EXISTENCE AND UNIQUENESS OF HAAR MEASURE

JONATHAN GLEASON

Abstract. In this paper, we prove existence and uniqueness of left and right
Haar measures on a locally compact topological group, and show how one can
relate left and right Haar measure.

Contents
1. Introduction 1
2. Basic Definitions 2
3. Preliminary Results 3
4. Existence and Uniqueness 4
References 13

1. Introduction
The purpose of this paper is to prove existence and uniqueness of Haar measure
on locally compact groups. The paper is divided into four sections, including this
introduction.
The purpose of the second section is merely to present some basic definitions and
notation that will be used throughout the paper. Also included is brief motivation
and justification for some of the definitions. The reader may skip this section if
they prefer, and return to it later if needed. The reader should note, however (if
they decide to skip this section), that a locally compact topological group is, as a
topological space, both locally compact and T1 (and hence T31/2 ).
The third section contains statements and proofs of four lemmas. They have
been placed in a separate section because they are not immediately related to the
subject matter of the paper: topological groups and Haar measure. They are,
however, needed to complete the existence and uniqueness proofs, as well as the
theorem relating left and right Haar measures. They are placed in the order in
which they are used in Section 4. The reader may also skip this section, and return
to it later if interested in the details of a proof. No exposition is included in this
section at all.
The fourth section comprises the main body of the paper, and includes the
existence and uniqueness proofs, as well as a proposition relating left and right Haar
measures. In order to motivate only focusing on left Haar measure, we first show
that given a left Haar measure, one immediately obtains a right Haar measure, and
vice versa. We then provide a proof of the existence of left Haar measure on a locally
compact topological group. Then, after a couple of lemmas, we prove uniqueness

Date: June 24, 2011.

1
2 JONATHAN GLEASON

of left Haar measure on a locally compact topological group. We note here that, by
uniqueness, we mean that any two Haar measures on a locally compact topological
group are not exactly the same, but in fact only differ by a positive multiplicative
constant. We then briefly note how the relation between left and right Haar measure
immediately also implies existence and uniqueness of right Haar measure.

2. Basic Definitions
We first introduce some basic definitions and notation.
Notation 2.1. Throughout this paper, N = {0, 1, 2, 3, . . .}.
Notation 2.2. If X is a set, we shall denote the power set of X by 2X .
Notation 2.3. Let X be a set and let Σ ⊆ 2X . Then, we shall write σ[Σ] to denote
the σ-algebra generated by Σ.
Notation 2.4. If X is a topological case and A ⊆ X, then we shall denote the
interior of A by Ao and the closure of A by A.
Definition 2.5 (Borel Subset). Let X be a topological space with topology τ and
let A ⊆ X. Then, A is a Borel subset of X iff A ∈ σ[τ ].
Definition 2.6 (Topological Measure Space). A topological measure space is a
measure space (X, Σ, µ), where X is the space, Σ is the σ-algebra of measurable
subsets, and µ is the measure, such that X is a topological space and Σ is exactly
the collection of Borel subsets of X.
Definition 2.7 (Borel Measure). A measure µ on a topological measure space X
is called a Borel measure iff X is Hausdorff.
The reason we add the extra condition of T2 instead of doing things in complete
generality, is that, first of all, most spaces we care about in practice are going to be
Hausdorff anyways, and furthermore, we would like to know that compact subsets
are measurable (because in Hausdorff spaces compact subsets are closed), and in
general this won’t necessarily be the case.
Definition 2.8 (Regular Measure). Let (X, Σ, µ) be a Borel measure space. Then,
µ is said to be regular, or sometimes a regular Borel measure, iff
(1) Whenever K ⊆ X is compact, then µ(K) < ∞.
(2) Whenever A ∈ Σ, then1
µ(A) = inf {µ(U )|A ⊆ U, U is open.} .
(3) Whenever U ⊆ X is open, then2
µ(U ) = sup {µ(K)|K ⊆ U, K is compact.} .
Definition 2.9 (Locally Compact Group). A locally compact group is a topological
group G that is locally compact and T1 .

1This is sometimes referred to as outer regularity.

2This is sometimes referred to as inner regularity.
 EXISTENCE AND UNIQUENESS OF HAAR MEASURE 3

Here, we add in the extra condition of T1 because, one, it is a very weak assump-
tion, two, it is an assumption needed to prove the desired result (see Lemma 3.3 for
example), and three, by assuming just the fact that single points sets are closed,
we get T31/2 for free 3, although not necessarily T4 (for example, an uncountable
product of R)[3].
Definition 2.10 (Haar Measure). Let G be a topological group. A left Haar
measure (resp. right Haar measure) on G is a nonzero regular Borel measure µ on
G such that µ(gA) = µ(A) (resp. µ(Ag) = µ(A)) for all g ∈ G and all measurable
subsets A of G.

3. Preliminary Results
Lemma 3.1. Let f : X → Y and let E ⊆ 2Y . Then, σ f −1 (E) = f −1 (σ[E]).
 
 
Proof. Step 1: Show that σ f −1 (E) ⊆ f −1 (σ[E])
We shall show that f −1 (σ[E]) is a σ-algebra containing f −1 (E). Let A ∈
f (E). Then, there is some B ∈ E such that A = f −1 (B). Trivially, B ∈ E,
−1

so B ∈ σ[E], so A ∈ f −1 (σ[E]), so f −1 (E) is contained in f −1 (σ[E]). Now, we

wish to show that f −1 (σ[E]) is a sigma algebra, so let {An |n ∈ N} ⊆ f −1 (σ[E]).
Then,S for each An , there is some S Bn ∈ σ[E] suchSthat An = f −1 (Bn ). Now
B = n∈N Bn ∈ σ[E], so f −1 (B) = n∈N f −1 (Bn ) = n∈N An ∈ f −1 (σ[E]). Thus,
f −1 (σ[E]) is closed under countable union. Similarly, B0C ∈ σ[E], so f −1 (B0C ) =
f −1 (B0 )C = AC 0 ∈ f
−1
(σ[E]), so f −1 (σ[E]) is closed under complementation.
−1 −1 −1
Y ∈ σ[E], so f (Y ) =  X ∈−1f (σ[E]). Thus, f (σ[E]) is a σ-algebra containing
−1 −1
f (E), so σ f (E) ⊆ f (σ[E]).  
−1
Step 2: Show that  f (σ[E]) ⊆ σ f −1 (E) .
First, define Σ = A ⊆ Y |f −1 (A) ∈ σ f −1 (E) . We wish to show that Σ is
a σ-algebra containing E. Then, we will have shownthat σ[E]  ⊆ Σ. Let A ∈ E.
Of course, f −1 (A) ∈ f −1 (E), so trivially f −1 (A) ∈ σ f −1 (E) , and hence E ⊆ Σ.
Now, let {An |n ∈ N}
⊆ Σ. Then, f −1 (An ) = Bn for some  Bn ∈ σ  f −1 (E) . But
then, f −1 −1
(An ) = n∈N Bn ∈ σ f −1 (E)
S S S
n∈N An = n∈N f  , so Σis closed
under countable union. Now, f (AC −1
0) = f
−1
(A0 )C = B0C ∈ σ f −1 (E) , so Σ is
closed under complementation. And of course, f −1 (Y ) = X ∈ σ f −1 (E) , so Y ∈
Σ. Thus, Σ is a σ-algebra containing E, and so σ[E] ⊆ Σ. Now, let A ∈ f −1 (σ[E]).
−1
Then,
 −1 A =  f (B) −1 for some B ∈ σ[E]. But then, B ∈ Σ, so f−1 (B) = A ∈
σ f (E) . Thus, f (σ[E]) ⊆ σ f −1 (E) , and hence σ f −1 (E) = f −1 (σ[E]).

Lemma 3.2. Let (X, Σ, µ) be a topological measure space and let f : X → X be a
homeomorphism. Then, the following are equivalent:
(1) A ∈ Σ.
(2) f (A) ∈ Σ.
(3) f −1 (A) ∈ Σ.
Proof. Let τ be the topology on X. Then, by definition, Σ = σ[τ ].
((1) ⇒ (2)) Suppose A ∈ Σ. Then, by the above lemma,
f (A) ∈ f (Σ) = f (σ[τ ]) = σ [f (τ )] = σ[τ ] = Σ,

3This holds for a general topological group. The assumption of locally compact is not needed.
4 JONATHAN GLEASON

where we have used the fact that f (τ ) = τ because f is a homeomorphism.

((2) ⇒ (3)) Suppose f (A) ∈ Σ. Then, similarly as before,
A ∈ f −1 (Σ) = f −1 (σ[τ ]) = σ f −1 (τ ) = σ[τ ] = Σ.
 

Now, by (1) ⇒ (2) with the homeomorphism f −1 (instead of f as before), we have

that f −1 (A) ∈ Σ.
((3) ⇒ (1)) Suppose f −1 (A) ∈ Σ. Then, similarly as before,
A ∈ f (Σ) = f (σ[τ ]) = σ [f (τ )] = σ[τ ] = Σ.


Lemma 3.3. Let X be a Hausdorff space, let K be a compact subset of X, and let
U1 and U2 be open subsets of X such that K ⊆ U1 ∪ U2 . Then, there are compact
sets K1 and K2 of X such that K1 ⊆ U1 , K2 ⊆ U2 , and K = K1 ∪ K2 .
Proof. Define L1 = K − U1 and L2 = K − U2 . K is closed because X is Hausdorff,
so each Li is closed. Because each Li is a closed subspace of K and K is compact,
it follows that each Li is also compact. Furthermore, because K ⊆ U1 ∪ U2 , L1 ∩
L2 = ∅. Because L1 and L2 are disjoint compact subsets of a Hausdorff space,
we can separate them with disjoint open sets, say V1 and V2 respectively. Define
K1 = K − V1 and K2 = K − V2 . Similarly as before, both K1 and K2 are compact.
K1 = K − V1 ⊆ K − L1 = K − (K − U1 ) = K ∩ (K ∩ U1C )C = K ∩ (K C ∪ U1 ) ⊆ U1 .
Similarly, K2 ⊆ U2 . Furthermore, K1 ∪ K2 = K − (V1 ∩ V2 ) = K. 

Lemma 3.4. Let (X, µ) be a measure space, let f : X → R be measurable, and

let A ⊆ X be measurable. Then, if A = {x ∈ X|f (x) > 0} and µ(A) > 0, there is
some a > 0 such that µ ({x ∈ A|f (x) ≥ a}) > 0.
Proof. Suppose A = {x ∈ X|f (x) > 0} and µ(A) > 0. We proceed by contra-
diction: suppose that, for all a > 0, µ ({x ∈ A|f (x) ≥ a}) = 0. Write Sn =
x ∈ A|f (x) ≥ 21n . Then,

[
A= Sn ,
n∈N
so X
µ(A) ≤ µ(Sn ) = 0,
n∈N

so µ(A) = 0: a contradiction. Thus, there is some a > 0 such that

µ ({x ∈ A|f (x) ≥ a}) > 0.


4. Existence and Uniqueness

Before we prove anything about existence and uniqueness, we first show how to
obtain left Haar measure from right Haar measure, and vice versa.
Proposition 4.1. Let
G be a topological group, let µ be a Haar measure on G, and
define µ′ (A) = µ A−1 . Then, µ is a left (resp. right) Haar measure iff µ′ is a
right (resp. left) Haar measure on G.
 EXISTENCE AND UNIQUENESS OF HAAR MEASURE 5

Proof. (⇒) Suppose that µ is a left Haar measure on G.

Step 1: Show that µ′ is a Borel measure on G.
We first note that, because inversion is a homeomorphism of G, by Lemma 3.2,
µ′ is defined exactly on the Borel subsets of G. Trivially, µ′ is nonnegative and
µ′ (∅) = 0. Let {An |n ∈ N} be a collection
 of pairwise disjoint measurable subsets
of G. We would like to know that A−1 n |n ∈ N is also a collection of pairwise
disjoint measurable subsets. Once again, by Lemma 3.2, they are all measurable.
Suppose there is some x ∈ A−1 −1
m ∩ An for m 6= n. Then ,x = a
−1
= b−1 for some
 ∈−1Am and some b ∈ An , so that a = b ∈ Am ∩ An : a contradiction. Thus,
a
An |n ∈ N is a collection of pairwise disjoint measurable subsets, and hence
!  !−1  !
[ [ [ X
X ′
′ −1
µ An = µ  An =µ An = µ A−1n = µ (An ).
n∈N n∈N n∈N n∈N n∈N

Thus, µ′ is a Borel measure on G.

Step 2: Show that µ′ is regular.
Let K ⊆ G be compact. Then, K −1 is also compact, so
µ′ (K) = µ K −1 < ∞.

Let A be a measurable subset of G. A−1 ⊆ U and U is open
iff A ⊆ U

−1
and
−1 −1 −1
U is open, so that µ(U )|A ⊆ U, U is open. = µ U |A ⊆ U, U is open. .
Then,
µ′ (A) = µ A−1 = inf µ(U )|A−1 ⊆ U, U is open.



= inf µ U −1 |A ⊆ U, U is open. = inf {µ′ (U )|A ⊆ U, U is open.} .



Similarly, for A open,

µ′ (A) = sup {µ′ (K)|K ⊆ A, K is compact.} ,
and so µ′ is regular.
Step 3: Show that µ′ is a right Haar measure.
Trivially, µ′ is nonzero. Also,
µ′ (Ag) = µ (Ag)−1 = µ g −1 A−1 = µ A−1 = µ′ (A).

Thus, µ′ is a right Haar measure on G.

The other directions are essentially identical. 
This proposition tells us that, while left and right Haar measure on a group may
be different, they are related in a simple manner, and so we may as well simply
concern ourselves with the study of one or the other. Because of convention, we
shall restrict ourselves to proving existence and uniqueness of left Haar measure.
Before we prove existence of left Haar measure, however, one lemma is needed.
Lemma 4.2. Let G be a topological group, let K be a compact subset of G, and let
U be an open subset of G such that K ⊆ U . Then, there is an open set V containing
the identity such that KV ⊆ U .
Proof. For each x ∈ K, define Wx = x−1 U . Because x ∈ U , Wx is an open
neighborhood of the identity. Then, pick Vx to be an open neighborhood of the
identity such that Vx Vx ⊆ Wx . Then, the collection {xVx |x ∈ K} is anSopen cover
n
of K, so there is a finite collection of points x1 , . . . , xn such that K ⊆ k=1 xk Vxk .
6 JONATHAN GLEASON

Tn
Define V = k=1 Vxk . Let x ∈ K. Then, there is some xk such that x ∈ xk Vxk , so
that
xV ⊆ xk Vxk Vxk ⊆ xk Wxk = U.
Thus, KV ⊆ U . 

Theorem 4.3 (Existence). Let G be a locally compact group. Then, there exists a
left Haar measure on G.
Proof. Step 1: Define (K : V ).
Let K be a compact subset of G and let V be a subset of G with nonempty
interior. Then, {gV o |g ∈ G} is an open coverS of K, so there are a finite number
n
of elements of G, g1 , . . . , gn , such that K ⊆ k=1 gk V o . Let (K : V ) denote the
smallest nonnegative integer for which such a sequence exists.
Step 2: Define µU .
Let K denote the collection of compact subsets of G and let U denote the col-
lection of open subsets of G containing the identity. Because G is locally compact,
there is a compact subset of G with nonempty interior: call it K0 . For each U ∈ U,
define a function µU : K → R such that
(K : U )
µU (K) = .
(K0 : U )
Because K0 is nonempty, (K0 : U ) 6= 0, and so this is well-defined.
Step 3: Show that 0 ≤ µU (K) ≤ (K : K0 ).
As (K : U ) is always a nonnegative integer, µU is clearly always nonnegative.
We now show that (K : U ) ≤ (K : K0 )(K0 : U ) for K ∈ K and U ∈ U . For
the remainder of this paragraph, let us write m = (K : K0 ) and Sn = (K0 : U ).
m
let g1 , . . . , gm ∈ G and let h1 , . . . , hn ∈ G be such that K ⊆ k=1 gk K0o and
Then, S
n
K0 ⊆ k=1 hk U . Then,
 
[m [n
K⊆  g i hj U  ,
i=1 j=1

so that K can be covered by mn cosets of U , so that (K : U ) ≤ mn = (K : K0 )(K0 :

U ). It follows that
0 ≤ µU (K) ≤ (K : K0 ).
Step 4: Construct
Q the Haar measure on K.
Define X = K∈K [0, (K : K0 )]. Because 0 ≤ µU (K) ≤ (K : K0 ), each µU may
be thought of as a point in X. Thinking of each µU as a point in X, for each
V ∈ U , define C(V ) = {µU |U ∈ U, U ⊆ V }. We wish to show that the collection
{C(V )|V ∈TU} possess the finiteTintersection property, so let V1 , . . . , Vn ∈ U. Then,
n n
µ∩nk=1 Vk ∈ k=1 C(Vk ), so that k=1 C(Vk ) is nonempty. Thus, {C(V )|V ∈ U} sat-
isfies the finite intersectionTproperty, and because X is compact by Tychonoff’s
Theorem,
T it follows that V ∈U C(V ) is nonempty, so we may pick some µ ∈
V ∈U C(V ).
Step 5: Show that µ(K1 ) ≤ µ(K2 ) if K1 ⊆ K2 .
Let K1 , K2 ∈ K be such that K1 ⊆ K2 . We first show that, for each U ∈ U ,
µU (K1 ) ≤ µU (K2 ). But this is trivial, because the covering of K2 with (K2 : U )
cosets of U is also a covering of K1 with (K2 : U ) cosets of U , so that (K1 : U ) ≤
(K2 : U ), and hence µU (K1 ) ≤ µU (K2 ).
 EXISTENCE AND UNIQUENESS OF HAAR MEASURE 7

Thinking of elements f of X as functions from K to R, consider the map that

sends f ∈ X to f (K2 ) − f (K1 ). This is a composition of continuous functions, and
hence continuous.4. This map is also nonnegative on each C(V ) because µU (K1 ) ≤
µU (K2 ) for each U ∈ U (we need continuity so that we know it is nonnegative
on the entire closure.). It follows that this map is also nonnegative at µ, so that
µ(K2 ) − µ(K1 ) ≥ 0, so that µ(K1 ) ≤ µ(K2 ).
Step 6: Show that µ(K1 ∪ K2 ) ≤ µ(K1 ) + µ(K2 ).
Let K1 , K2 ∈ K. We first show that µU (K1 ∪ K2 ) ≤ µU (K1 ) + µU (K2 ) for each
U ∈ U . Thus this is trivial, because a covering of K1 with (K1 : U ) cosets of U
together with a covering of K2 with (K2 : U ) cosets of U , is a cover of K1 ∪ K2
with (K1 : U ) + (K2 : U ) cosets of U , so that (K1 ∪ K2 : U ) ≤ (K1 : U ) + (K2 : U ).
It follows that µU (K1 ∪ K2 ) ≤ µU (K1 ) + µU (K2 ).
Proceeding similarly as in step 5, the map that sends f ∈ X to f (K1 ) + f (K2 ) −
f (K1 ∪ K2 ) is continuous and nonnegative on each C(V ), and hence is nonnegative
for µ ∈ X. Thus, µ(K1 ∪ K2 ) ≤ µ(K1 ) + µ(K2 ).
Step 7: Show that µU (K1 ∪K2 ) = µU (K1 )+µU (K2 ) if K1 U −1 ∩K2 U −1 = ∅.
Let K1 , K2 ∈ K be such that K1 U −1 −1
Sn∩ K2 U = ∅. Let g1 , . . . , gn be such that
n = (K1 ∪ K2 : U ) and K1 ∪ K2 ⊆ k=1 gk U . If some gk U intersects both K1
and K2 , then gk ∈ K1 U −1 ∩ K2 U −1 : a contradiction. Thus, each gk U intersects
either K1 or K2 , but not both. Thus, we maySfind some natural number Sn m with
m
0 ≤ m ≤ n and reindex the gk s so that K1 ⊆ k=1 gk U and K2 ⊆ k=m+1 gk U .
Thus, (K1 : U )+(K2 : U ) ≤ (K1 ∪K2 : U ). Combining this result with the previous
step, it follows that µU (K1 ∪ K2 ) = µU (K1 ) + µU (K2 ) for each U ∈ U.
Step 8: Show that µ(K1 ∪ K2 ) = µ(K1 ) + µ(K2 ) if K1 ∩ K2 = ∅.
Let K1 , K2 ∈ K be such that K1 ∩ K2 = ∅. Then, we may find disjoint open
sets U1 and U2 such that K1 ⊆ U1 and K2 ⊆ U2 . By Lemma 4.2, there are open
neighborhoods of the identity V1 and V2 such that K1 V1 ⊆ U1 and K2 V2 ⊆ U2 .
Define V = V1 ∩ V2 . Then, K1 V and K2 V are disjoint because U1 and U2 are
disjoint. Thus, for any U ∈ U with U ⊆ V −1 , we have that K1 U −1 ∩K2 U −1 = ∅, so
that, by the previous step, µU (K1 ∪ K2 ) = µU (K1 ) + µU (K2 ). Thus, the continuous
map from X
to R that sends f ∈ X to f (K1 ) + f (K2 ) − f (K1 ∪ K2 ) is 0 for each
f ∈ S V −1 . In particular, µ(K1 ) + µ(K2 ) = µ(K1 ∪ K2 ).
Step 9: Extend µ to all subsets of G.
For U ⊆ G open, define
µ(U ) = sup {µ(K)|K ⊆ U, K ∈ K} ,
We must show that if K is compact and open, these two definitions of µ(K) agree.
That is, we must show that
µ(K) = sup {µ(K ′ )|K ′ ⊆ K, K ′ ∈ K} ,
T
where here the LHS is the original definition of µ as a point in U ∈U C(U ). Trivially,
since µ(K) ∈ {µ(K ′ )|K ′ ⊆ K, K ′ ∈ K}, µ(K) ≤ sup {µ(K ′ )|K ′ ⊆ K, K ′ ∈ K}. On
the other hand, by step 5, the set {µ(K ′ )|K ′ ⊆ K, K ′ ∈ K} is bounded above by
µ(K), so that sup {µ(K ′ )|K ′ ⊆ K, K ′ ∈ K} ≤ µ(K). Thus, this definition agrees
4The first map from X into R × R is the projection of f ∈ X onto the K th coordinate in the
1
first coordinate and the projection of f ∈ X onto the K2th coordinate in the second coordinate.
This map is continuous because it is continuous in each coordinate. Each coordinate is continuous
by definition of the product topology. This first map is followed by the map that subtracts the
second coordinate from the first, which is well-known to be a continuous map from R × R into R.
8 JONATHAN GLEASON

with the previous. It follows trivially that this extension still satisfies the property
µ(U1 ) ≤ µ(U2 ) if U1 ⊆ U2 .
Now, for an arbitrary subset A of G, define
µ(A) = inf {µ(U )|A ⊆ U, U is open.} .
Similarly as before, this indeed is an extension of our previous definition of µ to all
subsets of G. It again follows trivially that this extension still satisfies the property
that µ(A1 ) ≤ µ(A2 ) if A1 ⊆ A2 .
Step 10: Show that µ is an outer measure on G.
Trivially, µ(∅) = 0 because (∅ : U ) = 0 for every U ∈ U . Furthermore, to show
that µ is nonnegative, because of the definitions of the extensions, it suffices to
show that µ is nonnegative on K. For a fixed K, the map that sends f ∈ X to
f (K) is continuous (by similar reasoning as before). Furthermore, because this map
is nonnegative at each µU , it follows that this map is nonnegative on each C(V ).
Thus, this map is nonnegative at µ, so that µ(K) ≥ 0.
To show countable subadditivity, we first show that for each countable collection
of open sets {Un |n ∈ N}, we have that
!
[ X
µ Un ≤ µ(Un ).
n∈N n∈N

Let {Un |nS∈ N} be a countable collection

Sn of open subsets of G. Let K be a compact
subset of n∈N Un . Then, K ⊆ k=1 Uk for some n ∈ N. Applying Lemma 3.3
Sn
inductively, we may find compact sets K1 , . . . , Kn such that K = k=1 Kk and
Kk ⊆ Uk for 1 ≤ k ≤ n. Then, applying step 6 inductively,
Xn Xn X
µ(K) ≤ µ(Kk ) ≤ µ(Uk ) ≤ µ(Un ).
k=1 k=1 n∈N
It follows that
! ( )
[ [ X
µ Un = sup µ(K)|K ⊆ Un , K ∈ K ≤ µ(Un ).
n∈N n∈N n∈N
P
Now, let {An |n ∈ N} be an
arbitrary collection of subsets of G. If n∈N µ(An ) =
S P P
∞, then trivially µ n∈N An ≤ n∈N µ(An ), so suppose n∈N µ(An ) < ∞. Let
ε > 0, and for each n ∈ N, pick an open set Un such that An ⊆ Un and µ(Un ) ≤
µ(An ) + 2εn . Then,
! !
[ [ X X X 1 X ε
µ An ≤ µ Un ≤ µ(Un ) ≤ µ(An ) + ε n
= µ(An ) + ,
2 2
n∈N n∈N n∈N n∈N n∈N n∈N
but since ε > 0 was arbitrary, we have that
!
[ X
µ An ≤ µ(An ),
n∈N n∈N
so that µ is an outer measure on G.
Step 11: Show that the collection of Carathedory measurable sets
contain the Borel subsets of G.
To show that the collection of Carathedory measurable sets contain the Borel
subsets of G, it suffices to show that every open subset of G is measurable (because
the collection of measurable sets form a σ-algebra, if this collection contains the
 EXISTENCE AND UNIQUENESS OF HAAR MEASURE 9

topology of G, then it certainly contains the σ-algebra generated by the topology).

So let U ⊆ G be open and let A ⊆ G. If µ(A) = ∞, then trivially µ(A) ≥
µ(A ∩ U ) + µ(A ∩ U C ), so we might as well assume that µ(A) < ∞. Let ε > 0 and
pick V ⊆ G open and such that A ⊆ V and µ(V ) ≤ µ(A) + ε. Let K be a compact
subset of V ∩ U such that µ(V ∩ U ) − ε ≤ µ(K), and let L be a compact subset of
V ∩ K C such that µ(V ∩ K C ) − ε ≤ µ(L). Since K ⊆ U , V ∩ U C ⊆ V ∩ K C , so
µ(V ∩ U C ) − ε ≤ µ(V ∩ K C ) − ε ≤ µ(L).
Thus, by step 8,
µ(A ∩ U ) + µ(A ∩ U C ) − 2ε ≤ µ(V ∩ U ) + µ(V ∩ U C ) − 2ε ≤ µ(K) + µ(L)
= µ(K ∪ L) ≤ µ (V ∩ U ) ∪ (V ∩ K C )

≤ µ(V ) ≤ µ(A) + ε.
It follows that
µ(A ∩ U ) + µ(A ∩ U C ) ≤ µ(A) + 3ε.
Since ε is arbitrary, we have that
µ(A ∩ U ) + µ(A ∩ U C ) ≤ µ(A),
and hence U is measurable. It follows that µ restricts to a measure on the Borel
subsets of G, so that it is a Borel measure (G is completely regular, as mentioned
early, and in particular Hausdorff).
Step 12: Show that µ is regular.
Trivially, considering µ as an element of X, µ is finite on compact sets. Further-
more, as by construction µ(A) = inf {µ(U )|A ⊆ U, U is open.}, µ is trivially outer
regular. Similarly, µ is trivially inner regular (we showed that the extension agreed
with its definition for open sets which is by construction inner regular).
Step 13: Show that µ is nonzero.
µU (K0 ) = 1 for each U ∈ U , and the continuous function that maps f ∈ X to
f (K0 ) is a constant 1 on each C(U ), and in particular µ(K0 ) = 1, and hence µ is
nonzero.
Step 14: Show that µ is translation invariant.
Fix g ∈ G. The elements x1 , . . . , xn generate a cover for K iff the elements
gx1 , . . . , gxn generate a cover of gK, so that (K : U ) = (gK : U ) for each U ∈ U ,
and hence µU (K) = µU (gK) for each U ∈ U. It follows that the continuous function
that maps f ∈ X to f (K) − f (gK) is 0 on each C(U ), and hence µ(K) = µ(gK).
Thus, µ is translation invariant, and hence a left Haar measure on G. 
Before we dive into the proof of uniqueness, we first need to prove a couple of
lemmas about topological groups.
Lemma 4.4. Let G be a locally compact group and let f ∈ Cc (G). Then, for every
ε > 0, there is an open neighborhood U of the identity such that whenever y ∈ xU ,
it follows that |f (x) − f (y)| < ε.
Proof. Step 1: Construct the neighborhood.
Define K = supp[f ]. Let ε > 0. By continuity of f , for each x ∈ K, we may find
an open neighborhood Ux of the identity such that whenever y ∈ xUx , it follows
that |f (y) − f (x)| < ε. Then, for each x ∈ K, choose another open neighborhood
of the identity Vx such that Vx Vx ⊆SUx . By compactness of TK, there is a finite
n n
number of x1 , . . . , xn such that K ⊆ k=1 xk Vxk . Define V = k=1 Vxk and define
10 JONATHAN GLEASON

U = V ∩ V −1 . U is clearly an open neighborhood of the identity, and we claim that

this neighborhood works.
Step 2: Show that this neighborhood is a correct one.
Let y ∈ xU . If x, y ∈ / K, then |f (x) − f (y)| = 0, and so there is nothing to
worry about, so we may assume that either x ∈ K or y ∈ K. First suppose that
x ∈ K. Because x ∈ K, it follows that x ∈ xk Vxk for some 1 ≤ k ≤ n, and hence
that x ∈ xk Uxk . On the other hand, because x ∈ xk Vxk and V ⊆ Vxk , it follows
that y ∈ xV ⊆ xk Vxk Vxk ⊆ xk Uxk . Thus,

|f (x) − f (y)| ≤ |f (x) − f (xk )| + |f (xk ) − f (y)| < 2ε.

Now let us suppose that y ∈ K. y = xu for some u ∈ U , so x = yu−1 . But

U = V ∩ V −1 , so u−1 ∈ U , so that x ∈ yU . Then, we have that y ∈ K and x ∈ yU ,
so we may apply the same logic as in the previous paragraph (interchanging the
roles of x and y). 

Lemma 4.5. Let G beR a topological group

R and let µ be a Haar measure on G.
Then, for every x ∈ G, G f (xg)dµ(g) = G f (g)dµ(g) for f ∈ L1 (G).

Proof. Let x ∈ G.
Step 1: Prove for characteristics functions.
Let A be measurable and let f = χA . Then,
Z Z Z
χx−1 A (g)dµ(g) = µ x−1 A = µ(A)


f (xg)dµ(g) = χA (xg)dµ(g) =
G
ZG Z G
= χA (g)dµ(g) = f (g)dµ(g).
G G

Step 2: Prove for simple functions. Pn

Let f be a simple function. Then, f = k=1 ak χAK for some constants a1 , . . . , an
and some measurable sets A1 , . . . , An . Then,

Z n
X Z n
X Z
f (xg)dµ(g) = ak χAk (xg)dµ(g) = ak χAk (g)dµ(g)
G k=1 G k=1 G
Z
= f (xg)dµ(g).
G

Step 3: Prove for f nonnegative measurable.

Let f be a nonnegative measurable function on G. Then, there exists a mono-
tonic increasing sequence of simple functions φn that converges pointwise almost
everywhere to f . Thus, by the Monotone Convergence Theorem,
Z Z Z Z
f (xg)dµ(g) = lim φn (xg)dµ(g) = lim φn (g)dµ(g) = f (g)dµ(g).
G G G G

Step 4: Prove for real-valued integrable functions.

 EXISTENCE AND UNIQUENESS OF HAAR MEASURE 11

(
f (g) if f (g) ≥ 0
Let f be a real-valued integrable function on G. Define f+ (g) =
0 otherwise
(
−f (g) if f (g) ≤ 0
and f− (g) = . Then, f+ and f− are both nonnegative mea-
0 otherwise
surable functions, so
Z Z Z
f (xg)dµ(g) = f+ (xg)dµ(g) − f− (xg)dµ(g)
G
ZG Z G Z
= f+ (g)dµ(g) − f− (g)dµ(g) = f (g)dµ(g).
G G G
Step 5: Prove for f ∈ L1 (G).
Let f ∈ L1 (G). Define R = ℜ[f ] and I = ℑ[f ]. Then, R and I are real-valued
integrable functions, so
Z Z Z
f (xg)dµ(g) = R(xg)dµ(g) + i I(xg)dµ(g)
G G G
Z Z Z
= R(g)dµ(g) + i I(g)dµ(G) = f (g)dµ(g).
G G G

Theorem 4.6 (Uniqueness). Let G be a locally compact group, and let µ and µ′ be
two left Haar measures on G. Then, µ = aµ′ for some a ∈ R+ .
Proof. Step 1: Find a compact set of nonzero measure.
Because µ is nonzero, there is some set of nonzero measure (with respect to µ).
It follows by outer regularity that there is some open set (containing this set) that
is also of positive measure, and by inner regularity, it follows that there s a compact
set of nonzero measure (contained
R in this open set): call it K.
Step 2: Show that G f dµ > 0 for f ∈ Cc (G) nonnegative and not
identically 0.
Let f ∈ Cc (G) be nonnegative and not identically 0. Define U = f −1 (R+ ). U is
nonempty because f is not identically 0. By continuity, U is open, so because K
is compactSand U is nonempty, there is a finite number of elements g1 , . . . , gn such
n
that K ⊆ k=1 gk U , so that
n
X
0 < µ(K) ≤ µ(gk U ) = nµ(U ),
k=1
so that µ(U ) > 0. Then, by Lemma 3.4, it follows that there is some a > 0 such
that V = {g ∈ G|f (g) ≥ a} is of positive measure. It follows that
Z Z
f dµ ≥ f dµ ≥ aµ(V ) > 0.
G V
Step 3: Define h.
Let g ∈ Cc (G) be nonnegative and not identically 0, and let f ∈ Cc (G) be
arbitrary. g will remain the same throughout the remainder of the proof. Define
f (x)g(yx)
h(x, y) = R .
G
g(tx)dµ′ (t)
By step 2, the denominator never vanishes, and so h is well-defined on all of G × G.
Trivially, h is compactly supported because both f and g are.
12 JONATHAN GLEASON

Step 4: Show that h is continuous.

To show that h is continuous, it suffices to show that I(x) ≡ G g(tx)dµ′ (t) is
R

a continuous function. Define K = supp[g], let x0 ∈ G, and let U be an open

neighborhood of x0 whose closure is compact (which exists because G is locally
−1 −1
compact). K × U is compact by Tychonoff’s Theorem, so KU is compact
−1
because this is the image of K × U under a continuous function. Let ε > 0, and
−1 −1
choose δ > 0 so that δµ′ KU < ε, which we may do because KU is compact,
and hence of finite measure. By Lemma 4.4, there is an open neighborhood V of
the identity such that whenever y ∈ xV , it follows that |g(x) − g(y)| < δ.
Then, whenever x ∈ U ∩ x0 V , an open neighborhood of x0 , tx ∈ tx0 V , so that
Z  
−1
|I(x) − I(x0 )| ≤ |g(tx) − g(tx0 )| dµ′ (t) ≤ δµ′ KU < ε,
G
−1
where we have used the fact that integrand vanishes for t outside of KU . Thus,
I is continuous, and henceRh is continuous, and hence h ∈ Cc (G × G).
f (x)dµ(x)
Step 5: Show that RG g(x)dµ(x) = C, where C is some constant inde-
G
pendent of µ.
By a generalization of Fubini’s Theorem5, we have that
Z Z  Z Z 
′
h(x, y)dµ (y) dµ(x) = h(x, y)dµ(x) dµ′ (y)
G G G G
Z Z 
h y x, y dµ(x) dµ′ (y)
−1


=
G G
Z Z 
−1

′
= h y x, y dµ (y) dµ(x)
G G
Z Z 
−1

′
= h y , xy dµ (y) dµ(x),
G G

where we have applied Lemma 4.5 several times. Thus,

R
g(yx)dµ′ (y)
Z Z  
G
f (x)dµ(x) = f (x) R dµ(x)
G G G
g(tx)dµ′ (t)
Z Z  Z Z 
f (x)g(yx) ′ ′
= R
′
dµ (y) dµ(x) = h(x, y)dµ (y) dµ(x)
G G G g(tx)dµ (t) G G
Z Z 
h y −1 , xy dµ′ (y) dµ(x)


=
G G
Z "Z
#
f y −1 g(x) ′
= R
−1 ) dt
dµ (y) dµ(x)
G G G g (ty

!
f y −1
Z  Z
′
= g(x)dµ(x) R
−1 ) dµ′ (t)
dµ (y)
G G G g (ty
R
f (x)dµ(x)
Thus, RG = C, where C is some constant independent of µ.
G
g(x)dµ(x)
Step 6: Deduce that G f dµ′ = a G f dµ for some positive constant a.
R R

5See [1], pg. 243–244. Note that this is why we needed h ∈ C (G × G).
c
 EXISTENCE AND UNIQUENESS OF HAAR MEASURE 13

Because this constant does not depend on µ, it must be the case that
R R
f dµ f dµ′
RG = C = RG ,
G
gdµ G
gdµ′
and hence that Z Z
f dµ′ = a f dµ,
R G G
gdµ′
where a ≡ RG .
G
gdµ
Step 7: Show that µ′ = aµ.R R
For f ∈ Cc (G), define φ(f ) = G f dµ and ψ(f ) = G f dν, where ν is a measure
defined by ν = 1/aµ′ . Both φ and ψ are positive linear functions on Cc (G), and
Z Z Z
′
φ(f ) = f dµ = 1/a f dµ = f dν = ψ(f ).
G G G
6
Thus, by the Riesz Representation Theorem , it follows that µ = ν, i.e. that
µ′ = aµ with a ∈ R+ . 
This theorem tells us that left Haar measure on G is “essentially” unique, in
the sense that any two left Haar measures differ only by a positive multiplicative
constant. Furthermore, if we add the requirement that a certain fixed subset has a
specified measure, this completely determines the measure on the group.
This concludes the proof of the existence and uniqueness of left Haar measure on
a locally compact topological group. Given existence and uniqueness of left Haar
measure, the corresponding results for right Haar measure follow immediately from
Proposition 4.5.

References
[1] D. L. Cohn, Measure Theory, (2009), 199,206,211,243–244,297–311.
[2] G. B. Folland, Abstract Harmonic Analysis, (1995), 36.
[3] J. R. Munkres, Topology, 2nd Ed., (2000), 213.
[4] E. M. Stein; R. Shakarchi, Real Analysis: Measure Theory, Integration, & Hilbert Spaces,
(2007), 274.

6See [1], pg. 209–210.