ebook img

On Jacquet modules of representations of segment type PDF

0.33 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview On Jacquet modules of representations of segment type

ON JACQUET MODULES OF REPRESENTATIONS OF SEGMENT TYPE IVAN MATIC´ AND MARKO TADIC´ 5 1 0 2 n Abstract. Let Gn denote either the group Sp(n,F) or SO(2n+1,F) over a local non- a archimedeanfieldF. WestudyrepresentationsofsegmenttypeofgroupGn,whichplaya J fundamentalroleintheconstructionsofdiscreteseries,andobtainacompletedescription 6 of the Jacquet modules of these representations. Also, we provide an alternative way for determination of Jacquet modules of strongly positive discrete series and a description of ] T top Jacquet modules of general discrete series. R . h t 1. Introduction a m [ Let F bealocalnon-archimedean fieldofcharacteristic different thantwo. Representations of reductive groups over F that we shall consider in this paper will be always smooth and 1 v admissible. We shall use standard notationfrom the representation theory of general linear 9 groups over F introduced by Bernstein and Zelevinsky (see [17]). Recall that Levi factors 8 2 of maximal parabolic subgroups of general linear groups are direct products of two smaller 1 general linear groups. This fact enables one to consider the representation parabolically 0 induced by the tensor product π ⊗π of two representations of general linear groups, which . 1 2 1 is denoted by 0 5 π1 ×π2. 1 The parabolic induction that we consider in this paper will always be from the parabolic : v subgroups standard with respect to the subgroup of upper triangular matrices (the same i X will be the case for Jacquet modules). The Grothendieck group of the category of finite r length representations of GL(n,F) is denoted by R . The parabolic induction × defines in a n a natural way the structure of a commutative graded algebra with unit on R = ⊕n∈Z≥0Rn. The induced map from R ⊗ R will be denoted by m. (Sums of semi simplifications of) Jacquet modules with respect to maximal parabolic subgroups define mapping m∗: R → R⊗R. This gives R the structure of a graded coalgebra. Moreover, R is a Hopf algebra. Denote ν: GL(n,F) → R×, g 7→ |det(g)| F Date: January 7, 2015. I. Mati´c: Department of Mathematics, University of Osijek, Osijek, Croatia, e-mail: [email protected] M. Tadi´c: Department of Mathematics, University of Zagreb, Zagreb, Croatia, e-mail: [email protected] Mathematics Subject Classification: 22E35 (primary), 22E50, 11F70 (secondary) 1 2 IVAN MATIC´ ANDMARKOTADIC´ where | | denotes thenormalized absolute value. Asegment is aset ofthe form{ρ,νρ,ν2ρ, F ...,νkρ}, where ρ is an irreducible cuspidal representation of a general linear group. We denote this set shortly by [ρ,νkρ]. To such a segment the unique irreducible subrepresen- tation of νkρ×...×ρ is attached, which we denote by δ([ρ,νkρ]). These are the essentially square integrable representations, and one gets all such representations in this way. A very important (and very simple) formula of Bernstein-Zelevinsky theory is k m∗(δ([ρ,νkρ])) = δ([νi+1ρ,νkρ])⊗δ([ρ,νiρ]), i=−1 X which by the transitivity of Jacquet modules, describes all Jacquet modules of irreducible essentially square integrable representations of general linear groups. Onewouldalsolike tohavesuch aformulatodetermine Jacquet modulesofrepresentations of classical groups. It is of particular interest to determine Jacquet modules of classes of representations of classical groups whose role in the admissible dual is as important as the role of essentially square integrable representations in the admissible dual of a general linear group. Besides being interesting in itself, such description would have applications in the theory of automorphic forms and in the classification of unitary duals. In the present paper we are concerned with representations of segment type of symplectic and special odd-orthogonal groups over p-adic field F. This prominent class of representa- tions, consisting ofcertainirreduciblesubquotients ofrepresentations inducedbythetensor product of an essentially square integrable representation of a general linear group and a supercuspidal representation of a classical group, hasbeen introduced by the second author in [14]. Such representations have also appeared as the basic ingredients in classifications of discrete series and tempered representations of classical groups (we refer the reader to [9] and [16]). Representations of segment type can be viewed as irreducible subquotients of generalized principal series induced from representation having a supercuspidal classical- group part. We note that composition series of such representations have been obtained by Mui´c in [10] (in fact, a more general class of generalized principal series, having a strongly positive representation on the classical-group part, has been studied there). In determina- tion of the composition series of induced representation, the fundamental role is played by Jacquet modules of the initial representation. Thus, our results provide a starting point for investigation of representations induced by those of segment type. We emphasize that in several cases our results provide complete description of Jacquet modules of certain non-tempered representations. Our description can be used to analyze asymptotics of matrix coefficients of such representations and, consequently, to determine some prominent members of the unitary dual. In the case of generic reducibilities, representations of segment type are always tempered or discrete series representations. However, for general reducibilities, representations of segment type might also be non-tempered. The structural formula, which is a version of ON JACQUET MODULES OF REPRESENTATIONS OF SEGMENT TYPE 3 the Geometrical Lemma of Bernstein-Zelevinsky, together with certain properties of the representations of segment type obtained in [14], enables us to use an inductive procedure which results in a complete description of Jacquet modules of GL-type and top Jacquet modules of such representations. These results, enhanced by the transitivity of Jacquet modules and some results regarding Jacquet modules of representations of general linear groups, allow us to determine Jacquet modules of representations of segment type with respect to all standard maximal parabolic subgroups. Since representations of segment type can appear in three technically different composition series, we obtain a description of their Jacquet modules considering three technically different cases (but the general strategy in all the cases is the same). However, we introduce a convention regarding irreducible constituents of considered composition series, which enables us to state our results uniformly. An analogousproblem to determine Jacquet modules has been studied for strongly positive representations by the first author ([6]), but it was mostly based on the fact that the Jacquet module of strongly positive discrete series has a representation of the same type on its classical-group part. On the other hand, an approach similar to the one presented here has recently been used by the first author to provide a description of Jacquet modules with respect to maximal parabolic subgroups of certain families of discrete series which contain an irreducible es- sentially square integrable representation on the GL-part ([7]). In that paper one starts with determination from Jacquet modules which are not of GL-type. Then, to deduce to which irreducible subquotient obtained Jacquet modules belong, one uses transitivity of Jacquet modules and representation theory of general linear groups. Let us now describe the contents of the paper in more details. In the following section we introducesomenotationwhich will beusedthroughoutthepaper, while inthethirdsection we recall some important properties of representations of segment type and introduce a certain convention which will keep our results uniform. The next three sections aredevoted to determination of Jacquet modules of the representations of segment type, considering three technically different cases. Also, some results obtained in section four and five help us to shorten the proofs in the sixth section. In the last two sections we derive some interesting Jacquet modules of discrete series. Firstly we provide an alternative way to determine Jacquet modules of strongly positive discrete series and secondly we provide a description of top Jacquet modules of general discrete series. For the convenience of the reader, we cite the main description of Jacquet modules here. Representations of segment type are irreducible subqutients of δ([ν−cρ,νdρ])⋊σ, where ρ is an irreducible cuspidal representations of a general linear group and σ is an irreducible cuspidal representations of a classical group (c,d ∈ R, c + d ∈ Z ). Then directly from ≥0 4 IVAN MATIC´ ANDMARKOTADIC´ previously mentioned formula for m∗ and [12] we get d d µ∗ δ([ν−cρ,νdρ])⋊σ = δ([ν−iρ˜,νcρ˜])×δ([νj+1ρ,νdρ])⊗δ([νi+1ρ,νjρ])⋊σ i=−c−1 j=i (cid:0) (cid:1) X X (ρ˜ denotes the contragredient of ρ). The case which interest us is when δ([ν−cρ,νdρ])⋊σ reduces (square integrable subquotients can show up in this case only). Then we can take selfcontragredient ρ and assume d ∈ (1/2)Z (only in this case we can have reducibility). We shall consider the case d−c ≥ 0 (changing signs of c and d simultaneously gives the same composition series). We shall assume that there exists α ∈ (1/2)Z such that for ≥0 β ≥ 0, νβρ⋊σ reduces if and only if β = α. This always holds for ρ selfcontragredient (it is a very non-trivial fact which we shall not discuss here; we shall simply assume that it holds for ρ and σ). Also, we assume d−α ∈ Z (only then we can have reducibility). The length of δ([ν−cρ,νdρ])⋊σ is at most three. This is a multiplicity one representation. It is reducible if and only if [−c,d] ∩ {−α,α} 6= ∅. It has length three if and only if {−α,α} ⊆ [−c,d] and c 6= d. Below we shall define terms δ([ν−cρ,νdρ] ;σ), δ([ν−cρ,νdρ] ;σ) and L (δ([ν−cρ,νdρ]);σ). + − α Each of them is either irreducible representation or zero. They satisfy (1.1) δ([ν−cρ,νdρ])⋊σ = δ([ν−cρ,νdρ] ;σ)+δ([ν−cρ,νdρ] ;σ)+L (δ([ν−cρ,νdρ]);σ) + − α in the corresponding Grothendieck group. Suppose first that δ([ν−cρ,νdρ]) ⋊ σ is irreducible. Then we take δ([ν−cρ,νdρ] ;σ) = 0. − Furthermore, in this case we require δ([ν−cρ,νdρ] ;σ) 6= 0 if and only if [−c,d] ⊆ [−α + + 1,α−1]. For irreducible δ([ν−cρ,νdρ])⋊σ, this requirement and (1.1) obviously determine L (δ([ν−cρ,νdρ]);σ). α Suppose now that δ([ν−cρ,νdρ]) ⋊ σ reduces. If c = d, let L (δ([ν−cρ,νdρ]);σ) = 0. α Otherwise, L (δ([ν−cρ,νdρ]);σ) will denote the Langlands quotient L(δ([ν−cρ,νdρ]);σ) α of δ([ν−cρ,νdρ]) ⋊ σ. If α > 0, then there is the unique irreducible subquotient of δ([ν−cρ,νdρ]) ⋊ σ which has in its minimal standard Jacquet module at least one irre- ducible subquotient whose all exponents are non-negative (for more details, we refer the reader to Sections 2 and 3). We denote such irreducible subquotient of δ([ν−cρ,νdρ])⋊σ by δ([ν−cρ,νdρ] ;σ). If α = 0, we write ρ⋊σ as a sum of irreducible subrepresentations + τ ⊕ τ (we fix the choice of signs ±, which is arbitrary and can be compatible with 1 −1 the one from [8], but this is not essential for our paper). Then there exists the unique irreducible subquotient of δ([ν−cρ,νdρ])⋊σ that contains an irreducible representation of the form π ⊗ τ in Jacquet module with respect to appropriate standard parabolic sub- 1 group, and we denote it by δ([ν−cρ,νdρ] ;σ). If c = d or the length of δ([ν−cρ,νdρ])⋊σ is + three, then this induced representation contains the unique irreducible subrepresentation different from δ([ν−cρ,νdρ] ;σ) and we denote it by δ([ν−cρ,νdρ] ;σ). Otherwise, we take + − ON JACQUET MODULES OF REPRESENTATIONS OF SEGMENT TYPE 5 δ([ν−cρ,νdρ] ;σ) = 0. We note that the representation δ([ν−cρ,νdρ] ;σ) is square inte- − + grable if and only if c 6= d, {−α,α} ⊆ [−c,d] or α = −c. If δ([ν−cρ,νdρ] ;σ) is square + integrable, then δ([ν−cρ,νdρ] ;σ) is also square integrable, if it is non-zero. Furthermore, − if δ([ν−cρ,νdρ] ;σ) is square integrable, then δ([ν−cρ,νdρ] ;σ) is square integrable. − + We have the following equality: µ∗ δ([ν−cρ,νdρ] ;σ) ± c d (cid:0) (cid:1) = δ([ν−iρ,νcρ])×δ([νj+1ρ,νdρ])⊗δ([νi+1ρ,νjρ] ;σ)+ ± i=−c−1j=i+1 X X + δ([ν−iρ,νcρ])×δ([νj+1ρ,νdρ])⊗L (δ([νi+1ρ,νjρ]);σ)+ α −c−1≤i≤c i+1≤j≤c X X i+j<−1 ±α−1 + δ([ν−iρ,νcρ])×δ([νi+1ρ,νdρ])⊗σ. i=−c−1 X For c < α or α ≤ c < d, we have µ∗ L(δ([ν−cρ,νdρ]);σ) (cid:0) = L(δ((cid:1)[ν−iρ,νcρ]),δ([νj+1ρ,νdρ]) ⊗L (δ([νi+1ρ,νjρ]);σ))+ α −c−1≤i≤d i+1≤j≤d X X (cid:1) 0≤i+j d + L(δ([ν−iρ,νcρ]),δ([νi+1ρ,νdρ]))⊗σ. i=α X The authorswould like tothank thereferee forreading thepaper very carefully andhelping us to improve the presentation style. Also, the author’s thanks go to Sˇime Ungar for many useful suggestions and help with English language. This work has been supported by Croatian Science Foundation under the project 9364. 2. Notation We will first describe the groups that we consider. Let J = (δ ) denote the n×n matrix, where δ stands for the Kronecker n i,n+1−j 1≤i,j≤n i,n+1−j symbol. For a square matrix g, we denote by gt its transposed matrix, and by gτ its transposed matrix with respect to the second diagonal. In what follows, we shall fix one of the series of classical groups 0 −J 0 −J Sp(n,F) = g ∈ GL(2n,F) : n gt n = g−1 , J 0 J 0 n n (cid:26) (cid:18) (cid:19) (cid:18) (cid:19) (cid:27) 6 IVAN MATIC´ ANDMARKOTADIC´ or SO(2n+1,F) = g ∈ GL(2n+1,F) : gτ = g−1 (cid:26) (cid:27) and denote by G the rank n group belonging to the series which we fixed. n The set of standard parabolic subgroups will be fixed in a usual way, i.e., in G we fix n the minimal F-parabolic subgroup consisting of upper-triangular matrices in G . Then n ∼ the Levi factors of standard parabolic subgroups have the form M = GL(n ,F) × ··· × 1 GL(n ,F)×G . For representations δ of GL(n ,F), i = 1,2,...,k, and a representation k n′ i i σ of G , the normalized parabolically induced representation IndGn(δ ⊗···⊗δ ⊗σ) will n′ M 1 k be denoted by δ ×···×δ ⋊σ. 1 k Let R(G )denote theGrothendieck groupofthecategoryoffinitelength representations of n G and define R(G) = ⊕ R(G ). Similarly as in the case of a general linear group, sums n n≥0 n of semisimplifications of Jacquet modules with respect to maximal parabolic subgroups define the mapping µ∗: R(G) → R⊗R(G). Throughout the paper, the Jacquet module with respect to the smallest standard parabolic subgroup(s) admitting non-zero Jacquet modules for the representation in question will be called the minimal standard Jacquet module. For representation π ∈ R(G ) with n partial cuspidal support σ ∈ R(G ), the Jacquet module of π with respect to the maximal n′ parabolic subgroup having Levi factor equal to GL(n − n′,F) × G will be called the n′ Jacquet module of GL-type and will be denoted by s (π). The sum of all irreducible GL constituents (counted with multiplicities) of µ∗(π) of the form τ ⊗ϕ, where τ is cuspidal, will be denoted by s (π). top We define κ: R ⊗ R → R ⊗ R by κ(x ⊗ y) = y ⊗ x and extend contragredient to an automorphism of R in the natural way. Let M∗: R → R be defined by M∗ = (m⊗id)◦( ⊗m∗)◦κ◦m∗. e We recall the following formulas which hold foer ρ not necessary self-dual: b b M∗(δ([νaρ,νbρ])) = δ([ν−iρ˜,ν−aρ˜])×δ([νj+1ρ,νbρ])⊗δ([νi+1ρ,νjρ]) i=a−1 j=i X X or b−a+1 b−k M∗ δ([νaρ,νbρ]) = δ([ν−iρ˜,ν−aρ˜])×δ([νk+i+1ρ,νbρ])⊗δ([νi+1ρ,νi+kρ]). k=0 i=a−1 (cid:0) (cid:1) X X The following lemma, which has been derived in [12], presents a crucial structural formula for our calculations with Jacquet modules. ON JACQUET MODULES OF REPRESENTATIONS OF SEGMENT TYPE 7 Lemma 2.1. Let ρ be an irreducible cuspidal representation of GL(m,F) and a,b ∈ R such that b−a ∈ Z . For σ ∈ R(G ) we write µ∗(σ) = τ ⊗σ′. Then the following ≥0 n τ,σ′ equalities hold: P µ∗(π ⋊σ) = M∗(π)⋊µ∗(σ) and b b µ∗(δ([νaρ,νbρ])⋊σ) = δ([ν−iρ˜,ν−aρ˜])×δ([νj+1ρ,νbρ])×τ ⊗ i=a−1 j=i τ,σ′ X XX ⊗δ([νi+1ρ,νjρ])⋊σ′. We omit δ([νxρ,νyρ]) if x > y. We briefly recall the subrepresentation version of Langlands classification for general linear groups, which is necessary for determination of Jacquet modules of GL-type. Forevery irreducible essentially square integrablerepresentation δ ofGL(n,F), thereexists an e(δ) ∈ R such that ν−e(δ)δ is unitarizable. Suppose that δ ,δ ,...,δ are irreducible, 1 2 k essentially square integrable representations of GL(n ,F),GL(n ,F),..., GL(n ,F) with 1 2 k e(δ ) ≤ e(δ ) ≤ ... ≤ e(δ ). Then the induced representation δ × δ × ··· × δ has 1 2 k 1 2 k a unique irreducible subrepresentation, which we denote by L(δ ,δ ,...,δ ). This irre- 1 2 k ducible subrepresentation is called the Langlands subrepresentation, and it appears with the multiplicity one in δ ×δ ×...×δ . Every irreducible representation π of GL(n,F) 1 2 k is isomorphic to some L(δ ,δ ,...,δ ). For a given π, the representations δ ,δ ,...,δ are 1 2 k 1 2 k unique up to a permutation. Also, throughoutthepaperweusetheLanglandsclassificationforclassicalgroupsandwrite a non-tempered irreducible representation π of G as the unique irreducible (Langlands) n quotient oftheinducedrepresentationoftheformδ ×δ ×···×δ ⋊τ, whereτ isatempered 1 2 k representation of G , and δ ,δ ,...,δ are irreducible, essentially square integrable repre- t 1 2 k sentations of GL(n ,F),GL(n ,F),..., GL(n ,F) with e(δ ) ≥ e(δ ) ≥ ... ≥ e(δ ) > 0. 1 2 k 1 2 k In this case, we write π = L(δ ,δ ,...,δ ;τ). Again, for a given π, the representations 1 2 k δ ,δ ,...,δ are unique up to a permutation. 1 2 k Since the class of representations which we will study contains certain discrete series rep- resentations, we shortly recall basic ingredients of the classification of discrete series for classical groups due to Mœglin and second author ([8, 9]). According to this classification, discrete series are in bijective correspondence with admis- sible Jordan triples. More precisely, discrete series σ of G corresponds to the triple of n the form (Jord,σ′,ǫ), where σ′ is the partial cuspidal support of σ, Jord is the finite set (possibly empty) of pairs (c,ρ), where ρ is an irreducible cuspidal self-dual representation of GL(n ,F), and c > 0 an integer of appropriate parity, while ǫ is a function defined on ρ a subset of Jord∪(Jord×Jord) and attains the values 1 and −1. 8 IVAN MATIC´ ANDMARKOTADIC´ For an irreducible cuspidal self-dual representation ρ of GL(n ,F) we write Jord = {c : ρ ρ (c,ρ) ∈ Jord}. If Jord 6= ∅ and c ∈ Jord , we put c = max{d ∈ Jord : d < c}, if it ρ ρ ρ exists. Now, by definition ǫ((c ,ρ),(c,ρ)) = 1 if there is some irreducible representation ϕ such that σ is a subrepresentation of δ([ν−c−1ρ,νc−1ρ])⋊ϕ. 2 2 In the classification mentioned above, every discrete series of G is obtained inductively, n starting from a Jordan triple of alternated type which corresponds to a strongly positive representation, i.e., to the one whose all exponents in the supports of GL-type Jacquet module are positive. In each step one adds a pair of consecutive elements to Jordan block and the expanded ǫ-function equals one on that pair. For more details about this classification we refer the reader to [15] and [16]. 3. Representations of segment type Definition 3.1. Let ρ and σ be irreducible cuspidal representations of a general linear group and of a classical group, respectively. Let a,b ∈ R, b−a ∈ Z , be such that ≥0 0 ≤ a+b and δ([νaρ,νbρ])⋊σ reduces. Then any irreducible subquotient of the above representation which contains δ([νaρ,νbρ]) ⊗ σ in its Jacquet module (with respect to the standard parabolic subgroup), will be called a representation of segment type. We emphasize that representations δ([νaρ,νbρ])⋊σ and δ([ν−bρ,ν−aρ])⋊σ share the same composition series, but the choice 0 ≤ a+b enables us to obtain, using the known formulas for Jacquet modules of GL-type, that the representation of segment type is always a subrepresentation of δ([νaρ,νbρ])⋊σ. Reducibility of the induced representation δ([νaρ,νbρ])⋊σ implies that ρ is self-dual. Also, let α ∈ R be such that the induced representation ναρ⋊σ reduces. Given ρ and σ such ≥0 α is unique, by the results of Silberger [11]. Furthermore, recent results of Arthur imply that α ∈ (1/2)Z (for more details we refer the reader to [1]). Notethatifδ([νaρ,νbρ])⋊σ isirreducible, thenitsJacquetmodulecontainsδ([νaρ,νbρ])⊗σ and in this case we have a complete description of Jacquet modules with respect to the maximal parabolic subgroups of this representation. Furthermore, it has been proved in [13] that δ([νaρ,νbρ])⋊σ reduces if and only if [νaρ,νbρ]∩{ν−αρ,ναρ} =6 ∅. Thus, in the sequel we shall assume that a,b ∈ (1/2)Z. ON JACQUET MODULES OF REPRESENTATIONS OF SEGMENT TYPE 9 Also, we introduce the notion of proper Langlands quotient of the induced representation d: L(d), if the corresponding standard module reduces; L (d) = proper 0, if the corresponding standard module is irreducible. ( Definition 3.2. In the case of reducibility, we define δ([νaρ,νbρ] ;σ) to be any irreducible + subquotient of δ([νaρ,νbρ])⋊σ which has in its minimal standard Jacquet module at least one irreducible subquotient whose all exponents are non-negative. In the sequel, we take δ([νaρ,νbρ] ;σ) = 0 if δ([νaρ,νbρ])⋊σ is a length two representation − and if −a 6= b. In general case, the uniqueness of irreducible subquotients in Definition 3.2, is provided by the following lemma. Lemma 3.3. Suppose that α > 0. There exists a unique irreducible subquotient of the induced representation δ([νaρ,νbρ])⋊σ whose minimal standard Jacquet module contains at least one irreducible subquotient with all exponents being non-negative. Proof. The claim obviously holds if the induced representation δ([νaρ,νbρ]) ⋊ σ is irre- ducible. Thus, we may assume {−α,α}∩[a,b] 6= ∅. One can see directly that b+1 s (δ([νaρ,νbρ])⋊σ) = δ([ν−i+1ρ,ν−aρ])×δ([νiρ,νbρ])⊗σ. GL i=a X It follows immediately that if a > 0 then there is a unique irreducible subquotient of s (δ([νaρ,νbρ]) ⋊ σ) with all exponents being non-negative, and we obtain such a sub- GL quotient for i = a. Similarly, if a < 0 and a− 1 ∈ Z, we deduce that the unique irreducible 2 subquotient ofs (δ([νaρ,νbρ])⋊σ) having allexponents non-negativeisobtainedfori = 1 GL 2 (note that in this case the representation δ([ν1ρ,ν−aρ])×δ([ν1ρ,νbρ]) is irreducible). 2 2 Thus, it remains to prove the lemma for a ≤ 0, a ∈ Z. Obviously, α ≤ b. If −a = b, the only irreducible subquotients of s (δ([νaρ,νbρ])⋊σ) having all exponents GL non-negative is δ([νρ,νbρ])×δ([ρ,νbρ])⊗σ (which appears with multiplicity two). On the other hand, δ([ν−bρ,νbρ])⋊σ is a representation of the length two whose composi- tion series consists of two non-isomorphic tempered representations. Using Lemma 4.1 of [16], we deduce that there is a unique irreducible subquotient of δ([ν−bρ,νbρ])⋊σ having δ([ναρ,νbρ]) × δ([ναρ,νbρ]) ⊗ δ([ν−α+1ρ,να−1ρ]) ⋊ σ in its Jacquet module. Let us de- note such subquotient by τ . Since δ([ν−α+1ρ,να−1ρ])⋊σ is irreducible, δ([ναρ,νbρ])× temp δ([ναρ,νbρ]) ⊗ δ([νρ,να−1ρ]) × δ([ρ,να−1ρ]) ⊗ σ appears in the Jacquet module of τ temp with multiplicity two. Transitivity of Jacquet modules implies that there is some irre- ducible representation ϕ such that ϕ ⊗ σ appears in the Jacquet module of τ and temp 10 IVAN MATIC´ ANDMARKOTADIC´ m∗(ϕ) ≥ δ([ναρ,νbρ]) × δ([ναρ,νbρ]) ⊗ δ([νρ,να−1ρ]) × δ([ρ,να−1ρ]). From cuspidal sup- port of ϕ and structural formula for µ∗ it follows easily that ϕ = δ([νρ,νbρ])×δ([ρ,νbρ]). Since δ([ναρ,νbρ])×δ([ναρ,νbρ])⊗δ([νρ,να−1ρ])×δ([ρ,να−1ρ]) appearsinm∗(δ([νρ,νbρ])× δ([ρ,νbρ])) with multiplicity one, it follows that the Jacquet module of τ contains both temp copies of δ([νρ,νbρ])×δ([ρ,νbρ])⊗σ. We now turn to the case −a < b. In this case, irreducible subquotients of s (δ([νaρ,νbρ]) ⋊ σ) having all exponents non- GL negative are δ([νρ,ν−aρ]) × δ([ρ,νbρ]) ⊗ σ (which appears with multiplicity two) and L(δ([ρ,νbρ]),δ([νρ,νbρ]))⊗σ (which appears with multiplicity one). Several possibilities, depending on a, will be considered separately. Let us first assume a ≤ −α. By Theorem 2.1 of [10], δ([νaρ,νbρ]) ⋊ σ is a length three representationandwedenotebyπ itsdiscreteseriessubrepresentationwhosecorresponding ǫ-function ǫ satisfies ǫ ((−2a + 1,ρ),(2b + 1,ρ)) = ǫ (((−2a + 1) ,ρ),(2a + 1,ρ)) = 1. π π π If we denote by π′ a discrete series subrepresentation of δ([νaρ,νbρ]) ⋊ σ different than π, it easily follows that its ǫ-function ǫ satisfies ǫ ((−2a + 1,ρ),(2b + 1,ρ)) = 1 and π′ π′ ǫ (((−2a+ 1) ,ρ),(2a+1,ρ)) = −1. Using Lemma 4.1 of [7] and transitivity of Jacquet π modules, we obtain that δ([νρ,ν−aρ]) × δ([ρ,νbρ]) ⊗ σ is not contained in the Jacquet module of the Langlands quotient of δ([νaρ,νbρ])⋊σ. Furthermore, using Proposition 7.2 of [16] and transitivity of Jacquet modules, we obtain that δ([νρ,ν−aρ]) × δ([ρ,νbρ]) ⊗σ must be in the Jacquet module of π and it is not contained in the Jacquet module of π′. Condition ǫ (((−2a+1) ,ρ),(2a+1,ρ)) = 1 implies that there issome irreducible represen- π tation ϕ such that π is a subrepresentation of δ([ρ,ν−aρ])⋊ϕ. Using Frobenius reciprocity and formula for µ∗, we deduce that ϕ is an irreducible subquotient of δ([νρ,νbρ]) ⋊ σ. Proposition 3.1(i) of [10] implies δ([νρ,νbρ])⋊σ = L (δ([νρ,νbρ]);σ)+L (δ([νρ,να−1ρ]);σ′), proper proper whereσ′ istheunique stronglypositive discreteseries subrepresentation ofδ([ναρ,νbρ])⋊σ. Since π is a square integrable representation, ϕ 6= L (δ([νρ,νbρ]);σ). Consequently, ϕ proper equals L (δ([νρ,να−1ρ]);σ′) and it follows immediately that it is a subrepresentation proper of δ([νρ,νbρ])⋊σ. Thus, Jacquet module of π contains δ([ρ,ν−aρ])⊗δ([νρ,νbρ])⊗σ. Since Jacquet module of δ([νρ,ν−aρ]) × δ([ρ,νbρ]) does not contain δ([ρ,ν−aρ]) ⊗ δ([νρ,νbρ]), transitivity of Jacquet modules implies that L(δ([ρ,ν−aρ]),δ([νρ,νbρ]))⊗σ ≤ µ∗(π). Let us now assume −α+1 < a. In this case, by Theorem 4.1(ii) of [10], δ([νaρ,νbρ])⋊σ is a length two representation and we have δ([νaρ,νbρ])⋊σ = L (δ([νaρ,νbρ]);σ)+L (δ([νaρ,να−1ρ]);σ′), proper proper for the unique strongly positive discrete series subrepresentation σ′ of δ([ναρ,νbρ])⋊σ. It can be directly verified that Jacquet module of δ([νaρ,να−1ρ])⋊σ′ contains δ([νρ,ν−aρ])× δ([ρ,νbρ])⊗σ with multiplicity two and L(δ([ρ,ν−aρ]), δ([νρ,νbρ]))⊗σ with multiplicity

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.