydoon Shahidi, Generic transfer from GSp(4) to GL(4), Compositio Mathematica 142 (2006), no. 3, 541–550. 2, 15DOI
  2. [2]George Boxer and Vincent Pilloni, Higher Coleman theory, preprint, 2020. 5, 30, 40, 41, 42, 52arxiv.org/abs/2110.10251
  3. [3]———, Higher Hida and Coleman theories on the modular curve, Épijournal de Géométrie Algébrique 6 (2022). 34, 42, 43DOI
  4. [4]———, Higher Hida theory for Siegel modular forms, Invent. Math. 244 (2026), no. 1, 45–141 (English). 31, 41, 42DOI
  5. [6]Pierre Deligne, La conjecture de Weil: I, Publications Mathématiques de l’IHÉS 43 (1974), 273–307. 13, 49DOI
  6. [7]Glenn J. Fox, A method of Washington applied to the derivation of a two-variable p-adic L-function, Pacific Journal of Mathematics 209 (2003), no. 1, 31–40. 38DOI
  7. [8]Andrew Graham and Rob Rockwood, Nearly higher Coleman theory and p-adic l-functions for GSp(4) × GL(2) and GSp(4) × GL(2) × GL(2), 2024, preprint, available at https://arxiv.org/pdf/2411.04559.pdf. 5, 33, 39
  8. [10]Michael Harris, Automorphic forms of ̄σ-cohomology type as coherent cohomology classes, J. Differ. Geom. 32 (1990), no. 1, 1–63 (English). 2, 21, 22, 25DOI
  9. [11]Haruzo Hida, A p-adic measure attached to the zeta functions associated with two elliptic modular forms. II, Ann. Inst. Fourier (Grenoble) 38 (1988), no. 3, 1–83. MR 976685 6, 7, 8, 9, 10
  10. [12]———, Greenberg’s L-invariants of adjoint square Galois representations, Int. Math. Res. Not. (2004), no. 59, 3177–3189. 10
  11. [13]———, Constancy of adjoint L-invariant, J. Number Theory 131 (2011), 1331–1346. 10, 59
  12. [14]Aleksander Horawa and Kartik Prasanna, Motivic action for Siegel modular forms, Journal of the European Mathematical Society (2025), Published online. 13, 26arxiv.org/abs/2307.04115
  13. [15]Hervé Jacquet, Ilya I. Piatetski-Shapiro, and Joseph Shalika, Rankin-Selberg convolutions, Amer. J. Math. 105 (1983), 367–464 (English). 2, 7, 15DOI
  14. [16]Guido Kings, David Loeffler, and Sarah Livia Zerbes, Rankin-Eisenstein classes and explicit reciprocity laws, Camb. J. Math. 5 (2017), no. 1, 1–122. 7, 8DOI
  15. [17]Anthony W. Knapp, Representation theory of semisimple groups. An overview based on examples. With a new preface by the author., paperback ed. ed., Princeton, NJ: Princeton University Press, 2001 (English). 13
  16. [18]Francesco Lemma, Algebraic cycles and residues of degree 8 L-functions of GSp(4) × GL(2), Int. Math. Res. Not. IMRN (2020), no. 12, 3614–3648. MR 4120306 5
  17. [19]Zheng Liu, p-adic L-functions for GSp(4) × GL(2), preprint, 2023. 5, 33
  18. [20]———, p-adic L-functions for GSp(4) × GL(2) (ii), preprint, 2024. 5, 33, 60
  19. [21]David Loeffler, On some zeta-integrals for unramified representations of GSp(4), 2021, preprint, available at https://arxiv.org/pdf/2106.16076.pdf. 46, 47, 50, 52
  20. [22]———, On local zeta-integrals for GSp(4) and GSp(4) × GL(2), New York J. Math. 30 (2024), 1–23 (English). 2, 15
  21. [23]David Loeffler, Vincent Pilloni, Christopher Skinner, and Sarah Livia Zerbes, Higher Hida theory and p-adic L-functions for GSp(4), Duke Math. J. 170 (2021), no. 18, 4033–4121. 2, 13, 16, 17, 18, 19, 20, 24, 26, 27, 31, 32, 33, 37, 38, 39, 50, 58DOI
  22. [24]David Loeffler and Óscar Rivero, Algebraicity of L-values for GSp₄ × GL₂ and GSp₄ × GL₂ × GL₂, Q. J. Math. 75 (2024), no. 2, 391–412. MR 4765775 19, 30, 32, 46, 47, 48DOI
  23. [25]———, On p-adic L-functions for GSp(4) × GL(2), Pacific J. Math. 335 (2025), no. 2, 373–400. 3, 4, 5, 19, 30, 31, 32, 33, 34, 35, 36, 37, 39, 40, 42, 44, 45, 51, 54, 58
  24. [26]David Loeffler, Christopher Skinner, and Sarah Livia Zerbes, Euler systems for GSp(4), J. Eur. Math. Soc. (JEMS) 24 (2022), no. 2, 669–733. MR 4382481 13DOI
  25. [27]David Loeffler and Sarah Livia Zerbes, p-adic L-functions and diagonal cycles for GSp(4) × GL(2) × GL(2), preprint, 2020. 30, 32, 33arxiv.org/abs/2011.15064
  26. [29]David Loeffler and Sarah Livia Zerbes, Poles of p-adic asai L-functions and distinguished representations, preprint, 2025. 4, 5, 37, 38, 55DOI
  27. [30]Tomonori Moriyama, L-functions for GSp(2) × GL(2): archimedean theory and applications, Can. J. Math. 61 (2009), no. 2, 395–426 (English). 3, 13, 14, 15, 16DOI
  28. [31]Vincent Pilloni, Overconvergent modular forms, Annales de l’Institut Fourier 63 (2013), no. 1, 219–239. 38DOI
  29. [32]———, Higher coherent cohomology and p-adic modular forms of singular weights, Duke Math. J. 169 (2020), no. 9, 1647–1807. 5, 13DOI
  30. [33]Ilya I. Piatetski-Shapiro and David Soudry, L and ε factors for GSp(4), J. Fac. Sci. Univ. Tokyo Sect. IA Math. 28 (1981), no. 3, 505–530 (1982). MR 656034 14
  31. [35]Óscar Rivero and Victor Rotger, Derived Beilinson–Flach elements and the arithmetic of the adjoint of a modular form, J. Eur. Math. Soc. (JEMS) 23 (2021), no. 7, 2299–2335. MR 4269414 6, 10
  32. [36]Ralf Schmidt, The Saito–Kurokawa lifting and functoriality, American Journal of Mathematics 127 (2005), no. 1, 209–240. 61DOI
  33. [38]David Soudry, The L and γ factors for generic representations of GSp(4,k)×GL(2,k) over a local nonarchimedean field k, Duke Math. J. 51 (1984), 355–394 (English). 15DOI
  34. [39]Jun Su, Coherent cohomology of Shimura varieties and automorphic forms, Ann. Sci. Éc. Norm. Supér. (4) 57 (2024), no. 4, 1039–1099. MR 4773300 2, 24, 25DOI
  35. [40]J. Tilouine and E. Urban, Several-variable p-adic families of Siegel-Hilbert cusp eigensystems and their Galois representations, Ann. Sci. École Norm. Sup. (4) 32 (1999), no. 4, 499–574. MR 1693583 32
  36. [41]David A. Vogan and Gregg J. Zuckerman, Unitary representations with non-zero cohomology, Compositio Mathematica 53 (1984), no. 1, 51–90. 17
  37. [42]Nolan R. Wallach, Real reductive groups II, Pure and Applied Mathematics, vol. 132-II, Academic Press, 1992. 16

Paper details

vlnc:W_jq9mr64d
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

GPT-6 Astra (bare), Claude Opus 5.5 (bare)

View statement
Looking for related papers…
Contents