---
title: Steinberg Representation in Reductive Groups
url: https://www.emergentmind.com/topics/steinberg-representation
type: topic
---

# Steinberg Representation in Reductive Groups

The Steinberg representation is a canonical representation attached to a reductive group, defined in several closely related ways depending on the ambient category: as an alternating sum of parabolic inductions, as the unique nonzero reduced homology of the spherical Tits building, as compactly supported top cohomology of the affine Bruhat–Tits building, or as the sign eigenspace for an Iwahori–Hecke algebra [2107.00794] [2501.01701] [1708.00717]. Its structural behavior is category-dependent: it is irreducible for connected reductive groups over infinite fields and in the classical finite-field and smooth \(p\)-adic settings, whereas the locally analytic Steinberg has a finite-length Jordan–Hölder series, and distinction problems for symmetric pairs need not satisfy multiplicity one [2107.00794] [1008.4910] [2410.03247].

## 1. Core constructions

For a connected reductive algebraic group \(G\) over a field \(k\), with semisimple \(k\)-rank \(r\), one standard construction defines the Steinberg representation by the unique nonzero reduced homology of the spherical Tits building:
\[
\St_G \;=\; \widetilde H_{r-1}\!\bigl(\T(G);F\bigr),
\]
where \(\T(G)\) is the simplicial complex of proper \(k\)-parabolic subgroups ordered by reverse inclusion [2107.00794]. In the anisotropic case, where \(G\) has no proper \(k\)-parabolics and \(r=0\), one sets
\[
\St_G \;=\; \widetilde H_{-1}(\varnothing;F)\;=\;F
\]
with trivial \(G(k)\)-action [2107.00794].

A second standard construction is an alternating sum of parabolic inductions in the Grothendieck group:
\[
\St_G\;=\;\sum_{P\subseteq G\;k\text{–parabolic}}(-1)^{\rk(P)}\,\Ind_P^G(\mathbf1).
\]
Under the hypotheses recorded for finite or infinite base fields, this virtual combination is realized by a single genuine representation [2107.00794].

For split groups over finite and non-archimedean local fields, the same object is expressed directly in terms of the relevant building. If \(G=G(F)\) is split over a finite field, then
\[
\St_G\;=\;H_{r-1}(B_{\mathrm{sph}};\mathbf Q),
\]
the top nonzero homology of the spherical building. If \(F\) is non-archimedean, then
\[
\St_G\;=\;H^{{\mathrm top}}_c(B_{\mathrm aff};\mathbf C),
\]
the compactly supported top cohomology of the affine Bruhat–Tits building [2501.01701].

For split adjoint quasi-simple groups over a non-archimedean local field \(K\), and coefficients in a commutative ring \(R\), one also has the parabolic-induction quotient
\[
\St(G,R)\;:=\; C^\infty(G/P,R)\Big/\sum_{P\subsetneq Q\subset G\text{ parabolic}} C^\infty(G/Q,R),
\]
together with an equivalent Iwahori model in terms of \(C_c(G/B,R)\) modulo explicit parahoric relations [1708.00717].

A further generalization replaces the minimal parabolic by an arbitrary standard parabolic \(Q=MN\). One sets
\[
\St_{\bar Q}^G(R)\;=\;\Ind_{\bar Q}^G(R)\Big/\sum_{Q'\supsetneq Q}\Ind_{\bar Q'}^G(R),
\]
recovering the classical Steinberg when \(Q\) is minimal and obtaining the trivial one-dimensional module when \(Q=G\) [1707.06187].

| Setting | Realization | Structural note |
|---|---|---|
| Connected reductive \(G/k\) | \(\widetilde H_{r-1}(\T(G);F)\) | Anisotropic case gives \(F\) |
| Split \(G(F)\), \(F\) finite | \(H_{r-1}(B_{\mathrm sph};\mathbf Q)\) | Top nonzero homology |
| Split \(G(F)\), \(F\) non-archimedean | \(H_c^{\mathrm top}(B_{\mathrm aff};\mathbf C)\) | Affine-building model |

These constructions exhibit a recurring principle: the Steinberg representation isolates the top combinatorial or cohomological contribution of the parabolic geometry of \(G\). A plausible implication is that many later variants are best understood as deformations of this “top piece” mechanism rather than as unrelated objects.

## 2. Irreducibility, Hecke algebras, and Iwahori models

In the finite-field case, Steinberg showed that \(\St_G\) is irreducible and affords the sign action of the finite Hecke algebra \(H(G,B)\): every simple reflection acts by \(-1\) on \(\St_G^B\). In the \(p\)-adic split case, \(\St_G\) is again irreducible and realizes the one-dimensional sign representation of the Iwahori–Hecke algebra \(H(G,I)\); more generally, for any unramified character \(\chi:G\to\mathbf C^\times\), one obtains
\[
\St_\chi\;=\;\St_G\otimes\chi
\]
inside the unramified principal series [2501.01701].

For connected reductive groups over an infinite field \(k\), Putman and Snowden proved that the Steinberg representation \(\St_G\) of the abstract group \(G(k)\) over any field of coefficients is irreducible [2107.00794]. Their proof identifies \(\St_G\) with \(F[U(k)]\) as a vector space, using a minimal \(k\)-parabolic \(B=U\rtimes Z_G(T)\), and reduces irreducibility to two statements: any nonzero vector can be moved to have nonzero augmentation, and any nonzero left ideal in \(F[U(k)]\) stable under a positive torus action is the whole algebra [2107.00794].

The Iwahori-spherical realization becomes especially explicit for generalized Steinberg representations in unramified principal series. If \(G(F)\) is split reductive over a \(p\)-adic field, \(J\) is the standard Iwahori subgroup, and \(I(x)=\Ind_B^G(x\delta_B^{1/2})\), then the Steinberg representation \(\St_x\) is the unique irreducible quotient of \(I(x)\) whose \(J\)-fixed line affords the sign character of the finite Iwahori–Hecke algebra. In the Casselman basis \(\{f_w\}_{w\in W}\),
\[
\phi_{\mathrm{St}}(g)\;:=\;\sum_{w\in W}(-q)^{-\ell(w)}f_w(g)
\]
is, up to scalar, the unique \(J\)-fixed vector on which each simple reflection acts by \(-1\), so \(\dim \St_x^J=1\). On this one-dimensional space, the full extended Hecke algebra acts through the character
\[
\chi(X_{s_a})=-1,\qquad \chi(X_d)=(x\delta_B)(d),
\]
and the associated Whittaker function satisfies
\[
W(dw)=0 \text{ unless } X(d)\text{ is \(w\)-dominant,}
\]
while in the \(w\)-dominant case
\[
W(dw)=(x\delta_B)(d)(-q)^{-\ell(w)}.
\]
This generalizes the corresponding \(\GL_n(F)\) formula to arbitrary split reductive groups [2407.01448].

A common oversimplification is that “the Steinberg representation is irreducible” without qualification. This is correct in the smooth finite-field, smooth \(p\)-adic, and infinite-field settings just described, but it fails in the locally analytic category discussed below [1008.4910].

The Hecke-theoretic perspective also supports a classification problem for smooth \(p\)-adic representations with depth-zero Steinberg content. For an unramified reductive \(p\)-adic group \(G\), if an irreducible smooth \((\pi,V)\) has \(V^{K_1}\) containing the Steinberg representation of the finite reductive quotient \(K/K_1\), then \(\pi\) has depth zero, is Iwahori-spherical, hence is a subquotient of some unramified principal series \(I(\chi)\). In each principal series there exists exactly one irreducible subquotient \(\pi_\chi\) containing the Steinberg representation in its hyperspecial subgroup, and the assignment \(\chi\mapsto \pi_\chi\) induces a bijection
\[
\{\pi_\chi\}\;\longleftrightarrow\;W\backslash\{\text{unramified }\chi\}
\]
[2603.22931].

## 3. Building-theoretic and harmonic-cochain realizations

For split adjoint quasi-simple groups over a non-archimedean local field \(K\), the dual of the Steinberg representation admits a purely building-theoretic description in terms of harmonic cochains [1708.00717]. Let \(\Delta\) be the Bruhat–Tits building, \(X_\ell\) the set of pointed chambers, and \(L\) an \(R\)-module with \(G\)-action. A harmonic cochain is an \(R\)-linear map
\[
h:R[X_\ell]\to L
\]
satisfying two conditions: the orientation relation
\[
h(v_{\sigma(0)},\dots,v_{\sigma(\ell)})=(-1)^{\mathrm{sign}(\sigma)}h(v_0,\dots,v_\ell),
\]
and the codimension-one vanishing condition
\[
\sum_{C\supset f} h(C)=0
\]
for each codimension-one face \(f\). The main theorem identifies this harmonic space with the \(R\)-dual of the Steinberg representation:
\[
\Har'(R,L)\;\simeq\;\Hom_R(\St(G,R),L).
\]
The proof uses the Iwahori model of \(\St(G,R)\) and explicit combinatorial identities in the extended affine Weyl group [1708.00717].

In the special case \(G=\PGL_2(K)\), the building is a \((q+1)\)-regular tree, pointed chambers are oriented edges, and the harmonicity conditions reduce to
\[
h(\bar e)=-h(e),\qquad \sum_{e^-=v}h(e)=0.
\]
This realizes the Steinberg dual as the usual space of harmonic anti-symmetric edge functions [1708.00717].

A closely related formulation appears in the analysis of symmetric pairs. For a connected reductive group \(G\) over a non-archimedean local field \(K\), let \(\Ch(G)\) be the set of chambers of the reduced Bruhat–Tits building and define a harmonic cochain by the panel relation
\[
\sum_{C\supset D}f(C)=0
\]
for every panel \(D\). With the quadratic orientation character \(\varepsilon_G:G\to\{\pm1\}\), the \(G\)-action is
\[
(g\cdot f)(C)=\varepsilon_G(g)\,f(g^{-1}C),
\]
and the smooth part \(\sH(G)^\infty\) is naturally isomorphic to the Steinberg representation \(\St_G\) [2410.03247].

These harmonic models are not merely alternative descriptions. They provide the concrete linear functionals, adjacency relations, and summation identities used in distinction problems, Poincaré-series constructions, and explicit comparison with Hecke operators. This suggests that the building is the most uniform geometric carrier of Steinberg phenomena across smooth \(p\)-adic settings.

## 4. Locally analytic and generalized Steinberg representations

For a split reductive \(p\)-adic group \(G\) over \(L\), with Borel \(B\subset G\), maximal torus \(T\subset B\), and simple roots \(\Delta\), Orlik and Schraen define the locally analytic Tits complex
\[
0\to I_G^G(1)\to C^0\xrightarrow{d^0}C^1\xrightarrow{d^1}\cdots\xrightarrow{d^{|\Delta|-1}}C^{|\Delta|}\to0,
\]
where
\[
C^i=\bigoplus_{K\subset\Delta,\ |K|=i} I_{P_K}^G(1).
\]
Its degree-zero cohomology is the locally analytic Steinberg representation
\[
\St_G^{\mathrm{la}}:=C^0/\operatorname{im}d^{-1},
\]
which coincides with the classical continuous Steinberg space of locally analytic vectors [1008.4910].

The complex is acyclic in all degrees except at the left. The proof is by induction on semisimple rank and uses three specific ingredients: the parabolic BGG resolution of algebraic \(L_I\)-modules, the Orlik–Strauch functors \(F_{P_I}^G(-,-)\) together with their bi-exactness and the \(PQ\)-formula, and the fact that applying \(F_{P_I}^G\) to the parabolic BGG resolution recovers the rows of the double complex resolving \(I_{P_K}^G(V_K(\lambda)')\) [1008.4910].

For a dominant algebraic weight \(\lambda\in X^*(T)\), the Jordan–Hölder constituents of \(\St_G^{\mathrm{la}}(\lambda)\) are precisely
\[
F_{P_{I(w)}}^G\bigl(L(w\cdot\lambda),\,v_{P_{I(w)}}^J\bigr),
\]
as \(w\in W\) and \(J\subset I(w)\) vary, with multiplicity
\[
m_{w,J}=(-1)^{|J|}\sum_{\substack{w'\in W\\ \operatorname{supp}(w')=J}}(-1)^{\ell(w')}m(w',w).
\]
In particular,
\[
m_{w,J}\neq0\iff J\subset\operatorname{supp}(w).
\]
When \(\lambda=0\), the only locally algebraic subquotient is the smooth Steinberg \(v_B^G\); the classical smooth Steinberg appears with multiplicity one, and there are no new smooth subquotients. All extra constituents are genuinely locally analytic and are parametrized by the proper subsets \(J\subset I(w)\) [1008.4910].

The same work introduces an analogue of the Jacquet functor for locally analytic representations. If \(U\) is the unipotent radical of \(B\), then
\[
H_0(U,V)=V/\overline{\langle u\cdot v-v\rangle}
\]
is the largest Hausdorff \(U\)-coinvariants. For \(M=L(\chi)\) simple in \(\mathcal O\), \(P\) maximal for \(M\), and smooth \(L_P\)-representation \(V\),
\[
H_0(U,F_P^G(M,V))\simeq \chi^{-1}\otimes J_{U\cap L_P}(V),
\]
and dually
\[
H^0(U,F_P^G(M,V)')\simeq \chi\otimes J_{U\cap L_P}(V)'.
\]
This determines irreducible factors \(F_P^G(L(w\cdot\lambda),v_P^J)\) from their \(U\)-Jacquet modules [1008.4910].

In the smooth category, the generalized Steinberg functor
\[
\St_{\bar Q}^G(\sigma)=e_G(\sigma)\otimes_R\St_{\bar Q}^G(R)
\]
is exact, \(R\)-linear, and commutes with arbitrary direct sums. It is compatible with scalar extension, with ordinary parts, and with iteration along nested parabolics. In particular,
\[
\Ord_Q\bigl(\St_{\bar Q}^G(\sigma)\bigr)\simeq e_M(\sigma),
\]
and if \(Q_1\subset Q_2\) then
\[
\St_{\bar Q_2}^G\circ \St_{\overline{M_2\cap Q_1}}^{M_2}\xrightarrow{\sim}\St_{\bar Q_1}^G.
\]
When \(R\) is noetherian and \(p\) is nilpotent in \(R\), the functor is fully faithful on smooth and admissible subcategories [1707.06187].

## 5. Modular, algebraic, and quantum forms

For a finite group of Lie type \(G\) over defining characteristic \(p\), and a field \(k\) of characteristic \(\ell\neq p\), the \(\ell\)-modular Steinberg representation is constructed inside the permutation module \(kG\cdot b\simeq \Ind_B^G(k)\), where
\[
b=\sum_{b\in B} b,\qquad c=\sum_{w\in W}(-1)^{\ell(w)}n_w\cdot b.
\]
One defines
\[
\St_k:=kG\cdot c.
\]
Steinberg’s theorem gives that \(\St_k\) is free of rank \(|U|\) over \(k\) with basis \(\{u\cdot c\mid u\in U\}\), and \(\St_k\) is irreducible precisely when \([G:B]\) is a unit in \(k\) [1504.04157].

The socle of \(\St_k\) is always simple. Via the Hecke algebra
\[
H_k=\End_{kG}(kG\cdot b),
\]
one has
\[
T_w\cdot c=(-1)^{\ell(w)}c,
\]
so the \(B\)-fixed line in the simple socle corresponds to the one-dimensional Hecke character
\[
\zeta:H_k\to k,\qquad T_w\mapsto (-1)^{\ell(w)}.
\]
For \(G=\GL_n(q)\), the socle label is described using the integer
\[
e=\min\{i\ge2\mid 1+q+\cdots+q^{i-1}\equiv0\!\!\!\pmod\ell\},
\]
and the composition factors of \(\St_k\) are multiplicity-free. Their number \(C_n\) has generating series
\[
\sum_{n\ge0} C_n t^n=(1-t)^{-1}\prod_{j\ge1}(1-t^{ej})^{-1}.
\]
Analogous multiplicity-free statements hold for finite classical groups at linear primes [1504.04157].

For a simple, simply connected algebraic group over an algebraically closed field of characteristic \(p>0\), the \(r\)th Steinberg module is
\[
\St_r=L((p^r-1)\rho)\simeq V((p^r-1)\rho)\simeq \nabla((p^r-1)\rho).
\]
It is central to Donkin’s conjectures on good \((p,r)\)-filtrations and tilting modules. Bendel, Nakano, Pillen, and Sobaje reduce the tensor-product question to \(r=1\): if \(\St_1\otimes L(\mu)\) has a good filtration for every \(\mu\in X_1\), then \(\St_r\otimes L(\lambda)\) has a good filtration for every \(r\ge1\) and \(\lambda\in X_r\). They verify this under the conditions \(p\ge 2h-4\), for all rank-two groups, for fundamental weights when \(p\ge3\), and in many rank \(\le5\) cases [1804.00613].

For a semisimple, simply connected algebraic group \(G\) at an arbitrary complex root of unity \(q\), the quantum Steinberg module is
\[
\St_q=L((\ell-1)\rho).
\]
It is simple and self-dual in \(\Rep(G_q)\), remains simple on restriction to the small quantum group, and is both projective and injective in \(\Rep(G_q)\) and in \(\Rep(\mathbf u_q)\). The same work proves that \(\Rep(G_q)\) has enough projectives and injectives, and that projectivity or injectivity can be tested after restriction to the small quantum group [2306.14453].

Taken together, these variants show that “Steinberg” does not denote a single categorical behavior. In the modular finite-group setting, reducibility may occur but the socle remains rigid; in the algebraic and quantum highest-weight settings, the Steinberg object is tied to filtration and projectivity phenomena rather than only to irreducibility.

## 6. Resolutions, arithmetic duality, and topological appearances

For \(\GL_n\) over a principal ideal domain \(A\), with field of fractions \(F\), the Steinberg module is
\[
\St_n(A;R):=\widetilde H_{n-2}(\Delta_{n-1};R),
\]
where \(\Delta_{n-1}(F^n)\) is the spherical building of proper nonzero \(F\)-subspaces of \(F^n\) [1106.5034]. Ash, Gunnells, and McConnell compare three explicit resolutions: the Lee–Szczarba simplicial resolution \(C_*(A)\), the line-based resolution \(C'_*(A)\), and the sharbly complex \(Sh_*(A)\). For \(n\le4\), they also use a Voronoi-based complex \(\mathcal V_*\). Each resolves \(\St_n\), and the comparison maps are quasi-isomorphisms. These constructions are then applied to cohomology of congruence subgroups of \(\SL_4(\mathbf Z)\), proving that the Voronoi complex does not introduce spurious Hecke eigenclasses [1106.5034].

For number rings and symplectic groups, the symplectic Steinberg module is defined by the symplectic Tits building:
\[
\St^\omega_{2n}(K)=\widetilde H_{n-1}\bigl(\mathcal T^\omega_{2n}(K)\bigr).
\]
Borel–Serre duality identifies \(\St^\omega_{2n}(K)\) as the dualizing module for finite-index subgroups of \(\Sp_{2n}(R)\). An explicit projective resolution is constructed from tensor products of Lee–Szczarba sharbly groups over ordered orthogonal decompositions of the symplectic space, with a boundary combining omit-terms and split-terms. When \(R\) is a Euclidean number ring and \(p\in R\) satisfies the surjectivity condition on units, this yields a computation of the top-degree cohomology of principal level-\(p\) congruence subgroups of \(\Sp_{2n}(R)\) [2605.06499].

The Steinberg representation also appears in the hit problem. For \(G=\GL_n(\mathbf F_q)\), Hai studies quotients of \(\mathbf F_q[x_1,\dots,x_n]\) arising from Stanley–Reisner rings of matroid complexes. In a degree
\[
d=k\Bigl(\frac{q^{n-1}-1}{q-1}\Bigr)-n
\]
for suitable \(k\), one obtains
\[
Q_d\bigl(S^*(V^*)\bigr)\cong \St_n(\mathbf F_q)\otimes \det^{\,k-1},
\]
and
\[
\dim \St_n(\mathbf F_q)=q^{n(n-1)/2}.
\]
For \(q=2\), this specializes to the Walker–Wood degree \(2^n-1-n\), and the Steinberg summand admits a decomposition into suspensions of Brown–Gitler modules [2106.01537].

These constructions place the Steinberg representation at the intersection of building homology, arithmetic duality, computational cohomology, and unstable algebra. A plausible implication is that the persistence of Steinberg modules in these settings reflects a common dualizing or top-degree mechanism rather than an accident of notation.

## 7. Distinction, branching, and relative Langlands phenomena

For split symmetric spaces \(X=G/H\), with \(G\) split reductive and \(H=G^\sigma\) the fixed points of an \(F\)-rational involution, the distinction problem for the Steinberg representation is related to harmonic functions on hypergraphs built from \(B\)-orbits or Iwahori orbits on \(X\). In the finite-field case,
\[
\dim \Hom_H(\St,1)=\dim \mathcal H(\Gamma_F(X)),
\]
and in the \(p\)-adic case
\[
\dim \Hom_H(\St_{\chi_0},1)=\dim \mathcal H(\Gamma^{0}_{\mathrm aff,F}(X)).
\]
Shtotland proves that, over a non-archimedean local field, \(\St_{\chi_0}\) is \(H\)-distinguished if and only if its Langlands parameter factors through the dual group of \(X\). More precisely, if \(\phi_{\St}:W_F'\to G^\vee\) is the Steinberg parameter and \(\iota:{}^LX\to G^\vee\) is Takeda’s embedding, then
\[
\St_{\chi_0}\text{ is \(H\)-distinguished}\iff \phi_{\St}\text{ factors through }\iota.
\]
This occurs exactly when \(X\) is quasi-split and no simple adjoint factor is of type
\[
\PGL_{2n+1}/P(\GL_n\times \GL_{n+1})
\]
[2501.01701].

In relative rank \(1\), Broussous obtains a reciprocity law for symmetric spaces \(G/H\) with \(G\) and \(H\) semisimple of relative rank \(1\). If \(K_i=P_i\cap H\) are the anisotropic subgroups attached to the \(H\)-orbits on the flag variety, then for any irreducible smooth representation \(\pi\) of \(H\),
\[
\dim\Hom_H(\St_G,\pi)
=\sum_{i=1}^r \dim\Hom_{K_i}(\tilde\pi,1)+\chi_{\St_H}(\pi).
\]
Moreover, for \(\pi\neq 1_H\), one has
\[
\Ext_H^k(\St_G,\pi)=0,\qquad k\ge1,
\]
and for \(\pi=1_H\) the Euler–Poincaré characteristic is \(r-1\) [1810.06910].

A more recent harmonic-cochain approach studies distinction for general symmetric pairs \((G,H)\) over non-archimedean local fields. Let \(\sF^{\max}(G)\) denote the maximal \(\theta\)-stable facets in the Bruhat–Tits building. Then, for a character \(\chi\) satisfying the stated pro-\(p\)-triviality condition,
\[
\dim_\mathbf C\Hom_H(\St_G,\chi)\le |\sF^{\max}(G)\!/\!/H|.
\]
A refinement replaces maximal facets by effective connected components of an apartment-graph \(\Gamma(G,\theta)\), yielding
\[
\dim_\mathbf C\Hom_H(\St_G,\chi)\le
\#\{\text{\(\chi\)-effective connected components of }\Gamma(G,\theta)\}.
\]
Under additional Poincaré-series hypotheses, equality holds and explicit bases of distinguished linear forms are constructed [2410.03247].

In the concrete case \(G=\GL_n(K)\) and \(H^+=\SO(\varepsilon)\), the harmonic-cochain method yields a complete classification of \(\theta\)-stable apartments and exact multiplicity formulas. Writing \(n=2k\) or \(2k+1\),
\[
\dim_\mathbf C\Hom_{H^+}\bigl(\St_{\GL_n},\mathbf1\bigr)=
\begin{cases}
\dfrac{(k+1)(k+2)}2,& n=2k,\ H\text{ split},\\[6pt]
\dfrac{k(k+1)}2,& n=2k,\ H\text{ quasi-split but not split},\\[4pt]
\dfrac{(k-1)k}2,& n=2k,\ H\text{ non-quasi-split},\\[6pt]
\dfrac{(k+1)(k+2)}2,& n=2k+1,\ H\text{ split},\\[4pt]
\dfrac{k(k+1)}2,& n=2k+1,\ H\text{ non-quasi-split}.
\end{cases}
\]
For the full orthogonal group \(H=\mathrm O_n(K)\), the same formula holds except at \((n=2,H\text{ split})\), where the dimension is \(4\) [2410.03247].

These results correct two frequent expectations. First, multiplicity-one distinction is not a general property of the Steinberg representation: in the split orthogonal examples, the distinguished dimension grows like a triangular polynomial in \(k\) [2410.03247]. Second, the relative local Langlands picture can nevertheless remain precise: for split symmetric subgroups, Steinberg distinction is governed by explicit factorization of the Langlands parameter through the dual group of the symmetric space [2501.01701].

Source: https://www.emergentmind.com/topics/steinberg-representation