We establish nontrivial bounds for Kloosterman sums over p-adic fields associated with the longest Weyl element on arbitrary split reductive groups, at arbitrary level. As an application, we obtain the regularity of Bessel distributions (of Whittaker models) for all such groups.
Introduction
Bounding Kloosterman sums is a fundamental problem in number theory, with many important applications.
In the classical GL2 case, Salié [24] obtained early estimates. Weil [Wei48] established the celebrated square-root bound. Such estimates played an important role in Selberg’s work on the spectral theory of automorphic forms.
The study of Kloosterman sums beyond GL2 was pioneered by Bump, Friedberg and Goldfeld [2]. Subsequently, Friedberg [11] and Stevens [25] simultaneously and independently developed the theory for general GLn. Related developments for orthogonal groups were carried out by Cogdell, Li, Piatetski-Shapiro and Sarnak [6]. More recently, significant progress has been made for particular families of groups by Assing and Blomer [1], Blomer and Man [3], Man [19, 20], Linn [18], Mujdei [23], Huang (in the appendix of [12]) and Miao [22].
A uniform theory of Kloosterman sums for general split groups was initiated by Dąbrowski [8] and further developed by Dąbrowski and Reeder [9]. In particular, the latter obtained an explicit formula for the generating function of the cardinalities of Kloosterman sets, yielding general trivial bounds. Our recent works [4 CLYY26] further investigate Kloosterman sums on split groups and their connections with Bessel distributions.
In this paper, we establish nontrivial bounds for Kloosterman sums associated with the longest Weyl element on arbitrary split reductive groups at any level.
Our approach to obtain the power-saving counting estimates is through a representation-theoretic analysis of Plücker height zeta functions, using the Casselman-Shalika formula for spherical Whittaker functions and the Weyl dimension formula. This avoids explicit calculations of Plücker relations for individual groups.
Let G be a split reductive group over a p-adic field F with derived subgroup Gder. Let B=AN be a Borel subgroup of G with A a maximal torus of G and N the unipotent radical of B. Denote by δ the modulus character of B. We choose a representative w0∈NG(A) of the longest Weyl element. Let ψ0 be the generic character on N given as in § 2. For an open compact subgroup K of G such that ψ0∣N∩K=1, we consider the following Kloosterman sum
In the above, if [n1w˙0tn2]∈XK(w˙0t) with n1,n2∈N and t∈A, then
u([n1w˙0tn2])=[n1],u′([n1w˙0tn2])=[n2].
We say that the Kloosterman sum Kl has a nontrivial bound ([4 Definition 3.17]) if there exist constants ε>0 and C>0 (possibly depending on p) such that
∣Kl(t)∣δ1/2−ε(t)≤C,t∈A∩Gder.
The following is the main result of this paper.
Theorem 1.1. Let G be a split reductive group over F. Then for every compact open subgroup K⊂G such that ψ0∣N∩K=1, the Kloosterman sum Kl(⋅,ψ0−1,w˙0,K) has a nontrivial bound.
As an application of Theorem 1.1, we obtain the regularity of Bessel distributions Bπ on G as follows (See [4] for the survey of this problem).
Denote by S(G) the space of Schwartz functions on G. Let π be an irreducible smooth admissible representation on G with π∗ denoting its linear dual. Let π∨ be the contragredient representation of π, that is, the smooth part of π∗. The group G then acts on all these spaces and these actions induce actions of S(G). Note that for any ℓ∈(π∨)∗ and f∈S(G), π∨(f)ℓ is smooth so that π∨(f)ℓ∈(π∨)∨=π. For any ℓ1∈π∗ and ℓ2∈(π∨)∗, the relative character associated to ℓ1 and ℓ2 is defined as
Bℓ1,ℓ2(f)=ℓ1(π∨(f)ℓ2),f∈S(G).
When π is generic with respect to ψ0, and both ℓ1,ℓ2 are Whittaker functionals, the associated relative character Bπ=Bℓ1,ℓ2 is called Bessel distribution attached to π.
Theorem 1.2. Suppose G is a split reductive group over F. For any irreducible admissible generic representation π of G, the Bessel distribution Bπ with respect to the generic character ψ0 is regular.
Proof. By Theorem [4 Theorem 3.24], it is enough to prove that for any subset I of the set of simple roots Δ for G, the Kloosterman sum KlMI on the standard Levi MI associated to G with respect to the triple (ψ0∣NMI,w˙I,KI) has a nontrivial bound. Here, KI is some admissible group of MI ([4 Definition 2.25]) and {w˙I}I⊂Δ is a system of representatives of {wI}I⊂Δ. Applying Theorem 1.1 to each MI, we obtain the regularity. □
We sketch the proof Theorem 1.1. For this, we assume G is semisimple and simply connected. Stevens’ approach [25] is a basic method for bounding Kloosterman sums. By considering the action of A on the Kloosterman set, the contribution of each orbit in the Kloosterman sum on G can be decomposed into a product of Kloosterman sums on GL2. Applying the Weil bound for these Kloosterman sums on GL2, the problem is reduced to a counting problem for the Kloosterman set.
Plücker coordinates are generators of F[N\G]. Based on the homogeneity of Plücker coordinates, the counting problem is reduced to that of the Plücker-constrained sets Sr, r∈Z≥0m with m the rank of G, which are subsets of affine spaces defined by conditions on Plücker coordinates (Definition 3.6). This reduction is the main result of [CLYY26] (Theorem 5.1).
Here, fs is the normalized spherical section of certain unramified principal series on G.
The Casselman–Shalika formula expresses the spherical Whittaker function Js on G in terms of Weyl characters of finite-dimensional representations on the Langlands dual group Gad∨ of the adjoint quotient Gad of G (Theorem 4.6). We obtain that ZPl(s) extends holomorphically to the half-plane (Theorem 4.4(1))
Re(s)>1−MG1.
Here, MG=maxiai with 2ρ=∑i=1maiαi where 2ρ is the sum of positive roots and αi, 1≤i≤m are the simple roots.
Moreover, the Weyl dimension formula gives the bounding of ZPl(s) for s near its convergence boundary (Theorem 4.4(2)). This estimate for ZPl(s) implies the following bound for Plücker constrained sets Sr (Theorem 4.8)
∣Sr∣≪q(1−2MG1)R,R=i∑ri.
By Theorem 5.1, this gives a nontrivial bound for Kloosterman sums on G with level Kd, d≫0 (Corollary 5.2). Here, {Kd}d is the family of principal congruence subgroups of G.
By the germ expansion of Kloosterman integrals (a generalization of Kloosterman sums) in [4], together with the corresponding bounds for standard Levi subgroups, we obtain non-trivial bounds for Kloosterman sums with arbitrary level.
It is instructive to compare our approach with that of Dąbrowski and Reeder [9 Section 2], whose method yields the trivial bound for Kloosterman sums.
Acknowledgements. L. Cai is supported by the National Key R&D Program of China No. 2023YFA1009702 and National Natural Science Foundation of China, No. 12371012. J. Chai is supported by National Natural Science Foundation of China, No. 12571010 and No. 12671039.
Use of artificial intelligence. ChatGPT was used for exploratory computations, editing, searching for references and for verification of results.
Notations and conventions
Let F be a p-adic field. Denote by O the ring of integers, p the maximal ideal of O and q the cardinality of its residue field. Denote by ∣⋅∣ the p-adic absolute value on F with v the additive valuation. We use ϖ to denote a fixed uniformizer of F.
Let G be a split (connected) reductive group over F. We shall make no distinction between an algebraic group and its F-points.
We fix a maximal torus A of G and a Borel subgroup B=AN of G. Denote by Z the center of G. Let Φ+ be its positive roots and Δ its simple roots. Put L=∣Φ+∣ and m=∣Δ∣. Let α1,…,αm∈Δ be the simple roots and ω1,…,ωm be the fundamental weights.
Denote by
2ρ=γ∈Φ+∑γ.
Then
ρ=i∑ωi.
Denote by X∗(A) and X∗(A) the groups of characters and cocharacters on A. For any a∈A, there is a unique cocharacter ν(a)∈X∗(A) such that
⟨χ,ν(a)⟩=v(χ(a)),χ∈X∗(A).
Denote by δ the modulus character of B. Then
δ(a)=q−⟨2ρ,ν(a)⟩,a∈A.
By the existence theorem for split reductive group schemes (see [7 Exposé XXV, Theorem 1.1 and Corollary 1.2]), there exists a split reductive group scheme GZ over Z, a split maximal torus AZ and a Borel subgroup BZ=AZNZ, together with an isomorphism
GZ⊗ZF∼G
such that under this isomorphism, AZ⊗ZF and BZ⊗ZF go to A and B, respectively.
We fix a Chevalley system for GZ ([7 Exposé XXIII, Proposition 6.2])
Xγ:Ga,Z∼Nγ,Z,γ∈Φ.
We shall use the same notation Xγ for their base changes to F.
Denote by G=GZ⊗ZO. Then G is a split reductive O-model of G. Denote by A, B, and N the corresponding subgroup schemes obtained by base change from AZ, BZ and NZ respectively. In the following, all notations involving O-point refer to this model.
Let W=NG(A)/A be the Weyl group. For each simple root αi, let wi∈W be the corresponding simple reflection. Consider
w˙i=Xαi(1)X−αi(−1)Xαi(1).
Then w˙i∈NGZ(AZ)(Z) represents wi. For each w∈W, choose a reduced expression w=wk1⋯wkr and set
w˙=w˙k1⋯w˙kr.
The definition of w˙ is independent on the choice of the above reduced expression of w. In particular, we fix a representative w˙0 of the longest Weyl element w0 which satisfies
w˙0∈NGZ(AZ)(Z)⊂NG(A)(F)∩G(O).
We use this representative throughout the paper.
We fix an ordering Φ+={γ1,…,γL} by increasing height. With respect to this ordering, every element in N can be written uniquely as
n(x)=γ∈Φ+∏Xγ(xγ),x=(xγ)γ∈Φ+∈FΦ+.
If n=n(x) for some x∈FΦ+, then for any γ∈Φ+, we write xγ(n) for xγ.
We fix a nontrivial character ψF on F. We assume that ψF∣O=1, and that ψF∣ϖ−1O=1. For each c=(ci)i∈Fm, denote by
ψc(n)=ψF(i∑cixαi(n)).
We shall write ψ0=ψc if ci=1 for all i.
The Plücker height
In this section, we assume G is semisimple and simply connected. For each i=1,…,m, let (πi,Vi) be an irreducible (algebraic) representation of G with highest weight ωi∈X∗(A).
Lemma 3.1. There exists a G(O)-stable lattice Λi in Vi with a O-basis {eij}j of Λi such that eij are weight vectors and the basis containing a generator ei+ of the highest-weight space.
Proof. By [15 Lemma in I.10.4], there is a G-stable lattice Λi in Vi and hence is a G(O)-stable lattice of Vi. Consider the restriction of the action G on Λi to the split torus A. As the group A is diagonalizable, by [15 (3) in I.2.11], Λi decomposes into A(O)-eigen subspace:
Λi=μ∈X∗(A)⨁Λi,μ.
Here, X∗(A)=Hom(A,Gm,O) and Λi,μ consists of v∈Λi such that πi(a)v=μ(a)v for all a∈A(O).
After extending scalars to F, this is the weight-space decomposition of $V_i
Λi,μ⊗OF=Vi,μF,Λi,μ=Λi∩Vi,μF.
Here we use the following canonical isomorphism ([17 Fact B.1.2])
X∗(A)∼X∗(A),μ↦μF:=μ⊗F.
The highest-weight space Vi,ωi is one-dimensional, so its lattice Λi,ωi has rank one. Choose an O-basis in each Λi,μ, choosing a generator ei+ for the highest-weight summand. The union of these bases is a weight basis of Vi and an O-basis of Λi. □
For each i, denote by Λi∗=HomO(Λi,O) the dual lattice of Λi. A linear functional ℓi∈Λi∗ is called primitive if there is no ℓ′∈Λi∗ such that ℓ=ϖℓ′.
Lemma 3.2. For each i, there exists a unique primitive linear functional ℓi∈Λi∗ such that
ℓi∘πi(n)=ℓi for any n∈N.
ℓi∘πi(a)=w˙0ωi(a)ℓi for any a∈A.
ℓi(πi(w˙0)ei−)=1.
Proof. We give a construction of the required ℓi. Consider the weight decomposition in Lemma 3.1
Λi=μ⨁Λi,μ.
Denote by λi=w˙0ωi. Then λi is the lowest weight of Vi and Λi,λi=Oei− is of rank one. Consider ℓi∈Λi∗ such that
ℓi(ei−)=1,ℓi∣Λi,μ=0,μ=λi.
For any weight vector vμ∈Vμ and n∈N, we have
π(n)vμ=vμ+ν>μ∑vν.
In particular, for any μ>λi, ℓi(π(n)vμ)=0=ℓi(vμ) and ℓi(πi(n)vλi)=ℓi(vλi). Therefore, ℓi is N-invariant.
As the torus A acts on Vλi via λi=w˙0ωi, ℓiπi(a)=λi(a)ℓi for any a∈A.
We have πi(w˙0)ei+∈Vλi. As we require that w˙0∈G(O),
πi(w˙0)ei+=uei−,u∈O×
so that ℓi(πi(w˙0)ei+)=u. Modifying ℓi to u−1ℓi, we obtain the required ℓi∈Λi∗. The uniqueness of such linear functional is clear. □
Definition 3.3 (The Plücker height). For each i, we choose a G(O)-stable lattice Λi of Vi and an O-basis {eij}j of Λi as in Lemma 3.1. We also choose a primitive linear functional ℓi∈Λi∗ as in Lemma 3.2. Denote by
For each i, Hi(n)≥1 and Hi(nn0)=Hi(n) for any n∈N and n0∈N(O). In particular, for each n∈N, there exists a unique positive integer hi(n)≥0 with Hi(n)=qhi(n) and we denote by h(n)=∑ihi(n).
Let Gad=G/Z be the adjoint quotient associated to G. Denote by Aad=A/Z and Bad=B/Z. For each s∈C, consider the induced representation on Gad
Let fs∈I(s−1) be the unique spherical vector with fs(1)=1. Then
fs(w˙0n)=H(n)−s,n∈N.
Remark 3.5. Here, the reason we prefer Gad to G is for the Fourier analysis of the zeta function. See Theorem 4.4 (1).
The proof of Proposition 3.4 (1) The highest-weight vector ei+ is contained in the integral weight basis of Λi chosen in Lemma 3.1, say eij0=ei+. The unipotent subgroup N of G fixes this vector. Hence for every n∈N,
Since Hi(n)=maxj∣pij(n)∣, this gives Hi(n)≥∣pij0(n)∣=1. In particular, we can write Hi(n)=qhi(n) for some hi(n)≥0.
Let n0∈N(O). Because Λi is G(O)-stable, πi(n0) preserves Λi. Let M be the matrix of πi(n0) with respect to the O-basis {eij}j. Then the entries of M are contained in O. For any n∈N, we have
In this section, we still assume that G is semisimple and simply connected. We normalize the measure on N such that Vol(N(O))=1.
Definition 4.1 (The Plücker height zeta function). We define the following zeta function for s∈C
ZPl(s)=∫DH(n)−sdn.
Lemma 4.2.The series defining ZPl(s) is absolutely convergent for Re(s)>1. In particular, for any ℓ≥0, the cardinality bℓ is finite and for any s with Re(s)>1
ZPl(s)=ℓ≥0∑bℓq−ℓs.
Proof. By Proposition 3.4(2), for any s∈C and n∈N,
H(n)−s=fs(w˙0n)
with fs the normalized spherical function on the induced representation I(s−1). In particular,
ZPl(s)=∫Dfs(w˙0n)dn.
The absolutely convergence follows from the absolutely convergence of the integral defining the intertwining operator from I(s−1) to I(1−s). This is a special case of [26 Proposition IV.2.1].
As we normalize the measure on N such that Vol(N(O))=1, for any s with Re(s)>1
Before we state the main result for the zeta function, we introduce some notations. Denote by ωi∨∈X∗(A)Q, 1≤i≤m be the fundamental coweights of G, that is, ⟨αi,ωj∨⟩=δij. As we assume that G is semisimple and simply connected, ωi∨∈X∗(Aad). For each z∈Z≥0m, denote by
μz=i∑ziωi∨∈X∗(Aad).
Denote by Gad the Langlands dual group of Gad. In particular, μz can be viewed as a dominant weight of Gad. Let πz the irreducible representation on Gad with the highest weight μz. Denote by
Pz(u)=χ∑m(πz,χ)u⟨ρ,μz+χ⟩.
Here, χ∈X∗(Aad) and m(πz,χ) is the multiplicity of χ in πz.
Lemma 4.3. For each z∈Z≥0m, Pz(u) is a polynomial with nonnegative integral coefficients and degree dz=⟨2ρ,μz⟩.
Proof. We shall prove that for any weight χ∈X∗(Aad) of πz,
0≤⟨ρ,μz+χ⟩≤dz.
In fact, for any weight χ of μz, we have μz−χ=∑i=1mciαi∨ with some ci∈Z≥0. Therefore, ⟨ρ,μz−χ⟩≥0 and ⟨ρ,χ⟩≤⟨ρ,μz⟩. On the other hand, w0χ is also a weight of μz so that
−⟨ρ,χ⟩=⟨ρ,w0χ⟩≤⟨ρ,μz⟩
We obtain that −⟨ρ,μz⟩≤⟨ρ,χ⟩≤⟨ρ,μz⟩ and the Lemma holds.
Theorem 4.4. We have the following.
In the region Re(s)>1, we have
ZPl(s)=(1−q−1)mE(q1−s)z∈Z≥0m∑q−ZPz(q1−s).(2)
Here, the polynomial E(u)=∏γ∈Φ+(1−q−1u⟨ρ,γ∨⟩) and for each z=(zi)i∈Z≥0m, we denote by Z=∑izi.
Write 2ρ=∑iaiαi with ai≥1 and denote by
MG=imaxai.
The right hand side of Equation (4.1) is absolutely convergent for Re(s)>1−MG1. In particular, this gives the holomorphic continuation of ZPl(s) to Re(s)>1−MG1. Moreover, if s=1−MG1+δ for a positive real number δ near to 0, then
Theorem 4.6 (Casselman-Shalika). Let c=(ci)i∈Om with ci=uiϖzi, ui∈O× and zi≥0. Denote by z=(zi)i∈Z≥0m. Then
Js(c)=E(q1−s)Pz(q1−s).
Proof. This is in fact the Casselman-Shalika formula for spherical Whittaker functions. Let Ws(g) be the unique spherical Whittaker function of I(s−1)
Ws(g)=∫Nfs(w˙0ng)ψ0(n)dn.
Then the Casselman-Shalika formula can be stated as following ([13 Theorem 6.5.1])
Ws(tz)=δ1/2(tz)E(q1−s)trπz(As),z∈Z≥0m.
Here,
for each z∈Z≥0m, tz=μz(ϖ)∈Aad and trπz is the trace for πz.
the Satake parameter As∈Aad(C) for I(s−1) is characterized by
Moreover, as ∣u∣<q1/MG, there exists a real number x<1 (depending on ∣u∣) such that q−1∣u∣ai<x for all i. Hence we have
z∈Z≥0m∑q−Z∣Pz(u)∣≤z∈Z≥0m∑(1+Z)LxZ<∞.
In particular, the right hand side of Equation (4.1) is absolutely convergent for any s with Re(s)>1−MG1. Combining with Theorem 4.4(1), this gives the holomorphic continuation of ZPl(s) to Re(s)>1−MG1.
We estimate ZPl(s) near its convergence boundary. Consider
Since bℓ≥0, Landau’s theorem on Dirichlet series with nonnegative coefficients [21 Theorem 1], together with Lemma 4.2 and Theorem 4.4, implies that
ZPl(s)=ℓ≥0∑bℓq−ℓs
with absolute convergence for Re(s)>1−1/MG. For any real s>0, since each ℓ≤R satisfies 1≤qs(R−ℓ), we have
∣Sr∣≤qsRℓ=0∑Rbℓq−ℓs≤qsRZPl(s),s>1−MG1.
By Theorem 4.4 (2), for s=1−MG1+δ with the number δ positive and near to 0, we have
ZPl(s)≪δ−(L+m)
and
∣Sr∣≪q(1−MG1)RqδRδ−(L+m).
We may choose δ=2MG(1+R)1 (so that s<1). Then
∣Sr∣≪(1+R)L+mq(1−MG1)R.
For each ε>0, (1+R)L+m≤CεqεR for some constant Cε. Consider ε=1/(2MG). Then
∣Sr∣≪q(1−2MG1)R.
The proof of Theorem 1.1
In this section, we firstly still assume that G is semisimple and simply connected. Let {Kd}d be the family of principal congruence subgroups of G, that is, the kernel of the reduction map G(O)→G(O/pd). Let {Kld}d be the corresponding Kloosterman sums, that is,
Kld(t)=Kl(t;ψ0,w˙0,Kd),t∈A.
We also use Xd(w˙0t) to denote the Kloosterman set in this case.
For t∈A, write t=t0λt(ϖ) with t0∈A(O) and λt∈X∗(A). By Dąbrowski–Reeder [9 Theorem 0.3], Xd(w˙0t) is nonempty only if λt∈Z≥0Φ∨,+. It therefore suffices to consider
t=t0i=1∏mαi∨(ϖri),t0∈A(O),ri∈Z≥0.
Theorem 5.1 (Theorem 1.3 in [CLYY26]). Let G be a split semisimple and simply connected group. Assume there exist constants ε>0 and C>0 such that for any r∈Z≥0m
∣Sr∣≤Cq(1−ε)R,R=i∑ri.
Then for all sufficiently large d, the Kloosterman sum Kld has a nontrivial bound.
In fact, there is the condition that p∤det(CG) with CG the Cartan matrix of G in [CLYY26, Theorem 1.3]. This condition is from the decomposition of the Kloosterman sum on G to products of Kloosterman sums on GL2 in [4 Proposition 5.3]. We remove this condition in the Appendix (Theorem A.6).
Corollary 5.2. Let G be a split semisimple and simply connected group. Then for all sufficiently large d, the Kloosterman sum Kld has a nontrivial bound
Now, we let G be a general split reductive group over F. As before, we fix a Borel subgroup B=AN of G and ψ=ψ0 is the generic character on N given before [4 Definition 2.1]. Denote by {Gi}1≤i≤s the set of minimal nontrivial normal smooth connected subgroups of Gder. By [14 Theorem in Section 27.5], the groups Gi are simple and the multiplication homomorphism
G1×⋯×Gs⟶Gder
is an isogeny map. These groups Gi are called simple factors of G. We have the following isogeny map
G1×⋯×Gs⟶Gder
where Gi, 1≤i≤s is the universal cover of the simple factor Gi of Gder. For each 1≤i≤s, the group Gi is simple, simply connected group.
An open compact subgroup K⊂G with ψ0∣N∩K=1 is called admissible (with aspect to ψ0) if for any z∈Zder and n∈N with zn∈K, we have z=1 and n∈K, where Zder is the center of Gder ([4 Definition 2.25]).
Proposition 5.3. Let G be a split reductive group. Then there exists an admissible group KG such that the Kloosterman sum Kl(⋅,ψ0−1,w˙0,KG) has a nontrivial bound.
Proof. By Corollary 5.2, for every 1≤i≤s, there exists ϵi>0, 1≤i≤s such that the Kloosterman sum KlGi(ti,ψ0−1Ni,w˙0,i,KGi) has a nontrivial bound. By Lemma [4 Lemma 3.23 (1)], the product of KGi gives an admissible group KGder of Gder. By Lemma [4 Lemma 3.20] and Lemma [4 Lemma 3.23], for a representative w˙0 of the longest Weyl element in W, the Kloosterman sums KlGder(⋅;ψ0−1,w˙0,KGder) has a nontrivial bound. Here, we use the fact that the modular characters are compatible under isogeny maps. By Lemma [4 Lemma 3.22], there exists an admissible group KG of G and a representative w˙0 of the longest Weyl element in W such that the Kloosterman sums KlG(⋅;ψ0−1,w˙0,KG) has a nontrivial bound. By [4 Lemma 3.21 (2)], the above bound holds for any representative of longest Weyl element w0∈W. □
The proof of Theorem 1.1. Now let K⊂G be a compact open subgroup such that ψ0∣N∩K=1 and w0 is the relevant representative of w0 considered [4]. Consider the Kloosterman integral I(t,1w0K). By [4 Lemma 3.19], we have
I(t,1w0K)≐Kl((w0)−2t;ψ0−1,w0,K),t∈A.
Apply [4 Theorem 2.12], we have the following germ expansion of Kloosterman integrals
I(t,1w0K)=J⊂Δ∑(K∅J∗ωJ)(t),t∈A.
where ωJ∈S(AJ) for every J⊂Δ and {K∅J}J⊂J is a system of Shalika germs. By [4 Proposition 2.29], for each J, we can choose the germ function K∅J such that
where σ is the permutation of {1,…,m} defined by −w0αi=ασ(i).
Proposition A.3. We have,
Kld(t)=∣Hd,ℓ∣1x∈Xd(w˙0t)∑SH(x).
Proof. For fixed x, the function t↦ψ(u(t⋅x))ψ(u′(t⋅x)) is constant on cosets of Ad,ℓ:=ker(Ad→Hd,ℓ). Therefore the orbit integral over Ad can be computed by summing over Hd,ℓ; the rest of the proof is identical to that of [4 Proposition 5.3]. The details are omitted. □
We now prove that Hd,ℓ always contains a product subgroup of depth D.
Proposition A.4. If ℓ>D, then UD,ℓm⊆Hd,ℓ.
Proof. Since cCZm⊆im(f), tensoring with OF gives
(pcC)m⊆im(f⊗ZOF).
Multiplying by pd yields
(pD)m⊆(f⊗id)(X∗(A)⊗pd).
Composing with the p-adic exponential map, we have UD,ℓm⊆Hd,ℓ. □
Let V:=UD,ℓm and choose a set of representatives R for the finite quotient Hd,ℓ/V. Then
Here we used min{v(c),−2D−k}≤τ(c) for c∈F, and C:=qm(D+v(cC)+k/2) depends only on F,G,d. Hence,
∣Kld(t)∣≤Cy∈[−ℓ,d]m∑∣Xd(w˙0t)y∣q∑i=1myi/2.
References
(L.C.) Academy for Multidisciplinary Studies, Beijing National Center for Applied Mathematics, Capital Normal University, Beijing, 100048, People’s Republic of China Email address:[email protected]
(J.C.) School of Mathematics, Physics and Finance, Anhui Polytechnic University, Wuhu, Anhui, 241000, People’s Republic of China Email address:[email protected]
(Y.L.) Academy for Multidisciplinary Studies, Beijing National Center for Applied Mathematics, Capital Normal University, Beijing, 100048, People’s Republic of China Email address:[email protected]
(M.Y.) School of Mathematics, Nanjing University, Jiangsu, 210093, People’s Republic of China Email address:[email protected]
References
[1][AB24] E. Assing and V. Blomer. The density conjecture for principal congruence subgroups. Duke Math. J. 173(2024), no. 7, 1359-1426.DOI
[2][BFG88] D. Bump, S. Friedberg, D. Goldfeld, Poincaré series and Kloosterman sums for SL(3, ℤ), Acta Arith. 50 (1988), 31–89.DOI
[3][BM24] V. Blomer and S. Man, Bounds for Kloosterman sums on GL(n). Math. Ann. 390 (2024), 1171-1200.
[4][CCL26] L. Cai, J. Chai and Y. Liu, Bessel distributions and Kloosterman sums, arXiv:2606.29852v2 (2026).arxiv.org/abs/2606.29852
[6][CLPS91] J. Cogdell, J.-S. Li, L. Piateski-Shapiro and P. Sarnak. Poincare series for SO(n, 1). Acta Math. 167(1991), 229-285.DOI
[7][SGA3] M. Demazure, A. Grothendieck, Schémas en groupes I, II, III, Lecture Notes in Math 151, 152, 153, Springer-Verlag, New York (1970).
[8][Dab93] R. Dąbrowski, Kloosterman sums for Chevalley groups, Transactions of the American Mathematical Society, 337 (1993), no. 2, 757–769.DOI
[9][DR98] R. Dąbrowski, M. Reeder, Kloosterman sets in reductive groups. Journal of Number Theory 73 (1998), no.2, 228-255.DOI
[10][FH91] W. Fulton and J. Harris, Representation theory: a first course, Graduate Texts in Mathematics, vol. 129, Springer-Verlag, New York, 1991.DOI
[11][Fri87] S. Friedberg, Poincaré Series for GL(n): Fourier Expansion, Kloosterman Sums, and Algebreo- Geometric Estimates., Math. Z. 196 (1987), 165-188.DOI
[12][GSW21] D. Goldfeld, E. Stade, and M. Woodbury, An orthogonality relation for GL(4, ℝ) (with an appendix by Bingrong Huang). arXiv: 1910.13586. Forum of Mathematics, Sigma (2021), 1-83.
[13][HKP10] T. Haines, R. Kottwitz and A. Prasad. Iwahori-Hecke algebras. J. Ramanujan Math.Soc. 25, no.2 (2010), 113-145.
[14][H80] J. E. Humphreys, Linear Algebraic Groups, Graduate Texts in Mathematics, 021, Springer-Verlag, 1980.
[15][Jan03] J. C. Jantzen Representations of algebraic groups. Mathematical Surveys and Monographs, Vol. 107. American Mathematical Society, Providence, RI, second edition.
[16][JY99] H. Jacquet, Y. Ye, Germs of Kloosterman integrals for GL(3). Transactions American Math. Society, 351, no.3, (1999), 1227-1255.DOI
[17][KP23] T. Kaletha, G. Prasad, Bruhat-Tits Theory–A New Approach. New Math. Monogr., 44 (Cambridge University Press, Cambridge, 2023), xxx+718 pp.DOI
[18][Lin26] J. Linn, Bounds for Kloosterman Sums for GLn, Adv. in Math., vol.499, 2026.DOI
[19][Man22] S. Man, Symplectic Kloosterman sums and Poincare series. Ramanujan J. 57 (2022), 707-753.DOI
[20][Man24] S. Man. Fourier coefficients of Sp(4) Eisenstein series. Acta Arith., 213(3):227–271, 2024.
[21][Mau11] B. N. Maurizi, Extending Landau’s theorem on Dirichlet series with non-negative coefficients. Missouri J. Math. Sci. 23 (2011), no. 2, 105–122.arxiv.org/abs/1009.0228
[22][Mia24] Xinchen Miao. Bessel functions and Kloosterman integrals on GL(n). Journal of Functional Analysis. 286 (2024), 110267 University Press, Cambridge, 2017.arxiv.org/abs/2208.01016
[23][Muj24] C. Mujdai. Kloosterman sums on orthogonal groups. arXiv:2410.20569(2024).
[24][Sal32] H. Salie. Über die Kloostermanschen Summen S(u, v : q). Math. Z. 34(1932), 91-109.
[25][Ste87] G. Stevens, Poincaré series on GL(r) and Kloostermann sums. Math. Annalen 277 (1987), 25-51.DOI
[26][Wal03] J.-L. Waldspurger, La formule de Plancherel pour les groupes p-adiques (d’après Harish-Chandra). J. Inst. Math. Jussieu 2 (2003), 235–333.DOI