Inductive Algebras for Finite Heisenberg Groups

Abstract

A characterization of the maximal abelian sub-algebras of matrix algebras that are normalized by the canonical representation of a finite Heisenberg group is given. Examples are constructed using a classification result for finite Heisenberg groups. Comment: 5 pages

Full text

Page 1

arXiv:0803.2921v1 [math.RT] 20 Mar 2008
INDUCTIVE ALGEBRAS FOR FINITE HEISENBERG GROUPS
AMRITANSHU PRASAD AND M. K. VEMURI
Abstract. A characterization of the maximal abelian sub-algebras of matrix algebras
that are normalized by the canonical representation of a finite Heisenberg group is given.
Examples are constructed using a classification result for finite Heisenberg groups.
1. Introduction
Let G be a separable locally compact group and ρ an irreducible unitary representation of
G on a separable Hilbert space H. Let B(H) denote the algebra of bounded operators on H.
An inductive algebra is a weakly closed abelian sub-algebra A of B(H) that is normalized
by ρ(G), i.e., ρ(g)Aρ(g)−1= A for each g ∈ G. If we wish to emphasize the dependence
on ρ, we will use the term ρ-inductive algebra. A maximal inductive algebra is a maximal
element of the set of inductive algebras, partially ordered by inclusion.
The identification of inductive algebras can shed light on the possible realizations of H as
a space of sections of a homogeneous vector bundle (see, e.g., [4, 5, 6, 7]). For self-adjoint
maximal inductive algebras there is a precise result known as Mackey’s Imprimitivity The-
orem, as explained in the introduction to [4].
Inductive algebras were introduced in [7], where the maximal inductive algebras for the
real Heisenberg groups were classified. In this paper, we classify the maximal inductive
algebras for finite Heisenberg groups. All these algebras are found to be self-adjoint. By
Mackey’s imprimitivity theorem, they arise from induction. Therefore, we get a classifi-
cation of the realizations of the canonical representation as induced representations. Our
proof of the classification of maximal inductive algebras in the finite case is very direct and
reveals the simple algebraic idea behind such a result.
In Section 2, we recall the representation theory of a general Heisenberg group, and
formulate our main result. Our definition of Heisenberg group is taken from [2]. The
main result is proved in Section 3. In Section 4, we give a classification of finite Heisenberg
groups and use it to construct a large class of maximal inductive algebras for their canonical
representations.
2000 Mathematics Subject Classification. 20C15, 20C25.
Key words and phrases. Finite Heisenberg groups, inductive algebras.
1

Page 2

2AMRITANSHU PRASAD AND M. K. VEMURI
2. Formulation of the Main Theorem
Let K be a locally compact abelian group, and U(1) = {z ∈ C : |z| = 1}. Consider a
locally compact group G which lies in a central extension:
(1)
1
??U(1)
??G
π
??K
??0
Assume that π admits a continuous section s. Thus s : K → G and π ◦ s = IdK. Define
e : K×K → U(1) by e(k,l) = s(k)s(l)s(k)−1s(l)−1. Then e is well defined and independent
of the choice of the section s. Furthermore, e is an alternating bicharacter, i.e., it is
a homomorphism in each variable separately, and e(k,k) = 1 for each k ∈ K. Finally, e
determines G up to unique isomorphism of central extensions. Let? K denote the Pontryagin
dual of K. Then a homomorphism e♭: K →? K is obtained by setting
(e♭(l))(k) = e(k,l).
Definition 2.1. If e♭is an isomorphism, then G is called a Heisenberg group.
Let G be a Heisenberg group as in (1). Then, Z(G) = U(1). We will assume that K is
finite. The resulting Heisenberg groups are customarily called finite Heisenberg groups.
Definition 2.2. Let M ⊆ K be a closed subgroup. If e|M×M ≡ 1 then M is called an
isotropic subgroup. A maximal isotropic subgroup is one which is a maximal element of the
set of isotropic subgroups, partially ordered by inclusion.
Note that there is a one to one correspondence between isotropic subgroups of K and
abelian subgroups of G containing Z(G) given by M ?→ π−1(M). Consequently, there is a
one to one correspondence between maximal isotropic subgroups of K and maximal abelian
subgroups of G.
If M ⊆ K is isotropic then we have an exact sequence of abelian groups
1
??U(1)
??π−1(M)
π
??M
??0
The exactness of the dual sequence
0
Z
??
?
π−1(M)
??
?
M
??
0
??
implies that the character idZ(G): Z(G) → U(1) extends to a character of π−1(M).
Let M ⊆ K be a maximal isotropic subgroup, and let χ ∈?
idZ(G). Then the representation ρ = IndG
representation. The following is a famous theorem of Stone, von Neumann and Mackey.
π−1(M) be such that χ|Z(G)=
π−1(M)χ is irreducible, and called the canonical
Theorem 2.3. If ρ′is an irreducible representation of G on a vector space V such that
ρ′|Z(G)= (idZ(G)) · I then ρ′ ∼= ρ.
A proof of this theorem may be found in [2]. We may think of ρ as acting on the space
V = {f : G → C | f(hg′) = χ(h)f(g′) for all h ∈ π−1(M), g′∈ G}

Page 3

INDUCTIVE ALGEBRAS FOR FINITE HEISENBERG GROUPS3
by (ρ(g)f)(g′) = f(g′g). If u,v : G → C are functions, define mu(v) = uv. Thus muis the
operator of multiplication by u. Let
A(M) = {mu| u(hg′) = u(g′) for all h ∈ π−1(M), g′∈ G}.
Then A(M) ⊆ End(V ) is a maximal abelian subalgebra. Furthermore, it is inductive. So
it is a maximal inductive algebra. We note that A(M) is self adjoint. Our main result is
the following.
Theorem 2.4. Every maximal ρ-inductive algebra is of the form A(M) for a maximal
isotropic subgroup M of K. In particular, all maximal ρ-inductive algebras are maximal
abelian subalgebras and are self-adjoint.
3. The Proof
For k ∈ K, define κ(k) : End(V ) → End(V ) by by κ(k)T = ρ(s(k))Tρ(s(k))−1. Then
κ(k) is independent of the choice of section s, and κ : K → GL(End(V )) is a representation
of the finite abelian group K. For each l ∈ K, let
End(V )l= {T ∈ End(V ) | κ(k)(T) = e♭(l)(k)T}.
Lemma 3.1. End(V ) has a multiplicity-free decomposition into K-invariant subspaces
End(V ) =
?
l∈K
End(V )l.
Moreover, the subspace End(V )lis spanned by ρ(s(l)).
Proof. Since
κ(k)(ρ(s(l))) = ρ(s(k))ρ(s(l))ρ(s(k))−1
= ρ(s(k)s(l)s(k)−1s(l)−1)ρ(s(l))
= e♭(l)(k)ρ(s(l)),
ρ(s(l)) ∈ End(V )l. The vectors ρ(s(l)), l ∈ K are linearly independent, as they lie in
distinct eigenspaces for K. Since dim(End(V )) = |K|, they must form a basis of End(V ),
and the result follows.
?
To prove Theorem 2.4, note that every maximal ρ-inductive algebra A is an invariant
subspace of End(V ) under the action of K by κ. By Lemma 3.1, such a subspace is the
direct sum of those End(V )lwhich are contained in it. Hence, A is determined by the set
M = {l ∈ K | ρ(s(l)) ∈ A}.
The commutativity of A implies that for any l,k ∈ M, e(l,k) = 1. The maximality of
A implies that M is a maximal subset of K with this property. In other words, M is a
maximal isotropic subgroup of K.

Page 4

4AMRITANSHU PRASAD AND M. K. VEMURI
For each g ∈ G, there exists zg∈ U(1) such that g = zgs(π(g)). Now, for any f ∈ V ,
ρ(s(l))f(g) = f(zgπ(g)s(l))
= f(zg(π(g)s(l)π(g)−1s(l)−1)s(l)π(g))
= e(s(l),π(g))χ(s(l))f(g).
Hence, ρ(s(l)) = mu, where u(g) = e(s(l),π(g))χ(s(l)). Since mupreserves V , u(hg) = u(g)
for all h ∈ π−1(M) and g ∈ G. Hence ρ(s(l)) ∈ A(M) for each l ∈ M, and therefore,
A ⊂ A(M). Since e♭|Mdescends to an isomorphism K/M →ˆ M,
dimA = |M| = |ˆ M| = |K/M| = dimA(M).
Therefore, A = A(M).
4. Examples
We shall construct examples of maximal isotropic subgroups, and hence of maximal
inductive algebras, using a classification the finite Heisenberg groups up to isomorphism.
Since the pair (K,e) as in Section 2 determines the Heisenberg group G up to isomor-
phism, it suffices to classify such pairs. If A is a locally compact abelian group denote its
Pontryagin dual byˆA.
Theorem 4.1. Suppose that K is a finite abelian group and e : K × K → U(1) is a
non-degenerate alternating bicharacter. Then there exists a finite abelian group A and an
isomorphism ϕ : A ׈A → K such that
e(ϕ(x,χ),ϕ(x′,χ′)) = χ′(x)χ(x′)−1for all x,x′∈ A and χ,χ′∈ˆA.
Proof. Choose a basis x1,...,xn of K such that the order of xi is di with d1|d2|···|dn.
Let ζ denote a fixed primitive d1th root of unity. For each i,j ∈ {1,...,n}, there exists a
non-negative integer qijsuch that
e(xi,xj) = ζqij.
Observe that qijis determined modulo d1and is constrained to be divisible by d1/(di,dj).
Let Q(e) denote the matrix (qij) associated to e as above.
Any automorphism α of K is of the form
α??n
for a suitable matrix (αij) of integers, where αijis determined modulo diand is constrained
to be divisible by di/(di,dj). If e′(k,l) = e(α(k),α(l)), then
j=1ajxj
?=?n
i=1
?n
j=1αijajxi
Q(e′) =tαQ(e)α.
In the above expression, α is identified with its matrix (αij). In particular, Q(e′) can be
obtained from Q(e) by a sequence of column operations followed by the corresponding
sequence of row operations. Following Birkhoff [1], we use the operations
βiσ: multiplication of the ith row and ith column by an integer σ such that (σ,di) = 1.

Page 5

INDUCTIVE ALGEBRAS FOR FINITE HEISENBERG GROUPS5
πij: interchange of ith and jth rows and columns when di= dj.
αijσ: addition of σ times ith row to the jth row, and σ times the ith column to the jth
column when di/(di,dj) divides σ.
Since e is alternating, Q(e) can be taken to be skew-symmetric with 0’s along the diag-
onal. Since e is non-degenerate, there must be an integer in the first row of Q(e) coprime
to d1. First, βiσfor appropriate σ can be used to make this entry equal to 1. If this entry
is in the jth column, then the constraints on qij force dj= d1. Now, π2j transforms the
top-left corner of the matrix into (0 1
be used to make the remaining entries of the first two rows and columns vanish.
The form e identifies the Pontryagin dual of the subgroup generated by the first (trans-
formed) generator with the subgroup generated by the second one. After splitting off the
subgroup spanned by these two generators, we are left with a finite abelian group with
n − 2 generators. Thus Theorem 4.1 is proved by induction on n.
−1 0). In particular d1= d2. Finally, α1jσand α2jσcan
?
Let G be a finite Heisenberg group as in (1) and ϕ and A be as in Theorem 4.1. For
any subgroup B of A, let B⊥⊂ˆA denote the subgroup of characters which vanish on B.
Then ϕ(B × B⊥) is a maximal isotropic subgroup of K. This class of maximal isotropic
subgroups is likely to be very rich since the embedding problem for finite abelian groups
involves wild phenomena [3].
References
1. Garrett Birkhoff, Subgroups of abelian groups, Proc. Lond. Math. Soc. 38 (1934), 385–400.
2. David Mumford, Madhav Nori, and Peter Norman, Tata Lectures on Theta III, Birkh¨ auser, Boston,
1991.
3. Claus Michael Ringel and Marcus Schmidmeier, Submodule categories of wild representation type, J.
Pure and App. Alg. 205 (2006), 412–422.
4. Giovanni Stegel, Inductive Algebras for Trees, Pacific J. Math. 216 (2004), 177–200.
5. Giovanni Stegel, Even automorphisms of trees and inductive algebras, Int. J. Pure Appl. Math. 29
(2006), no. 4, 521–552.
6. Tim Steger and M. K. Vemuri, Inductive algebras for SL(2,R), Illinois J. Math. 49 (2005), no. 1, 139–151
(electronic).
7. M. K. Vemuri, Realizations of the Canonical Representation, Proc. Indian Acad. Sci. Math. Sci. 118
(2008), no. 1, 115–132.
The Institute of Mathematical Sciences, CIT campus Taramani, Chennai 600113.
Chennai Mathematical Institute, Plot H1, SIPCOT IT Park, Padur PO, Siruseri 603103.