Bernstein-Zelevinsky Theory in p-adic Representations

• Bernstein-Zelevinsky theory is a framework that classifies irreducible smooth representations of GL(n, F) using segments and multisegments.
• It employs derivative functors and Jacquet modules to bridge parabolic induction with Whittaker models and Bernstein components.
• The theory extends to include Hecke algebra formulations, geometric and automorphic analogues, and duality in both p-adic and Archimedean settings.

Bernstein-Zelevinsky theory is the framework in which irreducible smooth representations of $GL_n(F)$, for a non-Archimedean local field $F$, are described in terms of segments and multisegments, while derivative functors, Jacquet modules, Whittaker models, Bernstein components, and the Bernstein center organize the relation between local structure and parabolic induction. In the original Bernstein-Zelevinsky theory, for representations of $GL_n$ over non-archimedean local fields, derivatives are functors constructed via restriction and (twisted) Jacquet functors relating $\mathrm{Rep}(GL_n)$ and $\mathrm{Rep}(GL_{n-k})$; they are crucial for classification and for understanding Whittaker models of induced representations (Zhang, 2022). In current usage, the theory also includes Hecke-algebra realizations of derivatives, geometric and automorphic analogues, Archimedean filtrations, and several duality formalisms (Chan et al., 2017).

1. Local type-$A$ foundations

Let $F$ be a non-Archimedean local field, $G_n=\mathrm{GL}_n(F)$, and all representations smooth and over $\mathbb{C}$. A segment is of the form

$$[a,b]_{\rho}=\{\nu^a\rho,\nu^{a+1}\rho,\dots,\nu^b\rho\},$$

where $F$0 is an irreducible cuspidal representation and $F$1. A multisegment is a finite multiset of such segments. Zelevinsky showed that the irreducible smooth representations of $F$2 are parametrized by multisegments (Chan, 2 Jan 2026).

A complementary formulation, used in work on families, starts from a segment

$F$3

where $F$4 is a supercuspidal representation of $F$5. For such a segment, parabolic induction gives a representation of $F$6; the unique irreducible quotient $F$7 is the fundamental building block. For multisets of segments $F$8, suitably ordered, the unique irreducible quotient $F$9 yields all irreducible admissible representations of $GL_n$0 (Mundy, 2023).

Derivatives enter at several levels. The classical Bernstein-Zelevinsky derivative $GL_n$1 is a functorial construction involving a twisted Jacquet functor. A second operation, the $GL_n$2-derivative $GL_n$3, uses essentially square-integrable representations attached to segments. Given $GL_n$4 and a segment $GL_n$5, if there exists an irreducible $GL_n$6 such that

$GL_n$7

then $GL_n$8; otherwise $GL_n$9. For a multisegment $\mathrm{Rep}(GL_n)$0 in ascending order,

$\mathrm{Rep}(GL_n)$1

and this construction produces simple quotients of Bernstein-Zelevinsky derivatives (Chan, 2021).

The central representation-theoretic point is that the derivative formalism does not merely record highest Whittaker data. It also supplies a systematic way to pass between parabolically induced objects, generalized Steinberg representations, and simple quotients extracted from Jacquet modules. This suggests that the local classification and the calculus of derivatives are not separable parts of the theory but two presentations of the same structure.

2. Bernstein components, blocks, and the center

For a reductive $\mathrm{Rep}(GL_n)$2-adic group $\mathrm{Rep}(GL_n)$3, the irreducible smooth complex representations $\mathrm{Rep}(GL_n)$4 admit a Bernstein decomposition

$\mathrm{Rep}(GL_n)$5

where each Bernstein component $\mathrm{Rep}(GL_n)$6 corresponds, up to equivalence, to pairs $\mathrm{Rep}(GL_n)$7 with $\mathrm{Rep}(GL_n)$8 a Levi subgroup of $\mathrm{Rep}(GL_n)$9 and $\mathrm{Rep}(GL_{n-k})$0 a cuspidal representation of $\mathrm{Rep}(GL_{n-k})$1. The Bernstein center $\mathrm{Rep}(GL_{n-k})$2 is both the center of the category of smooth $\mathrm{Rep}(GL_{n-k})$3-modules, or the center of the Hecke algebra $\mathrm{Rep}(GL_{n-k})$4, and the algebra of invariant distributions $\mathrm{Rep}(GL_{n-k})$5 on $\mathrm{Rep}(GL_{n-k})$6 satisfying compatibility with convolution. By Schur-Quillen, each $\mathrm{Rep}(GL_{n-k})$7 acts on irreducibles via a scalar and hence defines a function

$\mathrm{Rep}(GL_{n-k})$8

These basic formulations are part of the Bernstein-Zelevinsky organization of the smooth dual (Braverman et al., 2015).

A precise relation between geometric support and spectral behavior is given by the theorem

$\mathrm{Rep}(GL_{n-k})$9

where $A$0 is the subspace of invariant distributions supported on compact elements of $A$1, and $A$2 consists of those $A$3 for which $A$4 is constant on every Bernstein component. Thus, an element of the Bernstein center is supported on compact elements if and only if it acts as a scalar on each Bernstein component of $A$5. The same result implies that $A$6 is a subalgebra of $A$7, even though the set of compact elements in $A$8 is not closed under multiplication (Braverman et al., 2015).

This identification is closely tied to central idempotents. The central idempotents corresponding to the projection onto a Bernstein component are supported on compact elements. Functions in $A$9 that are constant on components are thus finite linear combinations of these idempotents, so their preimages on the distribution side are supported on compact elements (Braverman et al., 2015).

An integral and modular analogue exists for $F$0. The category $F$1 of smooth $F$2-modules decomposes into blocks indexed by mod $F$3 inertial supercuspidal support,

$F$4

and the center of each block is a reduced, finite type, $F$5-torsion free $F$6-algebra. Moreover, the $F$7-points of the center of each block are in bijection with the possible supercuspidal supports of the smooth $F$8-modules that lie in the block (Helm, 2012).

On the Langlands side for classical groups, one can construct the cuspidal support of an enhanced Langlands parameter and obtain a decomposition of the set of enhanced Langlands parameters à la Bernstein; these constructions match under the local Langlands correspondence and yield compatibility of the correspondence with parabolic induction (Moussaoui, 2015).

3. Hecke-algebra formulations and branching

A major reformulation of Bernstein-Zelevinsky theory passes through Hecke algebras attached to Bushnell-Kutzko types. Bernstein components of the category of smooth representations of $F$9 are described by Hecke algebras arising from types, and this allows the derivative formalism to be translated into explicit module theory (Chan et al., 2017).

At Iwahori level, the key object is the Gelfand-Graev representation $G_n=\mathrm{GL}_n(F)$0, where $G_n=\mathrm{GL}_n(F)$1 is the maximal unipotent subgroup of a Borel and $G_n=\mathrm{GL}_n(F)$2 is a Whittaker character. Its space of Iwahori-fixed vectors is explicitly

$G_n=\mathrm{GL}_n(F)$3

where $G_n=\mathrm{GL}_n(F)$4 is the Iwahori-Hecke algebra, $G_n=\mathrm{GL}_n(F)$5 its finite Hecke subalgebra, and $G_n=\mathrm{GL}_n(F)$6 the sign character (Chan et al., 2016). This description makes genericity visible as a sign-isotypic condition in Hecke modules.

For an $G_n=\mathrm{GL}_n(F)$7-module $G_n=\mathrm{GL}_n(F)$8, the Hecke-algebra Bernstein-Zelevinsky derivative is defined by a sign projector: $G_n=\mathrm{GL}_n(F)$9 If $\mathbb{C}$0 is a smooth $\mathbb{C}$1-representation generated by its Iwahori fixed vectors, then

$\mathbb{C}$2

as $\mathbb{C}$3-modules (Chan et al., 2016). In this way, the classical derivative becomes a projector calculation inside a type-$\mathbb{C}$4 Hecke algebra.

Lusztig’s reductions then move the problem to graded Hecke algebras. The compatibility

$\mathbb{C}$5

allows derivative computations to be made in the graded setting (Chan et al., 2016). One concrete outcome is the computation of the Bernstein-Zelevinsky derivatives of generalized Speh modules: the $\mathbb{C}$6-th derivative is the direct sum of generalized Speh modules corresponding to all partitions obtained by removing $\mathbb{C}$7 boxes, at most one per row, such that the remaining diagram is still a valid Young diagram (Chan et al., 2016).

The same framework has branching consequences. For certain generic representations of $\mathbb{C}$8 restricted to $\mathbb{C}$9, the Iwahori-Hecke algebra action can be realized explicitly; in locally nice situations, the completed Hecke module is projective, and this is used to verify a conjecture on an Ext-branching problem of D. Prasad for a class of examples (Chan et al., 2016). In exceptional rank, explicit computations of the Aubert-Zelevinsky involution for principal and mediate series of $$[a,b]_{\rho}=\{\nu^a\rho,\nu^{a+1}\rho,\dots,\nu^b\rho\},$$0 translate to corresponding involutions on associated affine Hecke algebras and confirm several instances of the Bernstein conjecture for $$[a,b]_{\rho}=\{\nu^a\rho,\nu^{a+1}\rho,\dots,\nu^b\rho\},$$1 (Qin, 23 May 2025).

4. Multisegment combinatorics and simple quotients of derivatives

Recent work develops a refined combinatorics for the simple quotients produced by derivatives. For $$[a,b]_{\rho}=\{\nu^a\rho,\nu^{a+1}\rho,\dots,\nu^b\rho\},$$2 and irreducible $$[a,b]_{\rho}=\{\nu^a\rho,\nu^{a+1}\rho,\dots,\nu^b\rho\},$$3, one considers

$$[a,b]_{\rho}=\{\nu^a\rho,\nu^{a+1}\rho,\dots,\nu^b\rho\},$$4

where $$[a,b]_{\rho}=\{\nu^a\rho,\nu^{a+1}\rho,\dots,\nu^b\rho\},$$5 runs over all multisegments (Chan, 2 Jan 2026). The relevant partial order is the Zelevinsky ordering $$[a,b]_{\rho}=\{\nu^a\rho,\nu^{a+1}\rho,\dots,\nu^b\rho\},$$6, generated by elementary intersection-union operations on linked segments: if $$[a,b]_{\rho}=\{\nu^a\rho,\nu^{a+1}\rho,\dots,\nu^b\rho\},$$7 are linked, one replaces them by $$[a,b]_{\rho}=\{\nu^a\rho,\nu^{a+1}\rho,\dots,\nu^b\rho\},$$8 and $$[a,b]_{\rho}=\{\nu^a\rho,\nu^{a+1}\rho,\dots,\nu^b\rho\},$$9 (Chan, 2 Jan 2026).

The set $F$00 has strong order-theoretic structure. If $F$01 is nonempty, it contains a unique minimal element with respect to $F$02, and the set is convex in the Zelevinsky order: if $F$03 with $F$04, then every intermediate $F$05 also belongs to $F$06 (Chan, 2 Jan 2026). The minimal multisegment is called the minimal sequence associated to $F$07.

A second strand begins with the highest derivative multisegment. For each irreducible $F$08,

$F$09

and the main construction gives

$F$10

where $F$11 is the highest derivative (Chan, 2021). More generally, if $F$12 is admissible to $F$13, then the removal process produces a resultant multisegment $F$14 such that

$F$15

This is the double derivative result (Chan, 2021).

The minimal sequence has stability and commutativity properties. If $F$16 is minimal to $F$17, then every submultisegment $F$18 is minimal to $F$19, and

$F$20

Equivalently, within a minimal sequence, the order of the derivative operations can be permuted without changing the resulting simple quotient (Chan, 2 Jan 2026).

The proofs introduce fine chain orderings and local minimizability. Fine chains record successive “first segments” in the removal process, while local minimizability detects when a multisegment can be pushed downward in the Zelevinsky poset without changing the quotient (Chan, 2 Jan 2026). This suggests that the internal combinatorics of derivatives has acquired a canonical form that refines the original multisegment classification.

5. Geometric, categorical, and automorphic avatars

Bernstein-Zelevinsky derivatives admit a global automorphic analogue on $F$21. In that setting, restriction and Jacquet functors are replaced by constant term operators and degenerate Whittaker coefficient operators. The $F$22-th automorphic derivative operator is defined as a composition of the constant term along $F$23 and a Whittaker coefficient along a unipotent radical, and for an automorphic form $F$24 it can be expressed as

$F$25

Applied to Eisenstein series and their residues, these operators determine nonvanishing degenerate Whittaker coefficients, maximal nilpotent orbits, and Eulerianity of top coefficients (Zhang, 2022).

A geometric study of the partial Bernstein-Zelevinsky operator $F$26 on representations of $F$27 over a non-archimedean field relates it to Lusztig’s geometric induction. On standard modules,

$F$28

and the coefficients in the expansion on irreducibles satisfy

$F$29

In Grassmannian cases this becomes a Schubert-calculus computation, and a symmetric reduction reduces general cases to these special ones (Deng, 2024).

In categorified type-$F$30 settings, crystal derivative operators on quiver Hecke algebra modules categorify the Berenstein-Zelevinsky strings framework on quantum groups and generalize a graded variant of the classical Bernstein-Zelevinsky derivatives. For a BZ-sequence $F$31,

$F$32

and for RSK standard modules,

$F$33

Graded Specht modules for cyclotomic Hecke algebras arise as special cases of derived RSK modules, and graded cyclotomic decomposition numbers become a special subfamily of RSK decomposition numbers (Gurevich, 2021).

These developments show that the derivative calculus is no longer confined to smooth $F$34-adic representation theory. It now appears in automorphic Fourier analysis, geometric representation theory, and categorification, with the multisegment formalism surviving in each context in altered but recognizable form.

6. Archimedean, analytic, and duality extensions

For real and complex groups, the notion of derivatives and Bernstein-Zelevinsky filtration is subtler because of the richer topological and analytic structure. For Casselman-Wallach representations of $F$35 or $F$36, there is a canonical decreasing filtration on restriction to the mirabolic subgroup $F$37,

$F$38

directly analogous to the $F$39-adic case (Wu et al., 10 Jun 2026). More generally, for a real reductive group $F$40 and a parabolic $F$41 with abelian $F$42, one obtains a decreasing, closed $F$43-stable filtration whose bottom piece is the Casselman-Jacquet module and whose other subquotients are given functorially by twisted Jacquet functors and Schwartz inductions (Wu et al., 10 Jun 2026).

The same framework is tied to Casselman’s comparison conjecture. The conjecture is established for all $F$44, $F$45, and for quasi-split even orthogonal groups in some special cases, and derivative functors $F$46 are shown to be exact (Wu et al., 10 Jun 2026). A further extension develops Archimedean Bernstein-Zelevinsky theory for quasi-split real classical groups, proves that the homology groups $F$47 of the Jacquet functor are Casselman-Wallach, establishes the Euler-Poincaré characteristic formula

$F$48

and proves the vanishing of higher extension groups for irreducible generic Casselman-Wallach representations of $F$49 and $F$50 (Wu et al., 10 Sep 2025).

Duality is another major extension. For basic local Shimura varieties of Hodge-Newton reducible type, the Zelevinsky involution satisfies

$F$51

for supercuspidal $F$52, with a symmetric statement for $F$53-representations (Hamann, 2021). For locally analytic principal series representations, the Bernstein-Zelevinsky duality functor is defined by

$F$54

and duals of Kohlhaase-Schraen resolutions compute the expected dual principal series for the opposite Borel, with highest weight dual and inverse smooth character (Strauch et al., 8 Jan 2025). In exceptional rank, explicit Aubert-Zelevinsky duality computations for $F$55 exhibit the same interaction between parabolic induction, Jacquet functors, and Hecke algebras that characterizes the classical theory (Qin, 23 May 2025).

Across these Archimedean and analytic developments, Bernstein-Zelevinsky theory appears less as a theorem about one category than as a package of structures: filtrations by orbits, derivative or Jacquet-type functors, sign-projector or Whittaker mechanisms, and dualities compatible with induction, cohomology, and block decompositions.