Smooth.tex Last revised 9:17 p.m.

June 16, 2019

Introduction to admissible representations of p-adic groups

Bill Casselman
University of British Columbia
[email protected]

Chapter II. Smooth representations

The theory of representations of p-adic groups started off by looking at unitary representations on Hilbert
spaces. After [Jacquet-Langlands:1970] introduced the category of admissible representations, the subject
largely lost its analytical flavour and became quite algebraic—in many aspects not very different from the
theory of representations of finite groups.
In this chapter I’ll define smooth and admissible representations of arbitrary locally profinite groups and
prove their basic properties. I shall generally take the coefficient ring to be a commutative Noetherian ring
R, assumed to contain Q. The point of allowing representations with coefficients in a ring like R is to allow
dealing with families of representations in a reasonable way. For example, suppose k to be a p-adic field and
G to be the multiplicative group k× . The unramified characters of G are those trivial on o× , and may be
identified with characters of Z. In this case it is useful to take R to be the ring C[z ±1 ], the group algebra of
Z, whose maximal ideals parametrize its complex characters. The group acts on this through the left or right
regular representation.
Occasionally R will be taken to be a field. Often, it will be a field D, which I assume throughout to be an
algebraically closed field of characteristic 0. In places I shall take D to be C. There are good reasons, as we
shall see, for not making this choice always. One might be called philosophical—the theory of representations
of a locally profinite group is essentially a matter of algebra, and the choice of D emphasizes this.
Throughout this chapter, G will be a locally profinite group. At the beginning it will be more or less arbitrary.
However, I’ll assume always that it possesses a countable basis of neighbourhoods of the identity—i.e. that
it is what I call in [Casselman:2018] a König group. Also, it will be countable at infinity in the sense that the
discrete set G/K is countable for one, hence all, compact open subgroups K .
Thus this chapter will generally present results valid without significant assumptions about the structure
of G. This necessarily excludes almost everything interesting, but includes many basic items. The current
version incorporates parts of [Bernstein:1992].
Contents
1. Admissible representations and the Hecke algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2. The centre of G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3. The contragredient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4. Representations of a group and of its Hecke algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
5. Characters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
6. Tensor products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
7. Matrix coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
8. Compact representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
9. Unitary representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
10. Schur orthogonality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
11. Induced representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
12. Appendix. Semi-simple algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
13. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 2

1. Admissible representations and the Hecke algebra [hecke.tex]

A smooth G module over R is a representation (π, V ) of G on an R-module V such that each v in V is fixed
by an open subgroup of G. A smooth representation (π, V ) is said to be admissible if for each open subgroup
K in G the subspace V K of vectors fixed by elements of K is finitely generated over R. If R is a field, this
just means that V K has finite dimension.
The subspace of smooth vectors in any representation of G is stable under G, since if v lies in V K then π(g)v
−1
lies in V gKg .
The group G acts on functions on G by the left and right regular representations:

Rg f (x) = f (xg), Lg f (x) = f (g −1 x) .

I’ll call a locally constant function f on G uniformly locally constant if there exists a compact open subgroup
K of G such that Rk f = Lk f for all k in K . The left and right actions of G commute, hence give rise to a
smooth representation L of G × G on the space of all locally constant functions that preserves the uniformly
locally constant ones.
The rational Hecke algebra HQ (G) is the ring of smooth rational measures of compact support on G, in
which multiplication is convolution. Let
HR = R ⊗Q HQ .
♥ [smooth-distribution] Normally, the ring R will be implicit, and HR will be written as just H. It is a consequence of Proposition
I.6.6 that given a choice of rational Haar (right-invariant) measure dx, smooth R-valued functions may be
identified with smooth R-valued distributions:

ϕ 7−→ Dϕ = ϕ(x) dx ,

so that X
hDϕ , f i = meas(K) ϕ(x)f (x)
G/K

if both f and ϕ are fixed on the right by elements of K .

Suppose that D1 and D2 are two smooth measures of compact support, corresponding to smooth functions
ϕ1 and ϕ2 . Then
Z Z 
π(D1 )π(D2 )v = ϕ1 (x)π(x) ϕ2 (y)π(y)v dy dx
ZG G

= ϕ1 (x)ϕ(y)π(xy)v dx dy
G×G
Z Z 
= ϕ2 (y) ϕ1 (x)π(xy)v dx dy
G
Z ZG 
= ϕ2 (y) ϕ1 (zy −1 )π(z)v dz dy
G G
Z Z 
= ϕ1 (zy −1 )ϕ2 (y) dz π(z)v dz
G G
= π(Dϕ )v

where Z
ϕ(z) = ϕ1 (zy −1 )ϕ2 (y) dy .
G

The measure Dϕ is also smooth and of compact support. It is the convolution D1 ∗ D2 . This product makes
H into a ring.
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 3

The Hecke algebra H does not have a multiplicative unit unless G is compact. This is the source of some
technical problems.
There exists for every smooth representation (π, V ) of G a canonical homomorphism from the ring H to
EndR (V ). Fix for the moment a right-invariant Haar measure dx on G. Suppose D to be Dϕ . For v in V
define Z
π(D)v = ϕ(x)π(x)v dx .
G

If v is fixed by the elements of the compact open group K and ϕ is fixed by elements of K with respect to the
right regular representation, this is also
X
meas(K) ϕ(x)π(x)v .
G/K

This definition depends only on Dϕ . We can in fact characterize, if
not define, π(D) solely in terms of the
distribution D. If F is a linear function on V , then Φ(x) = F π(x)v is a locally constant function on G, and
we may apply D to it. Then
Z



F π(D)v = ϕ(x)F π(x)v dx = hD, Φi .
G

If K is a compact open subgroup of G, the Hecke algebra H contains the subalgebra H(G//K) of distributions
invariant under left and right multiplication by elements of K . This algebra has as basis the measures µKgK/K
of H: Z
1
hµKgK/K , f i = f (x) dx
meas(K) KgK

as g varies over K\G/K . In particular

Z
1
π(µK/K ) = π(k) dk
meas(K) K

amounts to projection onto K -fixed vectors. If π is any smooth representation of G then every element of
H(G//K) takes V K into itself, and
X
π(µKgK/K ) = π(x)π(µK/K )
KgK/K
X
(II.1.1) = π(kg)π(µK/K )
K/K∩gKg−1

[kgk/k] = |KgK/K| · π(µK/K )π(g)π(µK/K ) .

since the map from K/K ∩ gKg −1 to KgK/K , taking k to kg , is a bijection.

The projection µK/K is the unit of H(G//K).
For every closed subgroup H of G, define V (H) to be the subspace of V generated by the π(h)v − v for h
in H . This subspace of V is characterized by the property that every H -equivariant linear map from V to a
vector space on which H acts trivially factors through V /V (H). If H is compact, something better happens:
[projection] II.1.2. Proposition. For any compact open subgroup K and smooth representation V , we have an equality
 
V (K) = v ∈ V  π(µK/K )v = 0

and a direct sum decomposition

V = V (K) ⊕ V K .
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 4

Proof. If v is fixed by K∗ then

1 X
π(µK/K )v = π(k)v
[K: K∗ ]
K/K∗

and
1 X
v= v.
[K: K∗ ]
K/K∗

If we subtract the first from the second, we get

−1 X

v − π(µK/K )v = π(k)v − v
[K: K∗ ]
K/K∗

[abelian-smooth] II.1.3. Proposition. Suppose

0 −→ U −→ V −→ W −→ 0
to be an exact sequence of G-representations. If the representation on V is smooth so is that on U and W ,
and for every compact open subgroup K the sequence

0 −→ U K −→ V K −→ W K −→ 0

is also exact.
[exactness-cor] II.1.4. Corollary. If V is smooth, it is admissible if and only if both U and W are.
Thus the categories of smooth and admissible representations are abelian.
[restriction-to-K] II.1.5. Proposition. Suppose K
to be a fixed compact open subgroup of G, and suppose R to be D. A smooth
representation defined over R is admissible if and only if the restriction of π to K is a direct sum of irreducible
representations, each occurring with finite multiplicity.
One has to be a bit careful, because an irreducible representation over a field might become reducible over a
field extension. This is not a problem if the field is algebraically closed.
Proof. Choose a sequence of compact open subgroups Kn normal in K and with {1} as limit. Then
V = V (Kn ) ⊕ V Kn . But every smooth finite-dimensional representation of K factors through some Kn .
As I remarked earlier, in many ways things do not behave here very differently from how they do for finite
groups.
The following is trivial, but best to state formally:
[irr-hom] II.1.6. Proposition. If (πi , Vi ) are two smooth representations of G over D, then
n
D if π1 is isomorphic to π2
HomG (V1 , V2 ) =
0 otherwise.

Suppose that G possesses a countable basis {Kn } of compact open subgroups such that
[fg-field] II.1.7. Lemma.
HQ (G//Kn ) is a finitely generated ring. Then every admissible representation (π, V ) may be defined over a
field F generated over Q by a countable set of generators.
We shall see eventually that this hypothesis holds when G is the group of rational points on a Zariski-
connected reductive group defined over a p-adic field.
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 5

2. The centre of G [hecke.tex]

In this section, assume the coefficient ring to be D.

If (π, V ) is an admissible representation of G then each space V K is stable under the centre ZG of G. The
subgroup ZG ∩ K acts trivially on it.
[schur-scalars] II.2.1. Proposition. If (π, V ) is irreducible the centre acts as scalar multiplication by a single character.
Proof. Suppose V K 6= {0}. Since V K has finite dimension and D is algebraically closed, there exists a
character ω of ZG with values in D× such that the subspace
 
V K (ω) = v ∈ V K  π(z)v = ω(z) · v for all z ∈ ZG 6= {0} .

If U is the subspace of all v in V satisfying this equation, it is both non-trivial and stable under G, hence equal
to all of V .
In general, I call an admissible representation centrally simple if this occurs. If ZG acts through the character
ω then π is called an ω -representation. For any central character ω with values in D× the Hecke algebra HD,ω
is that of uniformly smooth functions on G compactly supported modulo ZG such that

f (zg) = ω(z)f (g) .

If π is centrally simple with central character ω it becomes a module over the Hecke algebra Hω−1 :
Z
π(f )v = f (x)π(x)v dx ,
G/ZG

which is well defined since f (zx)π(zx) = f (x)π(x).

But now I generalize this somewhat.
If C is a commuting set of linear operators acting on a vector space U of dimension n and γ a map from C to
D, let  
n
U ((γ)) = v ∈ U  c − γ(c) v = 0 for some n and all c ∈ C .

[commuting] II.2.2. Lemma. Suppose C to be any commuting set of linear operators acting on the finite dimensional vector
space U over D. There exists a direct sum decomposition of U into non-zero spaces U ((γ)).
It is called the primary decomposition of U .
Proof. The slight technical problem is that no assumption is made on the size of C .
Q
Suppose at first that C is made up of a single element c. Let P (x) = (x − γi )mi be the characteristic
polynomial of c. For m larger than all mi
Y
(c − γi )m = 0 .
i

Since there exist polynomials a(x) and b(x) such that

Y
1 = a(x)(x − γ1 )m + b(x) (x − γi )m ,
i6=1

a simple inductive argument implies that U is the direct sum of primary eigenspaces with respect to c.
If C is finite a similar inductive argument will imply the Lemma. For each finite subset of C the primary
decomposition of U assigns to that subset a partition of n = dim(U ), that determined by the dimensions of its
subspaces U ((γ)). There are only a finite number of these partitions; choose one of greatest length. Suppose
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 6

it belongs to the subset S in C . Any larger subset must determine a refinement of the decomposition for S ,
and hence must actually be the same. The decomposition for S is therefore one for all of C .
From this follows immediately:
(π, V ) is a finitely generated admissible representation of G, the restriction of π to ZG
[centre] II.2.3. Proposition. If
is a direct sum of primary components V ((ω)), where the ω vary over a finite set of homomorphisms from
ZG to D× .
The characters ω occurring in this decomposition are called the central characters of π .

3. The contragredient [hecke.tex]

Suppose (π, V ) to be a smooth representation of G. Let

Vb = HomR (V, R) .

The group G acts on Vb according to the recipe

hb
π (g)b v , π(g −1 )vi .
v , vi = hb

π , Ve ) is that on the
The point is that the canonical pairing is G-invariant. The contragredient representation (e
b
smooth vectors in V .
Suppose (π, V ) to be a smooth representation of G and K a compact open subgroup of
[K-fixed-dual] II.3.1. Proposition.
G. Restriction of f to V K is an isomorphism of Ve K with HomR (V K , R).
Proof. Because V = V K ⊕ V (K) and the functions in Ve K are precisely those annihilating V (K).
From the exact sequence of R-modules
Rn −→ V K −→ 0
we deduce
0 −→ HomR (V K , R) −→ HomR (Rn , R) ∼
= Rn .
Therefore Ve K is finitely generated over R, and π
e is again admissible.
In general, Ve may be very small. However, in special circumstances it will be sufficiently large. I’ll call (π, V )
free over R if each V K is free over R. There are many examples of such representations.
If V is admissible and free over R, then the contragredient is also free over R, and the
[free-c] II.3.2. Lemma.
canonical map from V into the contragredient of its contragredient is an isomorphism.
Furthermore:
[contraexact] II.3.3. Corollary. Suppose U , V , W all to be free over R. If

0→U →V →W →0

is a short exact sequence of admissible representations, then so is

f → Ve → U
0→W e → 0.

Suppose U and V both to be smooth representations. Given a G-equivariant map from U to V , we get by
duality one from Vb to U
b.
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 7

[dualdual] II.3.4. Proposition. Suppose U smooth, V admissible, both free over R. The canonical map defined above:

HomG (U, V ) −→ HomG (Ve , U

e)

is an isomorphism.
Proof. We can define an inverse. Dualizing once again we get an inverse map

e

e , Ve
HomG U e .

e
e and since V is admissible we know that V may be identified with Ve , so e
But U is canonically embedded in U
we get an inverse map


HomG Ve , U
e −→ HomG (U, V ) .

Remark. I am not completely happy with my treatment of the contragredient when the coefficient ring is not
a field, and I am not at all sure what the right approach should be. There are certainly some cases in which
the current treatment is entirely satisfactory.

4. Representations of a group and of its Hecke algebra [hecke.tex]

What is the relationship between a smooth representation of G and the associated representation of its Hecke
algebra H?
[hecke-same] II.4.1. Proposition. Suppose (πi , Vi ) are two smooth representations of G. Then

HomG (V1 , V2 ) = HomH (V1 , V2 ) .

A mild technical problem here is that H has no multiplicative identity.

Proof. Any G-homomorphism is clearly a homomorphism of modules over the Hecke algebra as well. So
suppose now that one is given a map F of modules over the Hecke algebra. Suppose v in V1 , g in G, and
choose a compact open subgroup K fixing v , π1 (g)v , F (v), and π2 (g)F (v). Then



F π1 (µKgK )v
F π1 (g)v =
|KgK/K|
π2 (µKgK )F (v)
=
|KgK/K|
= π2 (g)F (v) .

A smooth representation is said to be co-generated by a subspace U if every non-zero G-stable subspace of

V intersects U non-trivially. This is dual to the condition of generation, in the following sense:
[co-generation] II.4.2. Lemma. Suppose R to be a field and K to be a compact open subgroup of G. An admissible represen-
tation (π, V ) is generated by V K if and only if its smooth contragredient is co-generated by Ve K .

Proof. Suppose that V is generated by V K , and suppose U to be a G-stable subspace of Ve with U ∩ Ve K =

e
K
U K = 0. If U ⊥ is the annihilator of U in Ve = V , then V /U ⊥ e K = 0. Thus V K = (U ⊥ )K , and since
=U
V K generates V , V = U ⊥ and U = 0. The converse argument is similar.
[two-hecke] II.4.3. Proposition. Suppose that (πi , Vi ) are two smooth representations of G and that K is a compact open
subgroup of G. If
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 8

(a) the space V1 is generated as a G-space by V1K ;

(b) the space V2 is co-generated as a G-space by V2K .
Then the canonical map
HomG (V1 , V2 ) −→ HomH(G//K) (V1K , V2K )
is an isomorphism.
These conditions are satisfied if V1 = V2 is irreducible and K is small enough, for example.
Proof. If F lies in HomG (V1 , V2 ) then for any f in H we have


F π1 (f )v = π2 (f )F (v)

for every f in H and v in V1K . Conversely, if we are given F in HomH(G//K) (V1K , V2K ) then since V1K
generates V1 this formula will serve to define a G-map from V1 to V2 once we know that
if v lies in V1K , f in H, and π1 (f )v = 0 then π2 (f )F (v) = 0.
But if π1 (f )v = 0 then for every h in H

π1 (µK/K ∗ h)π1 (f )v = π1 (µK/K ∗ h ∗ f ∗ µK/K )v = 0 .

Since F is assumed to be H(G//K)-equivariant,

π2 (µK/K ∗ h ∗ f ∗ µK/K )F (v) = π2 (µK/K ∗ h)π2 (f )F (v) = 0

for every h in H. This means that the G-space U generated by π2 (f )F (v) satisfies U K = {0}. This means by
assumption that it U = 0.
[hk-irr] II.4.4. Proposition. Suppose R to be D and (π, V ) to be a smooth representation of G. Then
(a) if π is irreducible then V K is an irreducible module over H(G//K) for all K ;
♥ [two-hecke] (b) if V satisfies conditions (a) and (b) of Proposition II.4.3 and V K is an irreducible module over
H(G//K) then π is irreducible.
Proof. Suppose (π, V ) to be irreducible, and let U be any non-trivial H(G//K)-stable
P subspace of V K . Since
V is irreducible, U must generate V as a G-space, so every v in V is of the form ci π)gi ui with ui in U . But
then for v in V K
X ci
v = π(µK/K )v = ci π(µK/K )π(gi )ui = π(Kgi K)ui
|KgK/K|

which lies in U since U is assumed to be stable under H(G//K). So V K ⊆ U .

♥ [two-hecke] Conversely, assume conditions (a) and (b) of Proposition II.4.3 to hold for V , and assume V K irreducible. If
U is any non-zero G-stable subspace of V then by (b) U K 6= 0 must be a submodule of V K , but will equal it
because of irreducibility. But (a) implies that then U = V .
Suppose R to be D and (π, V ) to be an irreducible admissible representation of G, K
[hk-irr-jacobson] II.4.5. Proposition.
compact and open in G. The homomorphism from H(G//K) to EndD (V K ) is surjective.
♥ [density] Proof. This follows from Theorem II.12.1 in the Appendix to this Chapter.
Does every finite-dimensional module over H(G//K) arise as the space V K for some admissible V ? And
♥ [two-hecke] more particularly one satisfying the conditions (a) and (b) of Proposition II.4.3? We can obtain a partial
answer to these questions. It is motivated by a simple observation. Let V be an admissible representation of
G, U = V K . To each v in V we can assign the function

Fv : G −→ U, g 7−→ π(µK/K )π(g)v

 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 9

Then f ∗Fv = π(f )Fv for every f in H(G//K), and the map from V to C ∞ (G, U ) is equivariant with respect
to the right regular action of G.
Conversely, if U is a finite-dimensional representation of H(G//K), define IU to be the space of all functions
F : G → U such that f ∗F = π(f )F for all f in the Hecke algebra. There is a canonical embedding of U itself
into this, and let V be the subspace of IU generated by this copy. It is not hard to verify that V K = U , and
that V is also co-generated by U . But whether this representation of G is admissible presumably depends on
G.

5. Characters [characters.tex]

In this section, suppose R = D.

If (π, V ) is admissible then for every f in H(G) the trace of π(f ) is well defined since it may be identified with
an operator on some V K , which is finite-dimensional. This defines the character of π as a linear functional
on the Hecke algebra.
the (πi , Vi ) make up a finite set of inequivalent irreducible admissible representations
[character] II.5.1. Proposition. If
of G then their characters are linearly independent.
♥ [hk-irr] Proof. Choose K so small that ViK 6= 0 for all i. They then form, according to Proposition II.4.4 and
♥ [two-hecke] Proposition II.4.3, inequivalent modules over H(G//K). Because Q the πi are all distinct as well as irreducible,
♥ [density] Theorem II.12.1 implies that the map from the Hecke algebra into End(Ui ) is surjective. Suppose now that
X
ci · TRi = 0 ,

which means that X

ci · TR πi (f ) = 0

for all f in the Hecke algebra. But then we can choose f in the Hecke algebra such that πi (f ) = I but all the
other πj (f ) = 0, which implies that ci = 0.
♥ [abelian-smooth] The following is a consequence of Proposition II.1.3:
[char-exact] II.5.2. Proposition. If
0 −→ U −→ V −→ W −→ 0
is an exact sequence of admissible G-spaces, then the character of V is the sum of the characters of U and
V.
It implies easily one half of this refinement:
Two admissible representations of finite G-length have the same Jordan-Hölder factors if
[jh] II.5.3. Corollary.
and only if they have the same characters.
Proof. It remains to be seen that if U and V have the same characters then they have the same Jordan-Hölder
factors. But for this, by the previous result, it suffices to see that the semi-simplifications of U and V are
♥ [character] isomorphic. But this follows from Proposition II.5.1 and an induction argument.
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 10

6. Tensor products [characters.tex]

Suppose the (π, Vi ) (i = 1, 2) to be smooth representations of Gi , with D = R. The group G1 × G2 then acts
on the tensor product:
(π1 ⊗R π2 )(g1 , g2 ): v1 ⊗ v2 7−→ π(g1 )v1 ⊗ π(g2 )v2 .

[tensor-prod] II.6.1. Proposition. If the πi are irreducible, so is π1 ⊗R π2 . Every irreducible representation of G1 × G2 is of

this form.
Proof. Let G = G1 × G2 . Suppose V 6= 0 to be any G-stable subspace of V1 ⊗ V2 , and U 6= 0 an irreducible
♥ [irr-hom] G1 -stable subspace of V . As a representation of G1 , V is a direct sum of copies of V1 , so that by Proposition
II.1.6 the representation on U is isomorphic to that on V1 , and we may as well assume that U = V1 . and the
canonical map from U ⊗ HomR (U, V ) to V is an isomorphism. For the same reasons, the canonical map from
U ⊗ HomG (U, W ) to W is an isomorphism. For the same reasons again V is isomorphic to U ⊗ HomG (U, V ).
But the embedding of W into V induces an embedding of HomG (U, W ) into HomG (U, V ). The second is
isomorphic to V2 as a representation of G2 , and since V2 is irreducible this embedding is an isomorphism.
Hence W = V .
Similarly, if V is any irreducible representation of G1 × G2 and U 6= 0 is any irreducible G1 -stable subspace,
then V is isomorphic to U ⊗ HomR (U, V ), and Hom(U, V ) is an irreducible representation of G2 .
The group G1 × G2 also acts on the space HomD (V1 , V2 ). The pair (g1 , g2 ) takes f to

π2 (g1 ) · f · π1 (g2−1 ) .

(Confusing right and left is an inherent problem in this business.) The subspace of homomorphisms invariant
with respect to the diagonal copy of G is just HomG (V1 , V2 ). There is a canonical map from V2 ⊗ Vb1 to the
space HomD (V1 , V2 ) taking v2 ⊗ vb1 to the linear map

u 7−→ hb
v , uiv .

The image consists of the maps of finite rank.

7. Matrix coefficients [matrix-c.tex]

If (π, V ) is an admissible representation the matrix coefficient associated to the pair v in V , ve in Ve is the
R-valued function
Fṽ,v = he
v , π(g)vi ,
which is uniformly smooth. Let A(π) be the space of smooth functions spanned by the matrix coefficients of
π . It is a smooth representation of G × G (one factor acting on the left, one on the right), and the map from
Ve ⊗ V to C(G) is G × G-equivariant. In particular, for any fixed ve the map

Fṽ : v 7−→ Fṽ,v

is equivariant from V to C(G), with respect to the right regular action.

Now suppose that R = D. Let A(G) = AD (G) be the space of uniformly smooth functions on G contained
in a G × G-stable admissible subrepresentation of C ∞ (G). This is what Harish-Chandra called the space of
automorphic forms on G.
Suppose F to be a smooth function on G fixed on left and right by elements of some
[mc] II.7.1. Proposition.
compact open subgroup. The following are equivalent:
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 11

(a) the function F is contained in some A(π) with π admissible;

(b) the space AL (F ) spanned by all Lg F is an admissible LG -representation;
(c) the space AR (F ) spanned by all Rg F is an admissible RG -representation;
(d) the function F lies in A(G).
Proof. Suppose that V = AR (F ) is an admissible representation of G. We want to find to see ve in Ve and v in
V such that
F (g) = he
v , Rg vi .
It is natural to take v = F , and which case this equation can be written as

[Rg F ](1) = he
v , Rg F i .

This suggests defining

he
v , f i = f (1)
for all f in V . Then
v , Rk−1 f i = f (k −1 ) = [Lk f ](1)
hRk ve, f i = he
for all f in V . But the left action of G commutes with the right action, so if Lk F = F for all k in the compact
open subgroup K then Lk v = v for all v in AR F , and Rk ve = v , so that ve lies in Ve .

8. Compact representations [matrix-c.tex]

In this section, suppose R = D.

Suppose (π, V ) to be a smooth representation of G possessing a central character ω . All of its matrix
coefficients are then eigenfunctions for ZG . Using terminology introduced by [Bernstein:1992], I’ll call π
v , π(g)vi (ve in Ve , v in V ) have compact support modulo ZG . We’ll
compact if all of its matrix coefficients he
see later that, according to a theorem of [Jacquet:1971], that these can be characterized as super-cuspidal
representations (in the terminology of Harish-Chandra) when G is reductive. The following result is implicit
in the proof of Theorem I.6 in [Bernstein:1992].
[compact-fd] II.8.1. Proposition. Suppose (π, V ) to be compact, K a compact open subgroup of G, v in V . Then the space
 
π(µK/K )π(g)v  g ∈ G

has finite dimension.

Proof. Suppose that it did not have finite dimension. Suppose v fixed by the compact open subgroup K◦ .
One could then choose an infinite number of linearly dependent vectors vi = π(µK/K )π(gi )v . The union
S
gi K◦ must then be infinite modulo ZG . The vi are all in V K , so we may find ve in Ve K such that he
v , vi i = 1
for all i. This contradicts the assumption that the support of matrix coefficients is compact modulo ZG .
[B-eq] II.8.2. Corollary. A smooth ω -central representation is compact if and only if the function π(µK/K )π(g)v has
compact support on G.
[compact-admissible] II.8.3. Corollary. A finitely generated compact representation is admissible.
R = D. An irreducible compact representation is projective and injective in the
[cuspidals-projective] II.8.4. Proposition. Suppose
category of smooth representation of G with the same central character.
Proof. I’ll begin the proof here but postpone part of it to a later section. It is motivated by the analogous
case of modules over a commutative ring with unit, in proving that a module is projective if and only if it is
a summand of a free module. But in our case there are some minor difficulties because the Hecke algebra is
not commutative and does not possess a unit.
Let ω be the central character of π . By definition, the representation (π, V ) embeds into the right regular
representation of G on Hω . Identify V with its image. I’ll prove later that
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 12

there exists an equivariant projection from Hω onto V , which is to say an equivariant surjection P
such that P(v) = v for v in V .
It will be of the form
P(f ) = Rf ∨ v0
for some particular v0 6= 0 in V . Here f ∨ (g) = f (g −1 ). Given this, we now know V to be a summand of Hω .
Suppose now that (σ, U ) is an arbitrary smooth representation of G and that we are given a surjection U → V
of smooth representations. Choose u0 in U mapping to v0 . Let Π be the map from Hω to U taking f to
σ(f ∨ )u0 . This is G-equivariant. The diagram

Hω
Π P

F
images/h.eps
U V

is commutative, and the restriction of Π to the image of V in Hω amounts to a splitting of the surjection. This
shows that V is projective.
♥ [compact-admissible] Injectivity follows from the projectivity of its contragredient, since according to Corollary II.8.3 π is admissible
and hence the smooth dual of Ve is the same as V .

9. Unitary representations [unitary.tex]

In this section I take R to be C.

If V is a vector space over C, its conjugate V is the same additive group, but with the new complex
multiplication
c ◦v = c v .
By definition, an element of Vb is a function f : V → C such that f (cv) = c f (v), whereas the space Vb is that
of all C-linear maps from V to C, but with the multiplication [c ◦f ](v) = cf (v).

[conj-dual] II.9.1. Lemma. The map taking f to f is an isomorphism of Vb with Vb .

A Hermitian pairing of two spaces U and V is a C-linear function H: U ⊗ V → C. Equivalently, it is an

R-linear function taking u ⊗ v to u • v such that

(cu) • v = u ⊗ (c v) = c(u • v) .

It is also equivalent to a linear map ι = ιH : U → Vb :

ι: u 7−→ [v 7−→ u • v]

characterized by the equation

u • v = hι(v), ui .

A Hermitian form on a complex vector space U is a Hermitian pairing of U with itself. If H takes u ⊗ v to
u • v then the map
t
H: u ⊗ v 7−→ v • u
is also a Hermitian form, and this defines an involution on the space of Hermitian forms on U . The form is
called symmetric if tH = H , or in other words

v•u = u•v
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 13

for all u, v . Thsi involution is a linear transformation over R. The space of all Hermitian forms is the direct
sum of ±1 eigenspaces. These are isomorphic, since H 7→ i H swaps them. A form is symmetric when it lies
in the +1-eigenspace.
[symmetric-cond] II.9.2. Lemma. A form is symmetric if and only if u • u is real for all u.
Proof. The well known polarization identity asserts that



y • x = (1/4) (x + y) • (x + y) − (x − y) • (x − y) − i (x + iy) • (x + iy) − (x − iy) • (x − iy) .

Verifying this is straightforward. Swapping x and y , it implies that

y•x = x•y

if every u • u is real.
I’ll call the form non-degenerate if ι is an injection. Of course if V is finite-dimensional this happens if and
only if ι is an isomorphism.
Suppose that the form is non-degenerate, that V is finite-dimensional, and that the form is symmetric. The
map ι satisfies the equation

[iota-symm] (II.9.3) hι(v), ι−1 (e

v )i = he
v , vi .

Also in these circumstances the form on V gives rise to one on Vb :

[vdotv-conj] (II.9.4) b • vb = ι−1 (b

u v ) • ι−1 (b u, ι−1 (v)i = hb
u) = hb v , ι−1 (u)i, .

If c 6= 0 is real, and ι is replaced by c · ι, then u • v is replaced by c(u • v), and ι−1 is replaced by (1/c)ι−1 .

[vxv-conj] (II.9.5) (b
u ⊗ u) • (b u, ι−1 (b
v ⊗ v) = hb v )i · hι(v), ui

is therefore canonical.
A unitary representation of G is one with a positive definite G-invariant Hermitian inner product. According
♥ [symmetric-cond] to Lemma II.9.2, it is necessarily symmetric. This implies also that ι is G-equivariant. Unitary representations
are important in our subject because they are the ones that appear in orthogonal decompositions of arithmetic
quotients, and this has arithmetic consequences. In one classic example, unitarity is related to Ramanujan’s
conjecture.
If (π, V ) is admissible, then any G-invariant form on V induces Hermitian forms on each V K and the image
d
of ι in V K . Therefore we may apply the previous discussion to V itself.
[schurs-unitary] II.9.6. Lemma. If (π, U ) and (ρ, V ) are both irreducible unitary admissible representations, then
n
C if π is isomorphic to ρ
Hom(U, Vb ) =
0 otherwise.

[unitary-sum] II.9.7. Proposition. Everyadmissible unitary representation is a countable direct sum of irreducible unitary
representations, each occurring with finite multiplicity.
This requires the assumption that G possesses a countable basis of neighbourhoods of the identity.
It is easy to see that the matrix coefficients of a unitary representation are bounded. A much stronger condition
on matrix coefficients is fundamental. Suppose π to be an irreducible representation with central character
ω . It is said to be square-integrable modulo the centre ZG of G if |ω| = 1 (i.e. its central character is unitary)
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 14

and every matrix coefficient is square-integrable on G/ZG . A compact representation with unitary central
character is square-integrable modulo the centre, for example.
[irr-sqint] II.9.8. Proposition. If π is an irreducible admissible representation of G, then it is square-integrable if and
only if a single non-zero matrix coefficient is square-integrable.
A square-integrable representation may be embedded into L2 (G), and it is unitary. More precisely:
[sqint-unitary] II.9.9. Proposition. Suppose (π, V ) to be an irreducible, square-integrable, admissible representation. For
ve0 6= 0 in Ve the pairing Z
u•v = hπ(g)u, ve0 ihπ(g)v, ve0 i dg
G/ZG

defines a G-invariant positive definite inner product on V .

10. Schur orthogonality [schur.tex]

Throughout this section, R will be a field. At the beginning it will be C. Assume also:
Restriction to ZG induces an isomorphism of Hom(G, R× ×
>0 ) with Hom(ZG , R>0 ).

This is true, for example, if G is the group of rational points on a Zariski-connected reductive group defined
over a local field. As a consequence of this assumption:
[assumption-cor] II.10.1. Lemma. (1) If χ is any character of ZG , there exists a character ρ of G such that χ · ρ is unitary on ZG .
(2) The group G is unimodular.
The following is the basic version of Schur orthogonality:
[schur-unitary-2] II.10.2. Proposition. (Unitary Schur orthogonality) If π is square integrable modulo ZG then for some γπ > 0
Z
he
u, π(g)uihe
v , π(g)vi dg = γπ (e
u ⊗ u) • (e
v ⊗ v) .
G/ZG

The term on the right is the canonical inner product defined earlier.
♥ [schurs-unitary] Proof. An immediate consequence of Lemma II.9.6.
This has more general consequences. Suppose now that R = D. Suppose (π, V ) to be any irreducible
admissible representation of G, with central character ω . Matrix coefficients define an equivariant embedding
e ⊗ π into C(G):
of π
Φv̂⊗v (g) = he
v , π(g)vi .
Dually, if f is any function in Hω−1 Then the integral
Z
(II.10.3) f (g)Φv̂⊗v (g) dg
[mc-int] G/ZG

defines a map from Hω−1 to the dual of π e ⊗ π , which is (canonically) π ⊗ π

e. This last space may be identified
with the linear operators in End(V ) of finite rank, and it is easy to see:
[endo-int] II.10.4. Lemma. The image of f with respect to the map from Hω−1 to End(V ) is the same as π(f ).
This image is in some sense the Fourier transform of f evaluated at the representation π . Of course this
depends on the particular realization of π . More canonically, one might choose the Fourier transform to be
in the image of π(f ), considered as an element of V ⊗ Ve , with respect to the matrix coefficient map. Directly
in terms of an element of End(V ) this is the function

trace (π(g −1 )f ) .
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 15

From this discussion follows the first part of:

[schurs-cuspidal] II.10.5. Proposition. If (π, U ) and (ρ, V ) are irreducible, compact representations of G with the same central
character, then Z
he
u, π(g)uihe
π (g)e
v , vidg
G

is equal to 0 if π and ρ are not isomorphic, and equal to

cπ he
v , uihe
u, vi

for some constant cπ 6= 0 if they are isomorphic.

Proof. It remains to prove that cπ 6= 0.
♥ [fg-field] This can be reduced to the unitary case. Because of Lemma II.1.7, the representation π may be defined over
a field with a countable number of elements. As is well known, it may then be embedded into C. The proof
that cπ 6= 0 depends on this, and in the rest of this section I’ll prove the theorem when R = C.
The formula remains the same if π and ρ are replaced by π·χ and ρ·χ, where χ is a character of G. According
♥ [assumption-cor] to Lemma II.10.1, we may therefore assume the comon central character to be unitary, and both become
square-integrable modulo ZG .
♥ [schur-unitary-2] We may now apply Proposition II.10.2, and we may assume that π = ρ. This gives us
Z
he
u, π(g)uihe
v , π(g)vi dg = γπ (e
u ⊗ u) • (e
v ⊗ v) .
G

e etc. What we want now is to transform this to

for all u
Z
he
u, π(g)uihe e wi dg = γπ he
π (g)w, v , wihe
u, wi .
G

For this, in order to get the right hand sides to look similar we must set ve = ι(w), v = ι−1 (w)
e . But then one
of the terms in the first integral becomes (since π is unitary)

π (g)ι(w), ι−1 (wi

he e = hι(π(g)w), ι−1 (wi
e .

♥ [vdotv-conj] But by (II.9.4) this is the same as

e π(g)wi .
hw,

It is good to keep in mind:

Suppose G compact with total measure 1, and let (π, V ) be an irreducible admissible
[degree-is-dim] II.10.6. Proposition.
representation. Then cπ = 1/ dim(π).
The representation is necessarily of finite dimension. For this reason, the positive constant 1/cπ is called the
formal degree of π .
♥ [schurs-cuspidal] Proof. Let (ei ) be a basis of V , ebj the dual. Apply Proposition II.10.5:
Z
hb ej , π(g −1 )ek i dg = cπ hb
ei , π(g)ej i hb ei , ek ihb
ej , ej i = cπ hb
ei , ek i .
G

for all i, j , k . Sum over j , set i = k . We are looking at a diagonal entry of the matrix product π(k)π(k −1 ).
We deduce that cπ dim(π) = 1.
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 16

♥ [cuspidals-projective] Now to conclude the proof of Proposition II.8.4.

v0 in Ve . Fix v0 in V such that

We are given an irreducible compact representation (π, V ), and a vector e
he
v0 , v0 i = dπ . Map f in Hω to
Pf = Rf ∨ v0 .
This is a G-equivariant map from Hω to V . We compute

Pγv = Rγv∨ v0
Z
= γv∨ (x)π(x)v0 dx
G/Z
Z
= hπ(x−1 )v, ve0 iπ(x)v0 dx .
G/Z

This last is an element of V . But according to Schur orthogonality

DZ E Z 1
−1
hπ(x )v, ve0 iπ(x)v0 dx, ve = hπ(x−1 )v, ve0 ihπ(x)v0 , vei dx = hv, veihv0 , e
v0 i .
G/Z G/Z dπ

11. Induced representations [induced.tex]

If H is a closed subgroup of G and (σ, U ) is a smooth representation of H , the unnormalized smooth

representation Ind(σ | H, G) induced by σ is the right regular representation of G on the space of all uniformly
smooth functions f : G → U such that
f (hg) = σ(h)f (g)
for all h in H , g in G. Let
δH\G = δH \δG .
The normalized induced representation is

−1/2 

Ind(σ | H, G) = Ind σδH\G  H, G .

1/2
Why the δ -factor? Well, Ind(δH\G ) is the space of smooth functions on H\G. The normalization is moti-
−1/2
♥ [one-densities] vated by Corollary I.7.7, which asserts that Ind(δH\G ) is isomorphic to that of smooth one-densities. The
symmmetry between δ ±1/2 suggests a useful duality.
The compactly supported induced representation Indc is on the analogous space of functions of compact
support on G modulo H .
[induced-admissible] II.11.1. Proposition. If H\G is compact and (σ, U ) admissible then Ind(σ | H, G) is an admissible represen-
tation of G.
The hypothesis holds when G is a reductive p-adic group and H a parabolic subgroup.
Proof. If H\G/K is the disjoint union of cosets HxK (for x in a finite set X ), then the map


f 7−→ f (x)

is a linear isomorphism

Indc (σ | H, G)K ∼
−1
[ind-restr-K] (II.11.2) . = ⊕x∈X U H∩xKx
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 17

Suppose (π, V ) to be a smooth representation of G, (σ, U ) one of H . The map

Λ: Ind(σ | H, G) → U

1/2 −1/2
taking f to f (1) is an H -morphism from Ind(σ) to σδH δG . If we are given a G-morphism from V to
1/2 −1/2
Ind(σ | H, G) then composition with Λ induces an H -morphism from V to σδH δG .
If π is a smooth representation of G and σ one of H then
[frobenius] II.11.3. Proposition. (Frobenius reciprocity)
evaluation at 1 induces a canonical isomorphism

1/2 −1/2

HomG π, Ind(σ | H, G) → HomH π, σδH δG .

−1/2
For F in Ind(eσ | H, G) and f in Indc (σ | H, G) the product hF (g), f (g)i lies in Indc (δH\G ). On this space, fix
a positive G-invariant integral Z
ϕ(x) dx .
H\G

[ind-duality] II.11.4. Proposition. The pairing Z

hF, f i = hF (x), f (x)i dx
H\G

defines an isomorphism of Ind(e

σ | H, G) with the smooth dual of Indc (σ | H, G).
In particular, if R = C and σ is unitary so is Ind(σ | H, G).

12. Appendix. Semi-simple algebra [induced.tex]

In this section I summarize relevant results from [Bourbaki:1958]. Suppose R to be any algebra over D. The
main item is this:
Suppose given a finite collection of irreducible,
[density] II.12.1. Theorem.
Q finite-dimensional R-modules Vi that are
pair-wise non-isomorphic. The canonical map from R to i EndD (Vi ) is surjective.
This is Corollaire 1 to Proposition 4 of §4.3 in [Bourbaki:2011]. It is (as we shall see) elementary, but it is not
easy to extract a succinct account from the literature. The proof I give here proceeds in several steps.
A semi-simple module over R is any direct sum of irreducible modules.
[semi-simple] II.12.2. Lemma. A finite-dimensional R-module V is semi-simple if and only if every submodule is a sum-
mand.
Proof. The only non-trivial thing to prove is that any submodule U is a summand. This will be shown by
induction on the codimension of U . If it is 0, there is nothing to prove.
L
Suppose V to be the sum Vi of irreducibles. If all the Vi are contained in U then V = U . Otherwise, say
V1 is not contained in U . Since it is irreducible, we must have V1 ∩ U = {0}. The projection
M
U −→ V /V1 = Vi
i6=1

is then injective. The codimension of the image of U has smaller codimension than that of U in V , so induction
U possesses an R-stable complement W in V /V1 . But W may be identified with a submodule in V , and
W ⊕ V1 is a complement in V .
The commutant R′ of R in EndD (V ) is the ring of operators commuting with it. The bicommutant R′′ is the
commutant of the commutant. It contains R.
 Chapter II. Smooth representations (9:17 p.m. June 16, 2019) 18

[bicommutant] II.12.3. Lemma. finite collection of irreducible R-modules Vi of finite dimension

Given a L Q that are pair-wise
non-isomorphic. If V = Vi , the canonical map from the bicommutant of R to i EndD (Vi ) is surjective.
Proof. This is because the commutant is the direct product of copies of D, one for each representation.
♥ [density] Theorem II.12.1 will now follow from:
[bicommutant-density] II.12.4. Lemma. Given any semi-simple R-module V of finite dimension, the image of the bicommutant in
EndD (V ) is the same as that of R.
Proof. Suppose (ei ) (for 1 ≤ i ≤ m) to be a basis of V . It must be shown that if ρ lies in the bicommutant,
there exists r in R such that r(ei ) = ρ(ei ) for all i.
♥ [semi-simple] The representation of R on W = V m is also semisimple. Let e = (ei ) be diagonally embedded. By Lemma
II.12.2 the submodule U = R·e is a direct summand of W . The projection from V onto U lies in the commutant
R′ of W , so the bicommutant R′′ takes W into itself. But this means that R′′ · e = R · e, which implies that
there for every ρ in R′′ there exists r in R such that r(e) = ρ(e).

13. References [induced.tex]

1. Joseph Bernstein, Representations of p-adic groups, preprint based on lectures at Harvard, 1992.
2. N. Bourbaki, Modules et anneaux semi-simples, Chapter 8 of Algèbres. Hermann, 1958.

3. William Casselman, Introduction to the theory of admissible representations of p-adic groups, preprint,
1974/1995.
4. ——, ‘Analysis on profinite groups’, preprint, 2018.
5. Hervé Jacquet, ‘Représentations des groupes linéaires p-adiques’, in Theory of group representations
and Fourier analysis (proceedings of a conference at Monecatini), C. I. M. E. Edizioni Cremonese, Rome,
1971.
6. —— and Robert P. Langlands, Automorphic forms on GL(2), Lecture Notes in Mathematics 114, Springer,
1970.